Complex Manufacturing Systems Research Articles

Multimodal regression and mode recognition via an integrated deep neural network

Deep learning has emerged as a powerful tool to model complicated relationships between inputs and outputs in manufacturing systems. Existing deep learning approaches in manufacturing are often used to directly predict the Variables of Interest (VoI) such as the system status from a set of sensor measurements by supervised learning. However, in various complex manufacturing systems, components are operated under multiple modes that are not well known beforehand. The mapping of the VoI from sensor measurements highly depends on the mode information given that sensor measurements under different operation modes usually present different patterns. Therefore, predicting the VoI under multiple operation modes given sensor measurements is urgently necessary. This study develops a novel deep learning method for multimodal regression and mode recognition to predict the VoI under multiple modes and recognize the specific mode of a component from its sensor measurements. Specifically, we establish a deep neural network (DNN)-based regression- and classification-integrated framework. For model training, our innovative idea is to develop an Expectation–Maximum (EM)-based backpropagation algorithm, where the modes of components are set as latent variables, given that the mode information cannot be provided. Numerical experiments and a case study of degraded gas turbine engines are presented to validate the proposed model performance.

Data-driven aggregate modeling of a semiconductor wafer fab to predict WIP levels and cycle time distributions

In complex manufacturing systems, such as a semiconductor wafer fabrication facility (wafer fab), it is important to accurately predict cycle times and work-in-progress (WIP) levels. These key performance indicators are commonly predicted using detailed simulation models; however, the detailed simulation models are computationally expensive and have high development and maintenance costs. In this paper, we propose an aggregate modeling approach, where each work area, i.e., a group of functionally similar workstations, in the wafer fab is aggregated into a single-server queueing system. The parameters of the queueing system can be derived directly from arrival and departure data of that work area. To obtain fab-level predictions, our proposed methodology builds a network of aggregate models, where the network represents the entire fab consisting of different work areas. The viability of this method in practice is demonstrated by applying it to a real-world wafer fab. Experiments show that the proposed model can make accurate predictions, but also provide insights into the limitations of aggregate modeling.

Influence of Manufacturing Process Modularity on Lead Time Performances and Complexity

In principle, modular or integral character of manufacturing lines depends on the topological designs of products and determined operation tasks. On the other hand, in specific situations there is an articulated need for modular design in smart manufacturing systems since modular layouts are a crucial step towards agile production via smart manufacturing. The aim of this paper is to explore how the modular layout relates to manufacturing lead time (MLT) and to operational complexity of smart manufacturing systems. For this purpose, topologically different models of alternative process layouts were simulated and tested, while MLT values were obtained using Tecnomatix Plant Simulation. The obtained positive findings of this research could be useful not only in selection of the most suitable process design from the alternative ones, but especially in deepening the knowledge and bettering understanding of the concept of optimal network modularity.

Optimization of maintenance for complex manufacturing systems using stochastic Remaining Useful Life prognostics

This paper leverages Remaining Useful Life (RUL) prognostic information for preventive maintenance planning in complex manufacturing factories. Such a factory consists of multiple complex systems that use redundant components as backups to ensure system availability. The purpose is to minimize production fluctuations in the factory due to maintenance or system breakdown. To achieve this, we propose a Mixed-Integer Linear Programming (MILP) model that minimizes the overall production loss. Furthermore, we incorporate random RUL decrease rates according to reliability theory and extend the MILP model using chance-constrained programming. A novel approximation method for dealing with chance constraints is proposed to solve the stochastic model and approximation properties are analyzed. Computational results show that (i) our MILP model can provide more reasonable maintenance decisions compared to the literature. (ii) Sensitivity analysis reports that RUL prognostics and thresholds significantly impact maintenance decisions. (iii) The analysis of the impacts of confidence levels in chance constraints provides decision-makers with the flexibility to adjust maintenance planning under different levels of uncertainty.

Data-Based Estimation of the Dynamic Reliability and Performance Indicator of an Industrial Manufacturing System

The aim is to develop a more simple and effective method's performance and dynamic reliability assessment for complex industrial systems. By using the operating data of the industrial system characterized by a strong desynchronization and applying to it prediction algorithms of artificial intelligence applied to the time series, the model will have learned from the behavior of the complex manufacturing system allowing the operator or decision-maker to better orientate the maintenance, production, and quality policies. Furthermore, we propose this approach to avoid tedious mathematical methods related to dynamic reliability calculations and performance evaluation to make forecasts of the company's operation over a long period by identifying future bottlenecks in the system's behavior. The low-performance indicators and irrelevant reliability presented by many third-generation industries are due to the lack of efficient and simple tools for reliability assessment taking into account the dynamic aspect of the different elements of the production chain, maintenance department, production department, and quality department. We propose to develop a model that will abstract from conventional, complex, and inefficient mathematical methods for systems subject to combinatorial explosion problems in the manufacturing industry.

