Perspectives on Various Issues of Interest

Aiding Surgical Outcomes with AI and Machine Learning

At its 2022 annual meeting, the Surgical Outcomes Club—a leading consortium of surgeons and health services researchers dedicated to advancing surgical outcomes science—hosted a panel of four experts to discuss the growing role of predictive analytics and artificial intelligence (AI) in surgical research. The discussion centered on three core domains where AI is poised to make a significant impact: computer vision, digital transformation at the point of care, and the utilization of electronic health records (EHR) data. The panel addressed both the opportunities and inherent challenges associated with integrating AI into surgical practice.

Computer Vision: Giving Machines Eyes in the OR

The increasing capture of surgical video—routinely generated during minimally invasive and robotic procedures, and now expanding into open surgeries—offers a new frontier for AI. Real-time video annotation powered by computer vision can help evaluate surgical performance, identify complex anatomy, and provide intraoperative feedback to mitigate technical errors. Beyond performance assessment, this technology holds promise for surgical education, allowing skills training and behavior review through tool and hand tracking, and phase annotation.

With the advent of convolutional neural networks and other advanced models, video-based AI tools can now approach the visual complexity of surgery. These innovations can enhance surgeon training and potentially assist with decision-making during procedures. However, real-world implementation remains limited by several barriers, including the complexity of surgical environments, insufficient generalizability of current models, and a lack of large, annotated, and diverse datasets. Data sharing limitations and institutional barriers further complicate the creation of robust open-source datasets.

While current efforts rely on public video sources of inconsistent quality, a recent consensus suggests retrospective training tools may be feasible within two years, and real-time applications may emerge within the next decade. Recognizing the early-stage nature of this field is critical to fostering collaboration between surgeons and engineers, ensuring that AI tools ultimately support, rather than replace, surgical expertise.

Building Surgical Intelligence Through Video-Based Analytics

Video analytics offer the potential to assist surgeons during operations by identifying key anatomical landmarks, outlining tumor margins, or analyzing instrument usage patterns. Particularly in rare or unexpected events—such as intraoperative bleeding—AI could provide scenario-based recommendations to guide next steps. Though not yet deployed in operating rooms, emerging research outlines how real-time decision support systems could soon become a reality.

Leveraging Data for Surgical Innovation

Despite the influx of new surgical devices, there's limited insight into how they compare to existing techniques. Video analysis can quantify how new technologies influence surgical workflow and learning curves, particularly during the adoption of minimally invasive or robotic procedures. Side-by-side comparisons of similar cases can inform best practices and highlight improvements or setbacks introduced by novel tools.

Surgical video provides “ground truth” data, offering unparalleled insights into intraoperative behavior. Beyond enhancing individual performance, these data can serve broader purposes—from reducing OR inefficiencies to defending medical decisions in legal contexts, and informing medical device development. However, building the necessary infrastructure for data processing and analytics requires technical expertise far beyond traditional clinical training. Successful implementation hinges on interdisciplinary collaboration between clinicians and data scientists.

Addressing the Complexities of Surgical Video Analysis

While promising, the use of surgical video raises significant legal, ethical, and logistical questions. Ownership of the footage remains unclear, and concerns about patient privacy, staff exposure, and potential misuse can discourage open sharing—especially in complex or unfavorable cases. Nonetheless, video evidence can also serve as proof of adherence to standard care protocols.

Other concerns include potential conflicts of interest, data security, and a disconnect between those who generate the data (surgeons) and those capable of analyzing it (engineers). Despite these challenges, carefully selected use cases and clear goals can help bridge these divides.

Calls to Action: Accelerating AI Integration in Neurosurgery

The increasing public awareness of AI offers a strategic moment to integrate it meaningfully into neurosurgical practice. Four key actions are proposed:

1. Establish AI Task Forces: Professional societies such as the American Association of Neurological Surgeons (AANS) and the Congress of Neurological Surgeons (CNS) should form joint task forces to define best practices, set data standards, and facilitate clinician-scientist collaboration. Subspecialty task forces should address domain-specific use cases, backed by dedicated research funding and aligned with interdisciplinary partners across related surgical and technical fields.

2. Create Multi-Institutional Research Organizations: Single-institution efforts lack the scale and diversity needed to train robust AI models. Instead, we should foster independent, multi-institutional research entities—either nonprofit or for-profit—that can secure funding, manage cross-institutional data, and develop reusable tools for ML integration.

3. Launch Conferences and Challenge Frameworks: There is a need for clinician-led conferences and grand challenges to define and advance AI use cases in surgery. Inspired by the “Common Task Framework” (CTF) model, such initiatives can attract diverse collaborators and reward clinically meaningful innovation. Dedicated tracks within surgical and technical conferences can help bridge the divide between these communities.

4. Standardize Data Capture and Sharing: Video is just one of many valuable data streams in the modern OR. Integrating and standardizing these streams for AI use remains a challenge due to technical and regulatory hurdles. Collaborative efforts between surgical and technical communities, supported by new regulatory frameworks, can unlock the potential of OR data for clinical improvement.

By aligning clinical insight with technical innovation, the surgical community can unlock the transformative potential of AI. Through multidisciplinary efforts, structured collaboration, and a shared vision, we can bring next-generation surgical analytics from concept to clinical reality—benefiting patients, providers, and the entire healthcare ecosystem.

As an example, a study developed an AI-ready dataset for model training by programmatically querying open surgical procedures on YouTube, selecting and manually annotating a subset of videos. This dataset was used to train a multitask AI model, subsequently applied in two proof-of-concept studies: (1) to generate “surgical signatures” that characterize procedural patterns, and (2) to identify hand motion kinematics indicative of surgeon experience and skill level.

The resulting Annotated Videos of Open Surgery (AVOS) dataset comprises 1,997 videos spanning 23 procedure types, sourced from 50 countries over a 15-year period. To test real-world applicability, additional deidentified surgical videos were prospectively collected from a tertiary academic medical center (Beth Israel Deaconess Medical Center [BIDMC]), with IRB approval and patient consent.

Multitask Model Architecture and Training

A multitask neural network was trained on the AVOS dataset to perform spatiotemporal analysis of hands, tools, and actions in surgical video. The model captured procedural flow and fine motor behaviors in near real time, enabling simultaneous analysis across multiple tasks. To improve generalizability across varied operative conditions, data augmentation techniques—including flipping, scaling, rotation, and occlusion testing—were applied during training. An alternating task training strategy was used to optimize both spatial and temporal branches, with a dedicated training stream for hand-pose estimation.

Inference was performed by extracting batches of four frames at five-second intervals from each video. Background actions were filtered to ensure consistent comparisons across procedures.

Proof-of-Concept: Generating Surgical Signatures

The model was tested on previously unseen videos of appendectomies, pilonidal cystectomies, and thyroidectomies—procedures well-represented in the AVOS dataset. These videos were manually reviewed to confirm the presence of key operative steps, with durations ranging from 2 to 30 minutes. Using temporal averaging of model outputs, distinct surgical signatures were generated for each procedure, reflecting expected progressions in tool use and action (e.g., from cutting to suturing).

These signatures serve as procedural benchmarks, and significant deviations from them may reflect disruptions in surgical flow, variations in technique, or complexity in a given case. This functionality offers the potential for early detection of surgical anomalies or challenges requiring expert intervention.

Proof-of-Concept: Quantifying Surgical Skill

To assess skill, the model was retrospectively applied to 101 prospectively collected surgical videos at BIDMC, including live procedures and simulated wound closures. Participants included 14 operators categorized as either trainees (medical students, residents) or experienced surgeons (fellows, attendings). Hand movements were tracked using bounding boxes and nine anatomical key points (thumb, index finger, and palm).

Kinematic metrics—including velocity, rotation, and translation—were extracted and summarized into a single compound skill score using principal component analysis. Logistic regression analysis showed this compound feature significantly predicted surgeon experience, with each unit increase associated with a 3.6-fold increase in odds of being an experienced surgeon (95% CI: 1.67–7.62; p = 0.001).

Implications for Surgical Education and AI-Augmented Assessment

The multitask model demonstrated procedure-agnostic capabilities, performing reliably across variable video conditions such as lighting and camera angles. The ability to analyze both procedural flow and individual surgeon behavior marks a major advance toward automated, objective surgical feedback.

By linking hand motion patterns to surgical expertise, the model offers actionable insights for training. For instance, AI-driven feedback on motion economy and steadiness could allow trainees to iteratively improve performance, aligning with best practices observed in expert surgeons. This scalable, unbiased approach to skill assessment may facilitate faster and more reliable surgical training, especially in simulation-based environments.

Sources:

https://doi.org/10.1001/jamasurg.2022.5444

https://doi.org/10.3171/2023.4.JNS23640

https://doi:10.1001/jamasurg.2023.6262

Unsupervised NLP Methods for Analyzing User-Provided Data in Smart Product Service Systems

Product-Service Systems (PSS) are business models that go beyond delivering physical products by also offering complementary intangible services. With the rise of smart technologies and connected devices, these models have evolved into Smart Product-Service Systems (Smart PSS), enabling service providers to deliver personalized and data-driven offerings. By leveraging user-generated data from smart products, providers can tailor services to meet individual customer needs more effectively.

To harness the potential of smart devices in enhancing existing services or developing new ones, a structured Smart PSS design methodology is essential. Such a method enables enterprises and service providers to build personalized Smart PSS solutions using user-generated data, or to enhance current offerings through the application of deep learning techniques.

As a first step, service providers should adopt the customer's perspective to identify the root causes of any inconvenience experienced during service use. To support this, the use of a customer journey map is recommended as a tool to analyze and understand the customer’s mental and emotional experience throughout their interaction with a product or service.

A customer journey map is a visual method that outlines the steps a customer takes, combined with their subjective feelings at each stage. By illustrating the entire service experience through graphics and process flows, it helps organizations see their services through the eyes of the customer. This perspective enables companies to pinpoint weaknesses in the service process and use these insights as a foundation for improvement.

Furthermore, as many customer interactions today are closely tied to data collection, traditional customer journey maps may need to be adapted to incorporate data-driven insights that better align with the evolving needs of modern service providers. A modified customer journey map is needed to assist service providers in analyzing the customer experience, as illustrated in the following figure.

Two key enhancements include: the inclusion of “needed data” at each stage of the journey, and the addition of emotional states alongside their root causes, forming the basis of an emotional journey.

This modified approach is explained step by step in the following sections:

1. Service providers should place themselves in the role of a customer and break down the service process into distinct stages, reflecting how the customer experiences the journey.

2. Identify the key activities and touchpoints customers engage with at each stage. These are critical interactions that shape the overall experience and must be included in the journey map.

3. Consider the emotions customers might experience during each activity. Providers should empathize with the user to understand these feelings. Based on this, an emotional journey can be drawn to represent the fluctuations in customer sentiment. In this version, both the emotional state and its underlying causes are mapped, providing deeper insight into customer behavior.

4. Recognize the data customers rely on to complete key activities. This “needed data” should be clearly identified to help service providers understand what information supports each step of the journey.

5. Finally, service providers should reflect on potential problems or pain points within the process and explore corresponding solutions to enhance the customer experience.

The modified customer journey map is also applied to uncover user requirements across various industries, recognizing that different sectors may face unique service-related pain points. However, certain factors—such as system usability, user interface (UI) design, and pricing—remain consistently influential in shaping customer satisfaction across industries.

Therefore, when designing new services, providers must consider both industry-specific challenges and universal customer expectations.

In summary, the modified customer journey map allows service providers to gain a comprehensive understanding of the user experience, including the emotions associated with each step of the service process. By capturing these emotional insights, providers can more easily identify potential issues and opportunities for improvement, enabling them to make targeted adjustments that better align with customer needs and expectations.

After identifying user requirements, service providers must determine which pain points they aim to address or how to adapt existing services to better fulfill those needs. Once the specific challenges are defined, the next step is to gather the necessary data to support service redesign or improvement.

Data can be sourced from a variety of channels, including social media posts and replies, instant customer feedback, and open datasets provided by governments or academic institutions. Thanks to the widespread availability of online data, much of this information is easily accessible, well-organized, and often available at low cost.

However, publicly available data may not always provide the depth or specificity required for targeted service enhancements. To overcome this limitation, service providers are encouraged to collect their own data tailored to their objectives. For instance, one might scrape restaurant reviews from platforms like Google Maps to gain insights into customer sentiment and expectations.

The benefit of collecting data independently is that it can be customized to meet specific demands. However, this approach can also be time-consuming and resource-intensive, requiring careful consideration of the trade-offs between data quality and operational cost.

To effectively meet diverse customer needs, service providers must develop appropriate models that align with the data they’ve collected. Before selecting a suitable model, it is essential to clearly define the target problem and identify the type of data to be used. However, traditional machine learning models often struggle with unstructured data—such as text, images, and audio—due to their lack of inherent structure.

Focusing on text data one could demonstrate the advantages of integrating smart Product-Service Systems (Smart PSS) with deep learning techniques. Specifically, we utilize the Doc2Vec (Document-to-Vector) model, a deep learning method that extends the Word2Vec approach, and implement it using the Gensim® library.

Doc2Vec is an unsupervised learning algorithm designed to convert entire texts into fixed-length vector representations. It learns these representations by predicting words based on their surrounding context, leveraging the idea that the contextual relationships in a sentence can reveal its semantic structure. During training, sections of sentences are masked (hollowed out), and the model learns to predict these missing words using the surrounding context—without requiring labeled data.

Additionally, Doc2Vec incorporates a paragraph matrix that serves as a unique identifier for each document, capturing global information and maintaining coherence across different parts of the same text. This enables the model to retain not just statistical, but also semantic and contextual meaning, even across documents of varying lengths.

Compared to traditional techniques like TF-IDF, which rely heavily on frequency counts and disregard word order or meaning, Doc2Vec provides a richer representation of text. Unlike Word2Vec, which computes document vectors as averages of individual word vectors, Doc2Vec takes into account the structure and flow of entire paragraphs, making it more effective in capturing the topic or theme of a document.

Once trained, the resulting document vectors can be used for various tasks such as document classification, clustering, and similarity comparison. A commonly used technique for comparing these vectors is Cosine Similarity, which measures the angle between two non-zero vectors, offering a metric for semantic similarity between texts.

The final step focuses on encouraging service providers to develop a complete and practical solution tailored to customer needs. These solutions can take various forms—web platforms, mobile applications (such as Android apps), and other digital interfaces are commonly used to deliver services effectively.

For instance, the figure below illustrates the workflow of a smart Product-Service System (Smart PSS) designed for tourist recommendations, implemented by a network of independent taxi operators.

The system employs an Android application as the front-end interface for the proposed Smart PSS, while the back-end is powered by a Python server integrating a Doc2Vec model. The interaction process unfolds as follows:

1. In the first step (arrow 1), users input three types of information via the app: their current location, preferred travel distance, and a text description of personal preferences related to attractions and restaurants.

2. This information is then transmitted to a PHP and Python server (arrows 2, 3, and 4), where it is used as parameters to compute similarities with surrounding attraction data (arrow 5).

3. The top 5 most relevant results, based on similarity calculations, are sent back to the app (via arrows 6 and 7), and finally, the results are displayed to the user (arrow 8).

To enhance user experience, the recommendation system is also integrated with the Google Directions API to offer navigation services. It generates a personalized full-day itinerary, including suggestions for morning and afternoon attractions, as well as lunch and dinner spots.

The detailed workflow is as follows:

• Step 1: Users define a preferred travel radius and select types of attractions they are interested in.

• Step 2: They enter a brief text describing their preferences (e.g., “quiet places with scenic views”).

• Step 3: If users choose to skip planning for any part of the day (morning, lunch, afternoon, or dinner), they can do so.

• Step 4: The application uses the user’s current location to set the search radius.

• Step 5: For each attraction within that radius, the system retrieves the top five user reviews from Google Maps, then converts both the user’s input and the review texts into 50-dimensional vectors using the Doc2Vec model.

• Step 6: Using cosine similarity, the system compares the user's preferences to each attraction's reviews and ranks the nearby options accordingly.

Each recommended attraction includes details such as the name, user rating, estimated travel time, and distance.

From the ranked list, the system presents the top 10 recommended attractions, allowing users to make selections for each time period. Once selections are made, a custom day plan is generated and route-planned using the Google Maps API, which defaults to the shortest driving route. To support diverse travel preferences, the app also includes a transportation mode selector, enabling users to switch between walking, public transport, biking, or driving.

Source: https://doi.org/10.1016/j.compind.2021.103421

 

Features, Business Process Models and Architectural Design of Smart Warehouses in Industry 4.0

A traditional warehouse typically includes several core functions such as receiving, storing, tracking and tracing, picking, and shipping. It also involves interactions with both upstream and downstream stakeholders, as well as a centralized management system. While a planning function is recommended, it is often considered optional in traditional warehousing setups. When goods arrive at a warehouse, they need to be tagged. Tagging can occur at multiple levels—truck, pallet, tray, or individual product—but pallet and tray tagging are the most commonly used due to their efficiency and ability to distinguish between different product types. Once tagged, goods are stored either manually or through automated systems like conveyor belts. If the incoming products are hazardous, compatibility checks are performed before storage to ensure safe handling.