Design of Manufacturing Systems Based on Digital Shadow and Robust Engineering

In the era of digital transformation, industry is facing multiple challenges due to the need for implementation of the Industry 4.0 standards, as well as the volatility of customer demands. The latter has created the need for the design and operation of more complex manufacturing systems and networks. A case study derived from Process Industries (PIs) is adopted in this research work in order to design a framework for flexible design of production lines, automation of quality control points, and improvement of the performance of the manufacturing system. Therefore, a Digital Shadow of a production line is developed to collect, analyze and identify potential issues (bottlenecks). An edge computing system for reliable and low-latency communications is also implemented. The digital model is validated using statistical Design Of Experiments (DOE) and ANalysis Of VAriance (ANOVA). For the assessment of what-if scenarios, the Digital Shadow model will be used in order to evaluate and find the desired solution. Ultimately, the goal of this research work is to improve the design and performance of the industry’s production section, as well as to increase the production rate and the product mix.

Deep learning for predictive mechanical properties of hot-rolled strip in complex manufacturing systems

Deep learning for predictive mechanical properties of hot-rolled strip in complex manufacturing systems

Where to Place Monitoring Sensors for Improving Complex Manufacturing Systems? Discussing a Real Case in the Food Industry.

Industry 4.0 technologies offer manufacturing companies numerous tools to enhance their core processes, including monitoring and control. To optimize efficiency, it is crucial to effectively install monitoring sensors. This paper proposes a Multi-Criteria Decision-Making (MCDM) approach as a practical solution to the sensor placement problem in the food industry, having been applied to wine bottling line equipment at a real Italian winery. The approach helps decision-makers when discriminating within a set of alternatives based on multiple criteria. By evaluating the interconnections within the different equipment, the ideal locations of sensors are suggested, with the goal of improving the process's performance. The results indicated that the system of electric pumps, corker, conveyor, and capper had the most influence on the other equipment which are then recommended for sensor control. Monitoring this equipment will result in the early discovery of failures, potentially also involving other dependant equipment, contributing to enhance the level of performance for the whole bottling line.

A Discrete-Event Simheuristic for Solving a Realistic Storage Location Assignment Problem

In the context of increasing complexity in manufacturing and logistic systems, the combination of optimization and simulation can be considered a versatile tool for supporting managerial decision-making. An informed storage location assignment policy is key for improving warehouse operations, which play a vital role in the efficiency of supply chains. Traditional approaches in the literature to solve the storage location assignment problem present some limitations, such as excluding the stochastic variability of processes or the interaction among different warehouse activities. This work addresses those limitations by proposing a discrete-event simheuristic framework that ensures robust solutions in the face of real-life warehouse conditions. The approach followed embraces the complexity of the problem by integrating the order sequence and picking route in the solution construction and uses commercial simulation software to reduce the impact of stochastic events on the quality of the solution. The implementation of this type of novel methodology within a warehouse management system can enhance warehouse efficiency without requiring an increase in automation level. The method developed is tested under a number of computational experiments that show its convenience and point toward future lines of research.

Digital twin-enabled automated anomaly detection and bottleneck identification in complex manufacturing systems using a multi-agent approach

Digital twin (DT) models are increasingly being used to improve the performance of complex manufacturing systems. In this context, DTs automatically enabling anomaly detection, such as increase in orders, and bottleneck identification, such as shortage of products, can significantly enhance decision-making to mitigate the consequences of the identified bottlenecks. The existing literature has mainly focused on implementing top-down approaches for analysing the bottlenecks without considering the emergent behaviour of micro-level agents, including inventory levels and human resources, and their impact on the macro-level system’s performance. In order to handle the aforementioned challenges, this paper extends the current literature by proposing a novel DT integrated in a multi-agent cyber physical system (CPS) for detecting anomalies in sensor data, while identifying and removing bottlenecks that emerge during the operation of complex manufacturing systems. An extended 5 C CPS architecture, using multi-agent approach, is implemented to allow DT integration. The agent-based simulation technique enables capturing the probabilistic variability, and aggregate parallelism and dynamism of parallel dynamic interactions within the DT-CPS. A new single agent at the exo-level of the multi-level agent-based modelling structure, called the ‘monitoring agent’, is introduced in this research. The agent detects anomalies and identify bottlenecks through communicating with other agents in different levels automatically. The DT-CPS provides feedback automatically to the physical space to remove and mitigate the identified bottlenecks. The proposed DT based multi-agent CPS has been tested successfully on a real case study in a cryogenic warehouse shop-floor from the cell and gene therapy industry. The performance of the studied cryogenic warehouse is continuously measured using real-time sensor data. The analyses of the results show that the proposed DT-CPS improves the utilisation rates of human resources, on average, by 30% supporting decision making and control in complex manufacturing systems.