In traditional warehouses, tracking is usually done manually using handheld scanners. The planning department plays a key role in determining the order fulfillment schedule and allocating the necessary resources. After planning, the order-picking process begins, which is also typically manual. Orders can be picked by entire trays or pallets, but often require manual assembly of different items to fulfill custom orders. Once picked, orders are packed and shipped to the customers. The warehouse supply chain involves both upstream and downstream stakeholders. Upstream stakeholders, typically suppliers, are responsible for delivering goods to the warehouse. Downstream stakeholders, usually customers, place orders that the warehouse fulfills. In some cases, the same entity may serve as both an upstream and downstream stakeholder—for example, manufacturers storing surplus products in warehouses. Clear communication with stakeholders is crucial, particularly when notifying them about order readiness, inventory levels, or any disruptions in the delivery process.

Effective warehouse operations also require a robust management system that covers finance and accounting, data processing, and sales management. Sales management focuses on handling orders and inventory, ensuring supply meets demand. Finance and accounting oversee the warehouse's financial health and ensure operational continuity. Data processing manages information from scanners and other sources, working closely with sales management to maintain accurate stock levels and support efficient supply intake.

While smart warehouses share many of the same high-level components as traditional warehouses, smart warehouses make planning a mandatory component and incorporate a warehouse communication network for enhanced coordination and efficiency. Each feature in the top layer has various optional and obligatory elements.

A smart warehouse functions through the seamless communication and integration of multiple systems. Within its business process management (BPM) structure, various technologies support the overarching warehouse management system (WMS). Advanced Planning and Scheduling (APS) software is used to oversee the planning and scheduling operations, ensuring that resources are optimally allocated. Simultaneously, an inventory management system is employed to maintain optimal stock levels and streamline inventory control. Financial and sales management modules handle all monetary interactions—ranging from processing incoming orders to managing restocking events—ensuring smooth financial operations throughout the warehouse. Order picking is directed by an Automated Storage and Retrieval System (AS/RS), which interfaces with Automated Guided Vehicles (AGVs) and other material-handling equipment to carry out tasks efficiently. Furthermore, a Transport Management System (TMS) is integrated into the process, coordinating with AGVs to prepare and load shipments onto trucks.

The figure outlines three key roles in the system—represented as swimlanes: the Supplier/Client, the Warehouse Management System (WMS), and the Warehouse. The client initiates a request, which is processed by the supplier. The WMS manages the planning and coordination of that request, while the Warehouse role executes the operational tasks. Each role consists of specific actions that may trigger related tasks across the system.

Architecture design of smart warehouses need to consider various viewpoints. In the following context diagram for a smart warehouse, the Warehouse Management System (WMS) serves as the central hub, orchestrating the core operations of receiving, storing, picking, and shipping goods. The WMS interfaces with a range of human operators, including corporate supervisors, warehouse managers, and floor-level employees. These users interact with the system to log actions, monitor activities, make operational decisions, and extract performance reports from the collected data. Depending on warehouse policy, truck drivers may also interact with the WMS as external operators.

To function effectively, the WMS depends on real-time data inputs gathered through scanners and sensors. These devices are often mounted on Automated Guided Vehicles (AGVs) or embedded within Augmented Reality (AR) systems. The Transport Management System (TMS) uses this data to guide AGVs and employees—via AR devices—to specific locations within the warehouse. This operational flow is further supported by the Advanced Planning and Scheduling (APS) system, which determines which goods should be retrieved. In some configurations, an Order Picking Operation System (OPOS) is also integrated to enhance the efficiency of picking tasks, and shelves may be outfitted with RFID tags to enable adaptive, self-adjusting storage.

To further enhance warehouse performance, Multi-Agent System (MAS) methodologies can be implemented. In such setups, robotic agents operate collaboratively within a distributed architecture. Intelligent agent-based communication allows for decentralized task allocation, optimizing efficiency through distributed algorithms. This approach helps to reduce battery usage and latency while maximizing utilization through task decomposition techniques.

The finance/accounting and sales management components manage all monetary and transactional functions within the warehouse. These include procurement, order processing, employee payroll, and broader financial oversight. In an optimized smart warehouse, seamless information sharing occurs between internal systems and external partners, enabled through automated data exchanges and system integration. This real-time visibility enhances responsiveness, reduces uncertainty, and helps maintain operational stability.

Reliable communication is critical for the success of smart warehouse operations. To ensure low-latency and high-reliability connections, the WMS communicates with other systems via robust 4G LTE and 5G networks. These technologies provide the infrastructure needed for agile, responsive operations within the Industry 4.0 framework. Additionally, cloud-based smart warehouses often leverage hardware virtualization to enable resource pooling, improving scalability and resource efficiency.

The decomposition view shown below outlines all essential modules required for smart warehousing. It includes not only the top-level functional components but also the sub-modules associated with each enabling technology. Key technologies featured in this view include barcoding, Augmented Reality (AR), Automated Guided Vehicles (AGVs), the Internet of Things (IoT), Warehouse Management Systems (WMS), scanning, RFID, and communication infrastructure.

For example, within the IoT module, several sub-components are identified: Artificial Intelligence (AI), Ambient Intelligence (AmI), a Security Module, and Real-Time Information processing. These elements highlight the critical considerations a smart warehouse designer must account for when implementing IoT-based solutions.

The AI sub-module encompasses techniques such as machine learning and deep learning algorithms, which enable predictive analytics, anomaly detection, and adaptive process control. Closely related, Ambient Intelligence (AmI) builds on AI, sensor networks, and pervasive computing to create an environment that dynamically adapts to the needs and behaviors of users and stakeholders. In this context, AmI contributes to making the warehouse responsive, intelligent, and context-aware.

Real-time information gathered from distributed sensors and devices supports the execution of both AI and AmI functions, enabling timely decision-making and responsive automation. A critical aspect of any Industrial IoT system is security—hence, the inclusion of a dedicated Security Module. This module ensures that proper security controls are in place to protect IoT devices, data transmissions, and system integrity from potential threats.

Together, these sub-packages form the technological backbone of a smart warehouse, ensuring that operations are not only efficient and intelligent but also secure and adaptable to evolving operational demands.

The uses view is presented below. The two core modules in the smart warehouse architecture are the Warehouse Management System (WMS) and the warehouse communication network. The WMS serves as the central control unit, interacting with various back-end systems and external modules to coordinate operations. The communication network functions as a critical link between the WMS and front-end technologies such as Augmented Reality (AR) hardware and Automated Guided Vehicles (AGVs). Additionally, other systems—including the Transport Management System (TMS), Internet of Things (IoT) devices, RFID infrastructure, and various external platforms—connect directly to the communication network to enable seamless data exchange and real-time coordination across the warehouse ecosystem.

Finally, in the deployment view, the mapping of software modules to their corresponding hardware components is illustrated. The data processing module is deployed on both the Warehouse Management Server and Warehouse Manager nodes, which serve as the central hubs for managing operations and analytics. Additional nodes are dedicated to cameras, sensors, scanners, and augmented reality (AR) hardware.

Automated Guided Vehicles (AGVs) are equipped with their own onboard cameras, sensors, and scanners, enabling autonomous navigation and interaction with the warehouse environment. Similarly, smart shelves are outfitted with embedded sensors to monitor inventory and environmental conditions. Each of these hardware components is represented as a separate node within the deployment architecture, emphasizing the distributed and interconnected nature of the smart warehouse system.

Source: https://doi.org/10.1016/j.compind.2020.103343

 

Machine Learning Based Industry 4.0 Framework for Composite Autoclave Manufacturing

Artificial intelligence (AI), machine learning (ML), and data-driven techniques can be used to support and optimize the manufacturing of composite materials. The full model development lifecycle encompasses analysis, optimization, inverse problem-solving, and experimental validation. The Industry 4.0-inspired modeling pipeline tailored to composite processing combines both data-centric and model-centric knowledge engineering practices, with an emphasis on embedding domain expertise directly into the dataset curation and model development phases.

A high-level Industry 4.0 (I4.0) framework is composed of three interrelated sub-systems:

(a) Data acquisition from cyber-physical systems, which may include a combination of real-time production data and simulated outputs;
(b) A data pipeline that manages key preprocessing tasks such as data cleaning, dimensionality reduction, storage, and other logistical operations necessary to ensure data usability;
(c) A model development pipeline, which involves filtering incoming data, constructing relevant knowledge datasets, building and validating predictive models of the manufacturing system, and ultimately deploying these models to support real-time decision-making on the shop floor.

Focusing on the sub-system (c), the proposed framework builds upon the CRISP-DM (Cross-Industry Standard Process for Data Mining) methodology, a widely adopted and open-standard model for data analytics. It provides a structured, iterative approach to data-driven problem-solving, making it especially suited for dynamic, evolving manufacturing environments.

A key element in data-driven modeling is the anomaly detection system, which ensures that incoming data aligns with the distribution of previously collected data. This validation step can be implemented using a variety of techniques, such as classification, clustering, deep learning, or statistical models. If the chosen method determines that a new data point fits within the existing distribution, it is accepted and incorporated into the dataset; otherwise, it is rejected. To enhance this process, domain expertise can be integrated through a dataset knowledgeability exercise, allowing expert judgment to refine the dataset and establish a feedback loop. This hybrid approach contrasts with fully automated, knowledge-agnostic systems and is seen as a vital component in advancing toward Industry 5.0. By enforcing consistency in data distribution, the assumption of independently and identically distributed (IID) data is upheld—an essential condition for making reliable comparisons between models trained on distinct datasets.

USE CASE:

Composite materials have seen broad adoption across industries such as aerospace, automotive, and construction, thanks to their exceptional structural properties—namely, high stiffness-to-weight and strength-to-weight ratios—along with reduced maintenance needs and lower lifecycle costs. Despite these benefits, the production of composite components remains challenged by significant levels of uncertainty, both aleatoric (inherent variability) and epistemic (lack of knowledge), which hinder the consistent manufacture of high-quality parts.

In the aerospace sector, where structural reliability is paramount due to the potentially catastrophic consequences of failure, stringent qualification frameworks exist to ensure the integrity of materials, processes, and designs. While necessary for safety, these frameworks place considerable demands on manufacturers and often limit flexibility in production—particularly when it comes to cost optimization and real-time decision-making on the factory floor (Crawford et al., 2021a).

One example of these challenges can be seen in the use of “bus stop” autoclave cure cycles, where multiple parts, tools, and materials are batched together and cured simultaneously. This method offers practical advantages—such as reduced floor space requirements, more efficient use of capital equipment, and lower overhead—but also increases the complexity of planning and process control. In such settings, shop floor engineers and operators are often required to make in-situ decisions, relying on their expertise and tacit knowledge to maintain process conformance and part quality.

However, these decisions are typically unstructured and lack systematic optimization, revealing an opportunity for the integration of intelligent, technology-enabled decision-support systems. The success of such systems depends not only on access to historical process data, but also on the incorporation of expert insights—bridging human experience with data-driven methods to improve repeatability and efficiency in composite manufacturing.

In bus-stop autoclave curing runs for manufacturing composite aerospace structures, multiple small components—each with similar physical characteristics and qualified to undergo the same cure cycle—are stacked together in a single autoclave to improve production efficiency. Despite the shared cure cycle, each part must maintain a thermal history that stays within an acceptable thermal envelope to ensure product quality.

To achieve a successful cure across all components, parts are carefully selected based on physical similarities, such as laminate thickness, construction type (e.g., monolithic vs. sandwich panels), tooling material, and other relevant attributes. Two critical features derived from each part’s thermal profile—the peak exotherm (the maximum temperature reached during the cure) and the steady-state lag (the highest temperature differential between the part and the surrounding autoclave gas)—are used as indicators of process quality.

Lower values of these thermal metrics generally signify reduced process variability, which correlates with higher-quality outcomes. If these thresholds are exceeded, parts may develop defects like voids, ultimately compromising mechanical properties such as flexural strength, flexural modulus, and interlaminar shear strength. Thus, peak exotherm and steady-state lag serve as essential acceptance criteria for screening cured parts.

In the context of bus-stop autoclave cure cycles, interpretable surrogate models such as Logistic Rule Regression are integrated with expert knowledge through a fuzzy scoring system, enabling an assessment of dataset “knowledgeability” prior to the deployment of black-box models. To further support model validation, two metrics are introduced: specificity as a global confidence indicator, and the novel Decision Boundary Crispness Score (DBSC) as a local, sensitivity-based metric.

The modeling task at the factory level is framed as a binary classification problem, where the objective is to predict whether a carbon fiber prepreg part passes or fails the quality standards following autoclave curing. Specifically, the classification is based on two critical thermal processing criteria:

1. Pass if the peak exotherm temperature is less than 5 °C (Criterion 1)

2. Pass if the maximum lag temperature is less than 20 °C (Criterion 2)

Parts that meet both criteria are labeled as "pass" (class 1), while those that exceed either threshold are labeled as "fail" (class 0).

The model architecture, illustrated in the following figure, consists of two independent predictive models—one for each thermal outcome. This design allows for separate evaluation and parameter tuning tailored to each specific target, enabling more precise learning for each thermal metric.

Both models share an identical architecture, comprising five hidden layers with seven neurons per layer. This configuration was chosen through a parametric study, where it demonstrated the lowest error rate on the test dataset. Each neuron uses a sigmoid activation function, and model training is performed using the Adam optimizer with a binary cross-entropy loss function.

The results demonstrate that DBSC offers a more nuanced and conservative evaluation of both dataset and model quality, especially useful when dealing with uncertain or variable manufacturing conditions. The enhanced explainability and localized insight provided by these methods are particularly valuable to production engineers, supporting trust and accountability when using black-box models in high-stakes, real-time decision-making scenarios.

Source: https://doi.org/10.1016/j.compind.2021.103510

Augmented Reality for Training Applications in Small Enterprises

The fragmented and technically demanding nature of augmented reality (AR) content development continues to limit its broader adoption in industrial settings. Current AR authoring methods often require specialized knowledge in areas such as 3D modeling, programming, computer vision, tracking, and rendering. While automated tools attempt to streamline content creation, they are typically rigid and unsuitable for dynamic or undefined processes.

Many AR development approaches depend on existing resources like CAD models or PDF manuals. However, in cases where these materials are outdated, incomplete, or unavailable—such as with legacy equipment or obsolete electronics—content creation becomes a significant bottleneck.

Although model and symbol libraries are emerging to address this, they require constant updates and rely on fiducial markers for registration. This introduces further challenges, including the need for infrastructure changes (e.g., marker placement and visibility), time-consuming setup, and ongoing maintenance. Natural feature recognition can eliminate the need for markers but instead demands complete, high-quality 3D models and clear visibility of all object features, which is often impractical in industrial environments. Furthermore, while depth sensors and image capture can help generate 3D models without CAD data, these methods may interrupt workflows, require invasive equipment, and struggle to capture small or occluded components.

To address these challenges, AR content creation methods must evolve to accommodate non-experts, operate without dependencies on pre-existing resources or infrastructure modifications, and adapt to changes in processes or tasks. The proposed method introduces a low-disruption, template-based approach that bridges traditional and AR interfaces by capturing expert knowledge—both explicit and tacit—through eye-tracking and structured information mapping.

A Multi-Modal Framework for AR Content Creation and Delivery

The framework supports two user roles:

• Content creators (e.g., trainers or experts), and

• Content consumers (e.g., trainees or task novices).

It comprises five main steps:

1. Business Need Identification – Pinpoint a requirement for efficient knowledge transfer between experienced and inexperienced staff, particularly for hands-busy tasks.

2. Task Recording via Eye Tracking – The expert performs the task while wearing eye-tracking glasses, generating visual and audio data that are then broken down into clear training steps. These are populated into a Unity 3D-based AR content template.

3. Iterative Refinement – Content is refined through usability testing and feedback loops.

4. User Feedback Collection – Feedback during real-world use is captured to guide further optimization.

5. Knowledge Preservation and Empowerment – Organizations can retain critical knowledge through archiving and moderation, while users gain autonomy through self-paced learning.

(a) Task Recording & Information Mapping: The initial stage involves capturing the training task using eye-tracking technology. The recorded content is then structured through information mapping to define clear instructional steps. (b) Unity Development Environment: Unity is used as the core development platform, providing a visual interface for organizing and integrating multimedia elements into the AR training experience. (c) Augmented Repair Training Application Template with Mixed Reality Toolkit: The template, built on Unity and enhanced with the Mixed Reality Toolkit, offers a pre-configured structure for rapidly building AR training applications with intuitive interaction components. (d) Content Import: Trainers upload task-related media—such as video clips and images—into the template to visually support each instructional step. (e) Step Customization: Trainers define and organize the number of training steps, and for each one, they add supporting images, descriptive text, and other guidance to enhance user comprehension.

The training content (video, audio, gaze data) is captured directly during task execution. Eye-tracking glasses combine a front-facing video feed with a gaze fixation marker and integrated audio. This method enables the transfer of explicit procedural steps and tacit expertise, such as visual attention and workflow patterns. The resulting media is edited using standard software (e.g., Apple iMovie) and mapped into structured instructional steps.

Design of the AR Interface: Augmented Repair Training Application 

The template organizes content using user-centric design principles. A numbered step menu allows users to track progress and navigate between instructions. Controls such as “home,” “back,” and “next” replicate traditional interface metaphors, easing the transition for new AR users. To enhance clarity, content is displayed with high-contrast text (white font with blue accents).

Instructional content is structured using information mapping, a method that includes:

• Chunking related steps and concepts

• Highlighting essential information

• Ensuring consistency in formatting and structure

• Integrating visuals effectively

• Presenting content hierarchically for intuitive learning

Verification of Learning and Task Completion

Verification of knowledge transfer can be achieved through interactive assessments or expert review. While quizzes may suit well-defined tasks with quantifiable outcomes, they lack flexibility for variable, real-world processes. For this application, expert review was chosen for its simplicity and alignment with the goal of maintaining learner autonomy while ensuring training effectiveness.

USE CASE: 

The AR training assistance method was implemented to create shop floor training content for a small enterprise. While participants expressed enthusiasm about the innovative approach, many were unfamiliar with the tools involved—specifically Unity3D, eye-tracking software, and the AR device itself. This unfamiliarity led to several usability challenges during the content editing and deployment stages. Technicians required support to navigate unexpected issues such as accidentally deleted scripts or misplaced files.

Despite these challenges, subjective feedback from stakeholders was highly positive. Senior management and the sales team viewed the AR training tool as a valuable and forward-thinking investment. When showcased at trade events and to existing clients, the tool generated strong interest and favorable responses. Shop floor staff were also enthusiastic about the interface, interpreting it as a sign that their company was embracing modern technologies and investing in employee development. However, a small number of staff expressed discomfort using certain features of the AR device, such as voice commands and air tap gestures.

In response to the challenges identified during deployment, it became evident that a more accessible content creation interface would improve usability and reduce technical barriers. To this end, the development of an "intermediary interface" is proposed for future work. This interface would offer a form-based editing environment, simplifying the interaction between non-expert users and the underlying AR content formatting system.

Key features of the proposed intermediary interface include:

• Form-style input fields for guided data entry (e.g., text, video, audio annotations)

• Built-in tools to edit and align eye-tracking outputs into AR-friendly formats

• Tooltips and help documentation embedded within each section to guide users step-by-step

• Automated conversion of form inputs into AR code via a secure back-end service

• Responsive layout using a bootstrap-style grid system to accommodate various AR form factors

The Integrated Development Environment (IDE) would be decoupled from the front end and managed in the back end, ensuring secure storage, script execution, and formatting integrity. Additional layers of security and content moderation can be introduced to prevent accidental corruption, unauthorized access, or unapproved deployments.

This redesigned workflow aims to empower non-technical users to contribute to AR content creation while preserving the integrity, flexibility, and scalability of the training system.

Source: https://doi.org/10.1016/j.compind.2020.103208

Collaborative dynamics in open source software development: Team interaction and the role of project manager.

In a recently published study, we examine the impact of communication network characteristics on project performance, measured in terms of the number of issues closed within open-source software development projects. We also examine how this project outcome is affected by project managers' active participation in these communication networks. The results obtained from analyzing 120,243 observations of unbalanced dynamic panel show that with higher team interaction the technical problem-solving capability of team initially increases but then decreases. We find that project manager's participation in team discussions increases its problem-solving capability. That is, there is a moderating effect of manager participation in flattening the curvilinear relationship between team interaction and technical problem-solving performance. In essence, project managers' active participation alleviates the reduction in weekly issue closure rate once the density of team interaction goes beyond the inflection point.

The findings have implications for project managers and developers working on OSS projects. Our research provides recommendations on the attributes of the project manager that would be best suited for improving project team performance and sheds light on aspects of team communication that a project manager may need to manage. The project manager should promote team interaction to enable the flow of ideas and information to improve the issue resolution rate of the team. However, the project manager must remain aware of the growing fatigue from information overload in teams. Our research provides evidence for the inflection point beyond which such fatigue might start to impede project performance. We identify the specific time when the project manager should intervene to reduce the negative effects of information overload. Such an understanding enables project managers to make the best use of the project management hours and add value.

As issue closure is often required for bug fixes in new product development, our study shows that overall product development will substantially improve at high team interaction by employing a manager who actively engages in team interactions, enabling faster product development to the market cycle. As team member interaction intensity can seldom be controlled, our study provides guidance how expensive project management resources can be allocated depending on the extent of team interaction to shorten the new product development cycle.

The findings of this research also have implications for how OSS platforms can be made more effective in managing software development activities. Since team
performance decreases with additional team interactions after an inflection point, OSS platforms could consider developing a system that categorizes communication notifications that a team member receives (such as a thread or message broadcast) depending on the priority level. For example, suppose a software project on GitHub has several modules based on which developers are grouped. Developers will prefer to stay updated on the messages from their colleagues working on the same inflection point identified in this study could serve as a reference point for the platform when project performance is impacted by information overload. The system developed by the OSS platform could get triggered at this reference point so that communication notification to team members can be prioritized based on the module that a developer is working on. In turn, developers can focus on a few important messages to enhance team
performance.

The OSS platform may also aid project managers to participate more in team activities as our results suggest that active participation of project managers enhances team performance by flattening the curve. For instance, the platform may provide project managers with a dashboard that summarizes project progress. Such a service may enable the manager to get a quick and accurate update of the team and may help the manager to participate in a more meaningful way. The platform may also provide a metric that measures the relevance of communication by developers, enabling managers to efficiently streamline communication within their teams.

Source: https://doi.org/10.1002/joom.1324

Managing Project Cost Overruns Using Bayesian Network

High levels of project management maturity are often associated with improved operational outcomes—such as enhanced quality, timely delivery, and cost efficiency. However, this correlation does not always hold true. Many existing project management maturity models lack clarity and consistency; they are typically constructed from extensive lists of project domains, vague categories, ambiguous criteria, and loosely defined practices. These models often suffer from the absence of standardized terminology, fragmented integration of key project areas, and poor usability—frequently overwhelming practitioners with an excessive number of elements to assess during maturity audits.

In light of these limitations, Bayesian networks offer a promising alternative for modeling and analyzing the complexities of project management maturity. They provide a structured, visual, and data-informed approach to understanding causal relationships among variables. By leveraging causal reasoning, Bayesian networks help explain how multiple factors contribute to a particular outcome. Their adaptability makes them well-suited to the current landscape shaped by widespread IT adoption and the growing influence of machine learning.

A Bayesian network is a probabilistic graphical model that uses a Directed Acyclic Graph (DAG) to represent a set of variables and the causal dependencies between them. In this framework, nodes represent random variables, while directed edges (arcs) denote the influence of one variable (the cause) on another (the effect). Each connection reflects a conditional dependency, and the strength of these relationships is quantified using probability distributions.

Bayesian networks support two fundamental types of inference:

1. Backward (diagnostic) inference – used to identify the most likely cause of an observed effect.

2. Forward (predictive) inference – used to estimate the likelihood of an outcome based on known or assumed causes.

A critical strength of Bayesian networks lies in their ability to incorporate both expert judgment and empirical data. A robust network is built on a foundation of accessible and well-structured data, relevant in volume and quality to the issue at hand, and presented in a clear, visual, and interpretable format. This enables organizations to model project management maturity using a cohesive and data-driven causal structure, ultimately enhancing both understanding and decision-making.

To assess cost overrun risks linked to low project management (PM) maturity, the proposed method leverages two complementary data sources within a Bayesian Network (BN) framework: expert knowledge and historical project data. Experts contribute to structuring the BN by identifying input, intermediate, and outcome nodes. In contrast, a relevant dataset is used to train the algorithm, enabling the calculation of each node’s Conditional Probability Table (CPT).

It’s essential to differentiate between two modeling layers: semantic modeling and Bayesian (causal) modeling. Semantic modeling refers to the conceptual representation of project management maturity, while Bayesian modeling encompasses the systematic steps required to construct a causal BN structure. The methodology follows a seven-step process, starting with the definition of a streamlined framework for evaluating PM maturity.

Step 1: Define a Simplified PM Maturity Framework
Traditional Project Management Maturity Models (PMMMs) often involve exhaustive audits with over 150 criteria, which, while comprehensive, are impractical for correlational analysis due to the volume and lack of sufficient audit data. Instead, a more efficient two-tiered model is proposed:

• A macro-level model with a reduced number of key maturity indicators to help consultants quickly identify major weaknesses.

• A detailed level, informed by expert input, to be used when more granular analysis is needed.

This simplification allows for easier integration into causal models while maintaining diagnostic value.

To structure PM maturity, project features are organized into a three-layer classification system:

• Layer 1: Project Domains — Aligned with PMBOK® areas:

  • Social (S): Covers communication, HR, and integration management; emphasizes team dynamics and the project manager’s leadership.

  • Contract (C): Encompasses scope and risk management.

  • Results (R): Relates to cost, schedule, and quality management.

  • Interface (I): Includes procurement and stakeholder management.

• Layer 2: Project Phase Timing — Determines when a PM practice occurs:

  • Prepare (P): Before execution.

  • Monitor (M): During the project.

  • Valorize (V): At project closure or evaluation.

• Layer 3: Execution Characteristics — Describes how PM activities are conducted:

  • Activity Granularity (A): Level of task detail or time frame.

  • Resource Involvement (R): Who is involved and what tools are required.

  • Frequency (F): Timing and repetition of activities (e.g., daily, cyclical, or milestone-based).

These categories are used to build a tagged matrix of PM maturity criteria, creating a standardized nomenclature for assessing each PM task.

Step 2: Define Measurement Scales for BN Input Nodes
Maturity is typically assessed using a five-level performance scale, where advancing from level n to n+1 means all best practices at level n are fully implemented, and implementation of the next level is underway. The levels include:

• Level 1: Absent/Discovery — PM processes are undefined or applied inconsistently and informally.

• Level 2: Definition and Implementation – At this stage, the project manager and team define and establish key project management (PM) practices to be used throughout the project.

• Level 3: Measurement and Analysis – Performance data is collected from various sources (e.g., monitoring and forecasting systems), enabling the team to analyze results and make data-driven adjustments.

• Level 4: Managing Gaps and Interdependencies – The team becomes capable of handling complexity and uncertainty, especially when managing interrelated tasks across concurrent projects.

• Level 5: Capitalization and Continuous Improvement – The final stage focuses on institutionalizing learning. The team adapts quickly to changes, manages complexity, and shares improvements across the organization.

Step 3: Identify Drift Factors
This step links project maturity levels to performance deviations (drift). Drift factors represent recurring, high-impact causes of performance shortfalls such as cost overruns or delays. The identification process includes:

1. Drift Cause Identification – Experts compile a set of potential causes of deviation, based on project data.

2. Cause Selection – Non-recurring or low-impact causes are filtered out to ensure predictive reliability.

3. Cause Matching – Relationships between selected drift factors and PM maturity model (PMMM) components are established as tuples, forming the basis for Bayesian modeling.

Step 4: Define Aggregation Rules
Experts define how each maturity level correlates with the selected drift factors. Their insights are used to construct the conditional probability tables (CPTs) for the Bayesian Network (BN), effectively translating domain knowledge into mathematical rules.

Step 5: Prepare and Structure the Dataset
Since Bayesian Networks are data-driven, data preparation is critical. This involves cleaning the dataset by removing invalid values, standardizing formats, and handling outliers. Each row should represent a unique project instance. The quality and structure of the data must support the number of variables (columns) and ensure meaningful learning.

Step 6: Define the Target Node
The target node in the BN represents a measurable project performance issue (e.g., budget overrun or schedule delay). This variable should be specific enough to inform PM decision-making but generalized enough to avoid data fragmentation. The number of its possible states must balance interpretability with statistical reliability, typically optimized using metrics such as the Akaike Information Criterion (AIC) or Bayesian Information Criterion (BIC).

Step 7: Apply Learning and Inference Algorithms
Once the network structure is defined, it is trained using historical project data, with additional validation against a separate test set. The expectation–maximization (EM) algorithm is employed for parameter learning due to its robustness in handling incomplete data. BN models serve two core purposes:

• Diagnostic analysis – Identifying the most likely causes (drift factors) of current or future project failures and mapping them to key maturity criteria.

• Predictive analysis – Estimating the likelihood of future performance issues based on assumed changes in PM maturity levels.

Bayesian Networks rely on two types of algorithms:

1. Structure and parameter learning algorithms – Define the network’s architecture and CPTs.

2. Inference algorithms – Propagate evidence throughout the network for analysis and prediction.

Hyperparameter tuning in BNs is inherently iterative, requiring trial-and-error testing to find an optimal balance between model complexity and predictive accuracy.

BN Validation and Quality Metrics

To ensure a valid and effective BN, several evaluation criteria should be considered:

• Semantic coherence and interpretability – Results must be understandable and actionable for PM experts.

• Completeness and coverage – Nodes and states should accurately reflect key PM dimensions, supported by sufficient data.

• Relevance and granularity – The target variable should have meaningful, precise outcomes.

• Model conciseness – The structure must avoid combinatorial explosion, ensuring computational efficiency.

• Learning quality – The CPTs must be well-defined and complete. Metrics like AIC and BIC can assess whether the model will generalize well to new data.

A well-constructed BN supports more informed project decisions by linking PM practices to measurable performance outcomes through data-driven, causal modeling.

USE CASE:

This model is especially relevant for consulting firms working with large-scale engineering projects in sectors such as oil and gas. These firms are often brought in when project performance has deviated significantly from expectations. Their clients expect:

• A clear diagnosis of the issues causing underperformance,

• Immediate recommendations for corrective actions, and

• A forecast of the likely outcomes if those improvements are implemented.

With access to detailed long-term data—such as deviation dates, financial impacts, root causes, and implemented corrective actions—consultants can use Bayesian networks to create predictive models. These models not only explain why issues occurred but also help anticipate future risks and guide decision-making to prevent recurrence.

The application of the BN model predicted that the project had a 15% probability of overcost corresponding to less than 1% of the expenses. An overcost between 1% and 10% has a probability of 49%. An overcost between 10% and 100% had a probability of 21% and an overcost corresponding to more than the 100% of the budget had a probability of 15%. The training and testing accuracy were within the 90% to 96% range with the relative error between the training data and the test data being less than 6%.

The model helps to predict the probability of entering into an overcost situation, based on the level of maturity evaluated in the audited project. 

Source: https://doi.org/10.1016/j.compind.2020.103227

Platform for Intelligent Automation in a Pharmaceutical Laboratory

The acquisition, processing, and utilization of data have become fundamental drivers behind the adoption of emerging technologies such as Cyber-Physical Systems (CPS), the Internet of Things (IoT), predictive maintenance, dynamic scheduling, big data analytics, cloud-based platforms, and digital twins. As a result, efficient data exchange and management are critical—if mishandled, they can become significant obstacles.

In the development of decision support tools, data preparation stands out as the most vital and time-intensive phase. Real-world data often arrives in inconsistent formats, may be incomplete or error-prone, lacks proper structure or traceability, and is frequently contaminated with noise. The core challenge typically lies not with the analytical tools themselves, but with the underlying data communication and integration infrastructure that governs data acquisition and handling.

To enable the effective deployment of dynamic decision support systems (DSSs), next-generation industrial architectures aligned with Industry 4.0 must ensure real-time data consistency, especially in time-sensitive applications. Addressing this need calls for a robust platform capable of automating data acquisition, preprocessing, and management—achieved through the development of an intelligent middleware layer.

This middleware functions as a bridge between field-level devices and higher-level DSSs. It facilitates seamless integration with technologies such as industrial communication protocols, fog computing architectures, and big data storage solutions. A compelling example of this platform's application can be found in pharmaceutical industry quality control labs, where a prototype system integrates an autonomous robotic module for material preparation with a dynamic scheduling decision support tool.

The integration of production operations with business planning and logistics relies heavily on the development of data-driven industrial architectures. A key framework for achieving this integration is the automation pyramid, as defined in the ISA-95 standard.

This model consists of five hierarchical layers that collectively enable seamless information flow across industrial environments:

1. Field Layer: This base layer includes sensors and actuators, providing real-time data directly from the physical environment. Technologies such as barcodes and Radio-Frequency Identification (RFID) enable resource tracking and monitoring, supplying critical input to intelligent systems. In sectors like pharmaceuticals, smart sensors also support Quality by Design (QbD) approaches through Process Analytical Technologies (PAT).

2. Control Layer: At this level, Programmable Logic Controllers (PLCs)—robust industrial computers with IP65/67 protection—manage machine-level control with exceptional precision. Their deterministic cycle times (typically <1ms) make them ideal for time-sensitive applications such as motion control.

3. Supervisory Layer: The Supervisory Control and Data Acquisition (SCADA) system acts as a bridge between field devices, controllers, and supervisory systems. It is the de facto standard for industrial data acquisition and ensures reliable communication across various automation layers.

4. Manufacturing Execution Layer: Manufacturing Execution Systems (MES) monitor, analyze, and report real-time production activities, tracing the full transformation process from raw materials to finished products. With the decentralization of decision-making systems, MES must increasingly integrate with distributed and intelligent systems.

5. Enterprise Layer: Enterprise Resource Planning (ERP) systems manage core business functions through modular and integrated applications. With advancements in data science, ERP systems are increasingly leveraging machine learning and data mining to extract actionable insights from vast datasets.

Each layer in the automation pyramid plays a distinct role in acquiring, processing, and utilizing data. Traditionally, information flowed hierarchically from bottom to top. However, the rise of mass customization and the need for real-time responsiveness require high-volume, bidirectional data exchange across all layers—continuously, in parallel, and often over internet-based infrastructure.

While the structure of the automation pyramid remains relevant, its static, hierarchical nature is evolving. Emerging technologies are driving a shift toward highly interconnected systems with greater horizontal integration. Direct communication between lower-level devices (e.g., field buses) and higher-level applications now demands enhanced interoperability and standardized communication protocols.

In this transformed landscape, cloud computing has become a central enabler. It offers a globally accessible platform for data storage and management, along with scalable services that enhance flexibility, security, and cost-efficiency. Cloud connectivity eliminates the need for local data storage, allowing data to be accessed and analyzed remotely, anytime and anywhere. Cloud-based Industrial IoT (IIoT) platforms like MindSphere (Siemens) and PI System (OSIsoft) are already revolutionizing process industries with plug-and-play analytics capabilities.

To complement the cloud, fog computing has emerged as a critical intermediary. It enables local data processing, filters sensor data, and ensures low-latency responses necessary for time-sensitive tasks involving actuators and robotics. Fog nodes act as intelligent gateways between industrial assets and cloud services, managing unstructured data while balancing speed and bandwidth requirements.

As industrial systems become more interconnected, the need for machines and applications to communicate seamlessly becomes paramount. This requires shared standards, common languages, and unified infrastructure that supports both centralized and decentralized intelligence—ushering in a new era of cyber-physical convergence in the smart factory environment.

USE CASE:

In the pharmaceutical sector, many labor-intensive operations offer significant potential for automation. One particularly complex area is the dynamic scheduling of tasks within Quality Control (QC) laboratories. These labs are essential for ensuring that every drug batch meets strict safety and quality criteria through comprehensive sampling and testing procedures.

Each laboratory test typically comprises six main stages: system preparation, system suitability, sample preparation, analytical run, data processing, and final review. These stages break down into numerous detailed and often customized subtasks, each requiring specific analysts, equipment, or both, depending on the nature of the test.

While production workflows are generally organized through pre-defined schedules, QC operations are often interrupted by the arrival of urgent, high-priority samples, which must be addressed immediately. This requires the system to dynamically reschedule tasks in real-time. The challenge lies in coordinating both equipment capabilities and operator skill sets, taking into account resource availability and operational constraints. This creates a dual-resource constraint problem, where both personnel and machines must be optimized simultaneously.

A critical and precision-sensitive component of this workflow is material preparation, which falls under the sample preparation stage. To improve efficiency and reduce manual handling, a prototype automation module has been developed to autonomously manage material preparation tasks. This module is designed to share resources and operate across multiple test sequences concurrently, enhancing throughput, flexibility, and overall laboratory efficiency.

Despite these advancements, several challenges remain. Among the most notable are the complexity of test recipes and the short shelf life of prepared materials, both of which impose additional constraints on automation and scheduling systems.

The prototype system is composed of several integrated components, including two industrial identification platforms—PC-based and PLC-based, an automation module, a SCADA system, the Middleware for Intelligent Automation (MIA), a dynamic Decision Support System (DSS) with scheduling capabilities, and a Data Warehouse (DW).

The PC-based platform utilizes a personal computer configured to manage data acquisition from two identification technologies: Barcode scanners and the RFID RFH620 system.

The PLC-based platform consists of a PLC, an Extension Terminal, distributed I/O system, an HMI Basic Panel, and a SCADA PC, all working together to handle data from Siemens' Barcode and RFID systems. This setup is responsible for real-time monitoring and control within the automated environment.

The MIA platform is hosted on a dedicated PC running SQL Server Enterprise 2017 and a Python environment, serving as the core of data integration and communication between systems. It acts as a bridge linking lower-level control devices with higher-level decision tools.

The Data Warehouse (DW) plays a crucial role by aggregating and managing data extracted from both the Enterprise Resource Planning (ERP) system and the Laboratory Information Management System (LIMS). In this setup, LIMS functions similarly to a Manufacturing Execution System (MES) but is tailored specifically to meet the needs of laboratory operations, including test tracking, compliance, and quality management.

Together, these components create a cohesive infrastructure that supports automated identification, real-time data acquisition, intelligent scheduling, and data-driven decision-making in the quality control processes of pharmaceutical laboratories.

Source: https://doi.org/10.1016/j.compind.2020.103329

Detecting Hidden Failures in Cryogenic Pumps by Combining Artificial Neural Networks and Simulation

LNG tank cryogenic pumps are installed vertically inside a storage tank, with the suction (foot) valve positioned at the bottom. This design eliminates the risk of major tank leaks caused by issues with pipes or connections.

Both the pump and motor units are submerged, with the column serving as a guide during installation and also functioning as the discharge pipe that channels the pumped liquid to the top of the tank. Each pump is equipped with an inducer, an axial flow impeller positioned at the lowest possible level in the tank to enhance the Net Positive Suction Head Required (NPSHR). A suction valve is used to isolate the tank contents from the pump column; it is flanged to the lower end of the pump column and is closed by coil springs as well as the hydrostatic pressure of the liquid inside the tank.

Plant operators rely on information from various plant information systems to monitor the pump’s operating conditions and identify any potential issues. However, despite the wealth of information provided, it can be challenging for operators to make clear judgments about possible pump malfunctions due to the numerous operating variables, such as varying tank levels, flow rates, pressure, LNG density, and pump operating hours. Additionally, given the pump's functional setup, condition monitoring is restricted to operational variables, meaning data such as vibration levels, bearing temperatures, and motor temperatures are not available.

In these conditions, an Artificial Neural Network (ANN) can offer valuable support by predicting the motor's energy consumption over time based on any operating condition, and comparing it to actual energy usage to detect potential malfunctions.

The ANN dynamic simulation model considers seven inputs: flow, pressure, inlet temperature, output temperature, tank level, LNG density, and operating time since the last overhaul. The model has a single output, which is the energy consumption per hour. To achieve the desired accuracy in predictions, the hidden layer was set to contain a minimum of 20 neurons.

14,988 data points (simulation steps) were used for analysis.

• 10,082 data points were collected during the pump’s operation leading up to its overhaul. The first data point corresponds to the pump running for 2,828 hours, and the last before the overhaul was logged at 16,126 operating hours.

• 4,857 data points were recorded after the overhaul, starting from zero operating hours (i.e., as-new condition) until a sudden and catastrophic failure occurred at 5,171 hours.

The unexpected nature of the failure, with no prior warning signs, prompted strong interest from plant managers in applying machine learning techniques to explore the potential for predictive analytics.

For training the Artificial Neural Network (ANN), at least 70% of the pre-overhaul data was selected. During this period, the pump was operating under a wide range of conditions, yet performed well, with only normal wear detected during the overhaul—no significant damage was found.

Once the training phase achieved a satisfactory level of accuracy, the ANN model was applied to test pump performance both before and after the overhaul.

The outcomes revealed several noteworthy insights. During the testing phase prior to the overhaul, the accumulated error showed only a slight upward trend. It's important to note that during this period, the pump had already accumulated a significant number of operating hours—much higher than those seen during the training period. This discrepancy could account for the mild increase in the slope of the error curve.

In contrast, after the overhaul, there is a noticeable and consistent rise in the slope of the error curve. This steep increase persists over time, even though the pump's operating hours are comparable to those within the training dataset. Such behavior strongly indicates an abnormal condition or malfunction occurring immediately after the overhaul.

To evaluate the robustness of the ANN algorithm, its results were benchmarked against those produced by other established machine learning techniques used in predictive analytics. This comparative analysis helps support informed decision-making for asset management (Olivencia et al., 2015). For this purpose, two additional algorithms were implemented and tested on the same dataset: Support Vector Machines (SVM) and Random Forest.

The findings revealed that all three models—ANN, SVM (non-linear), and Random Forest—demonstrated comparable predictive accuracy, with Random Forest achieving the best performance in terms of error reduction. Specifically, the Root Mean Square Error (RMSE) values for the testing phase before the overhaul were:

• ANN: 0.00043

• SVM (non-linear): 0.00045

• Random Forest: 0.00038

For the testing phase after the overhaul, the RMSE values were:

• ANN: 0.0022

• SVM (non-linear): 0.0023

• Random Forest: 0.0020

These results confirm that while all models are effective, the Random Forest algorithm slightly outperforms the others in both scenarios.

Source: https://doi.org/10.1016/j.compind.2019.103164

Service Network Design By Combining Expert knowledge and Machine learning Techniques

Service network design and delivery have a wide range of applications, spanning industries such as manufacturing-oriented product service networks, emergency medical services, agricultural maintenance networks, supply chain logistics, and maintenance service networks. In many cases, it becomes crucial to select a group of service providers within the entire network. This problem can be categorized as a multi-objective optimization challenge, for which various AI-based decision tools, particularly artificial neural networks (ANNs), have been employed for classifying schemes. ANN-based approaches have been developed to assess and select potential service providers, determine industrial site locations, and map mineral resources. Furthermore, ANNs can be integrated with fuzzy logic, and a two-phase decision model can be used to select suppliers in third-party reverse logistics networks. Additionally, an adaptive neural network-based decision support system, combined with fuzzy expert-designed rules, can be utilized to select the optimal industrial site.

A flexible fuzzy-neuro approach can be developed to select service providers for building a maintenance service network and for making the final decision. To assist managers in making informed choices, it's crucial to categorize the set of potential schemes and narrow them down to a manageable number of carefully selected alternatives through decision-making methods.

PROCESS:

The process begins with identifying the evaluation criteria. Relevant data must be gathered to meet the basic requirements based on these criteria.

The second step, which is the most critical, involves classifying the set of service providers and presenting decision-makers with an optimized set of alternatives.

The final step is the human decision-making phase, where the optimal group is chosen.

For data collection, in certain contexts such as agricultural maintenance service network, geographic information systems (GIS) and enterprise information systems can be utilized to gather the necessary data for evaluation.

During the classification phase, an innovative adaptive fuzzy-neuro approach, informed by expert knowledge, can be employed to aid decision-making. This (combined) methodology can address the challenges of service provider selection and service network construction.

The fuzzy-neuro network integrates the reasoning capabilities of fuzzy logic with the interconnected structure of neural networks, offering an effective approach for a range of applications, including medical services, production scheduling, robotics control, product evaluation, and fault diagnosis. A fuzzy-neuro network is typically structured as a five-layer neural network, as described below:

• Layer 1 (Input Layer): This layer is responsible for receiving and importing the input samples.

• Layer 2 (Fuzzification Layer): Here, the inputs are fuzzified using a set of predefined membership functions, transforming crisp values into fuzzy values.

• Layer 3 (Rule Layer): This layer generates initial rules from the input samples, with each rule corresponding to a neuron in this layer. The connections between the fuzzification layer and the rule layer are determined by these rules.

• Layer 4 (Defuzzification Layer): In this layer, the fuzzy input is defuzzified, converting fuzzy values back into crisp values.

• Layer 5 (Output Layer): The final processed results are presented in this layer, providing the output of the fuzzy-neuro network.

Recently, the adaptive fuzzy-neuro network, which combines adaptive techniques with fuzzy logic, has gained widespread application. The Takagi–Sugeno model is often used to represent fuzzy rules in an IF-THEN format. Unlike the standard fuzzy-neuro network, the adaptive fuzzy-neuro network is structured with six layers.

• Layer 1 (Input Layer): This layer handles the import of sample data into the fuzzy neural network.

• Layer 2 (Fuzzification Layer): In this layer, input characteristic variables are transformed into fuzzy variables using a set of membership functions.

• Layer 3 (Rule Layer): This layer contains a set of fuzzy rules, which are expressed according to the Takagi–Sugeno model.

• Layer 4 (Normalized Firing Strengths Layer): Each rule is evaluated, and normalization is applied to the firing strengths of the rules.

• Layer 5 (Defuzzification Layer): The fuzzy values generated by the fuzzy inference system are processed. Various methods, such as the mean of maximum method and weighted average defuzzification method, are used to eliminate fuzziness.

• Layer 6 (Output Layer): This layer has a single node that generates the output data.

 

A fuzzy reference system is developed to capture and represent the expertise and knowledge of specialists. Meanwhile, an adaptive neural network is used to fine-tune the parameters. The combination of these components enhances both learning accuracy and generalization capabilities.

USE CASE:

The service provider plays a vital role in the maintenance service network, working alongside the manufacturer and the customer to form a comprehensive maintenance service system. Currently, such networks are in place across various industries, including agricultural machinery, the automobile sector, military equipment, and other complex machinery.

Agricultural machinery operates in a dynamic and continuous manner, which makes it prone to failures, particularly during peak farming seasons. Designing the layout for agricultural equipment maintenance services is therefore more complex. To address this, multiple service providers are selected to create a dynamic service network that can deliver cross-regional maintenance for agricultural machinery. This approach aims to enhance equipment reliability and extend the operational lifespan of the machinery.

The challenge of service network design and the joint selection of service providers can be framed as follows:

1. The service area is divided into multiple service districts.

2. Each service district has a set of potential service providers, and only one provider can be chosen per district.

3. The selected service providers together form the maintenance service network.

Given the large number of possible permutations and combinations in selecting service providers for different districts, the potential number of configurations for the maintenance service network can be enormous. As a result, it becomes essential to identify and select the best possible solutions from a vast array of options.

The goal of intelligent decision-making system is to determine the optimal group of service providers to deliver the best maintenance service for customers at the right time in maintenance service network. A set of evaluation criteria have to be defined prior to evaluate prospective service providers in the maintenance service network. These evaluation criteria could include reputation service provider, size and quality of fixed assets, geographic advantage, operational performance, experience in similar products, cost of services, quality of service, long-term relationship, relationship with neighboring service providers, total delivery cost of spare parts, service network construction cost, and geographical spread and range of service provided. Some features, such as cost of services, total delivery cost of spare parts and service network construction cost, are quantitative, while others are qualitative.

These evaluation criteria can be categorized into various clusters. The first cluster focuses on service provider's performance: reputation, size and quality of fixed assets, and geographical advantage. The second measures the business capability of service providers in four aspects: operational performance, experience in similar products, cost of services, and quality of service. The third is related to the manufacturer and whole service network: long-term relationship, relationship with neighboring service providers, total delivery cost of spare parts in the whole network, service network construction cost, geographical spread and range of service provided. Some features are developed to evaluate the service providers in whole service network by evaluating the service provider in each area, and others are defined to assess the performance of all service providers in the service network.

The decision-making methodology can be divided into four stages: generating alternatives and data collection, expert analysis, classification and final decision. 

The first step is to generate a set of potential alternatives and assess the value of evaluation criteria for each service provider group. The data collected includes information on the service providers, such as their location, the frequency of failures in the maintenance service network, geographic details, and service-related costs—such as labor, logistics, spare parts, and service station construction costs. Given the complexity and time-consuming nature of this task, data is gathered through various methods, including maps, GPS, computer records, and reports generated by the service providers. Additionally, the list of available service providers in each service area can be sourced from their enterprise information system. Similarly, the evaluation criteria values must be established for expert analysis. For quantitative features, values can be calculated using specific formulas based on the collected data. However, for qualitative features, direct calculation is not possible. In this case, a group of leading experts and relevant stakeholders is consulted to determine the value of these criteria and evaluate each service provider group. The experts are selected from the company’s expert database, and their roles vary from professors and maintenance service center managers to field service managers. The stakeholders involved come from various departments, such as finance, marketing, planning, and human resources.

As a second step, leading experts and stakeholders collaborate to review the information on each service provider and assess the performance of each group. For every service provider group, the decision group analyzes and discusses the available data, ultimately reaching a consensus among all members.

An adaptive fuzzy-neuro-based approach can be used in the third stage to assess the suitability of each service provider group by determining their ranking scores. This classification method combines the adaptive fuzzy-neuro approach with a two-class classification method, where varying ranking scores reflect the differences in the quality of service provider groups within the maintenance service network. Several parameters need to be adjusted in this approach, and the adaptive fuzzy-neuro method is found to be more appropriate than other machine learning techniques, such as the fuzzy inference system (FIS), particularly when sufficient training samples are available. Unlike the fuzzy rules used in FIS, which are typically defined by experts, the adaptive fuzzy-neuro method allows for both offline training and online learning, enabling the system to self-adjust its fuzzy inference rules.

Therefore, a fuzzy-neuro network-based selection system is better suited for evaluating service provider groups with a fuzzy knowledge base and large datasets. Additionally, the two-class classification technique is employed during the classification process, with the advantage of presenting the results in an ordered sequence.

n two-class classifiers can be used, and the method generates return values in a cascading mode, starting with i = 1. The computational process ranks the alternatives into n classes.

The adaptive fuzzy-neuro classifiers can be utilized to compute ranking scores representing the class of each service provider group. These classifiers can be refined using a set of training samples. In this way, the adaptive classification approach integrates FIS with neural networks. The adaptive fuzzy-neuro network consists of six layers, and unlike previous fuzzy-neuro implementations, detailed information for each layer is provided as follows:

• Layer 1 (Input Layer): The adaptive fuzzy-neuro approach receives a set of evaluation criteria for the service provider group.

• Layer 2 (Fuzzification Layer): Gaussian membership functions are used to fuzzify the inputs.

• Layer 3 (Rule Layer): Rules arre defined by the decision group, and fuzzy rules from the knowledge base are applied for approximate reasoning.

• Layer 4 (Normalized Firing Strengths Layer): The firing strengths are normalized.

• Layer 5 (Defuzzification Layer): The weighted average defuzzification method is applied in this layer.

• Layer 6 (Output layer): The floating-point data is the output and the range of output is between the values 1 to n. The outputs of the method are rounded to the nearest integers to represent the ranking score of these group.

Adaptive fuzzy-neuro method can be trained in a supervised manner. Least squares estimation method and back propagation approach can be integrated and applied to train the proposed n two-class adaptive fuzzy-neuro classifier.

In the final stage, based on the classification results obtained by the proposed adaptive fuzzy-neuro approach, the groups with the highest ranking score can be analyzed and assessed by the invited decision group and the appropriate group of service providers can be selected.

 

Source: https://doi.org/10.1016/j.knosys.2019.105263