Predicting future production system bottlenecks with a graph neural network approach

Predicting throughput bottlenecks has received an increasing research attention in the recent years. Several statistical and machine learning models have been developed, with the objective of identifying future bottleneck locations for more efficient and cost-saving productions. However, bottleneck prediction remains a challenging task due to the complexity of manufacturing systems and the interactive effects from many impacting factors. This paper proposed the BSTAN (bottleneck spatial-temporal attention network) – an interpretable modeling framework to improve system bottlenecks prediction capability for complex manufacturing systems by: (1) characterizing line topological features using graph representations; (2) modeling spatial-temporal relationships and interactions between stations using graph attention network with gated recurrent units. In this proposed framework, the production line is first transformed into the representation of a weighted undirected graph to depict network structural information of a production line. An attention-based graph neural network with gated recurrent units is applied to capture both the temporal trends and station interactions to predict each station’s blockage and starvation. Future bottleneck stations are identified by analyzing the forecasted blockage and starvation distribution. To evaluate the model performance, an industrial case study is performed with data collected from a one-year production at an automotive powertrain assembly line. With the proper model construction, BSTAN has outperformed both statistical and machine learning benchmark models, achieving up to 15% lower root mean square error compared to the best benchmark method in predicting overall system blockage and starvation time.

Simultaneous predictive maintenance and inventory policy in a continuously monitoring system using simulation optimization

Predictive maintenance (PdM) is a strategy which can help determine the appropriate timing for maintenance depending on the actual operational conditions of a manufacturing system. Many relevant studies have ignored the effects of resource dependency and only considered perfect maintenance to reduce the complexity and uncertainty of maintenance problems. In this study, we develop a novel and highly practical maintenance model to fill the gap. The proposed model simultaneously considers the predictive maintenance and inventory policies of spare parts. In light of the high degree of complexity in modern manufacturing systems and the profound stochasticity in component degradation, a metamodeling-based simulation optimization method is proposed to find the optimal component inventory policy and degradation level thresholds of each component. A numerical study is conducted to verify that the proposed model and solution method can find the optimal or nearly optimal maintenance cost effectively and efficiently. Furthermore, the impact of critical factors in the maintenance model on the optimal policy is analyzed and useful managerial insights are derived.

Constraint programming and logic-based Benders decomposition for the integrated process planning and scheduling problem

The integrated process planning and scheduling (IPPS) problem is of critical importance in achieving desirable performance for complex manufacturing systems. The IPPS problem is often categorized into two types, i.e., Type-I and Type-II, depending on how the process plan is represented. In recent years, several approaches have been proposed to solve the IPPS problem in the literature. However, due to the complexity of the problem, optimal solutions of some benchmark datasets still cannot be obtained in a reasonable time, and few of them can be used to simultaneously address both types of IPPS problem. To this end, this study constructs a constraint programming (CP) model considering both types of IPPS problem, and proposes two basic logic-based Benders decomposition (LBBD) algorithms: one for each type of IPPS problem. In order to ensure computational efficiency, an enhanced LBBD algorithm is designed for both types of IPPS problem with three effective enhancement strategies. The performance of proposed methods is rigorously evaluated and compared with the existing approaches in the literature based on thirteen datasets. The results show that our methods significantly outperform these approaches.

An assessment method of machine components for usage-based value creation in complex manufacturing systems

Rising demands on manufacturing technologies lead to an increase in the complexity of corresponding machines and systems. This complexity acts as an obstacle to the development and realization of digital, usage-based business models for complex manufacturing systems. Particularly in the case of pay-per-use business models, it is necessary to monitor the actual machine load in operation to minimize the business risk on the operator and supplier side. This complicates the transition towards data-oriented value creation in manufacturing system environment. To overcome this challenge, a novel method to assess machine components regarding their potential to monitor the actual load and usage behavior of the whole manufacturing system is proposed. On this basis, the transparency and therefore confidence in digital, usage-based value creation like pay-per-use business models can be increased. The resulting approach is demonstrated by taking the example of a sheet metal forming press.

A Real-Time Human-in-the-Loop Control Method for Complex Systems

Real-time control tasks are of great importance in complex production and manufacturing systems since real-time operation may not proceed as pre-planned mostly. In this paper, we focus on a realtime control problem in a complex system which mainly consists of decision-makers, operators, and mobile machines. Each decision-maker takes charge part of the machines and controls their visiting sequences towards an operator in real time with respect to a variety of practical constraints. Although massive sensing data is available, decision-makers may not be able to process all the data and control too many machines simultaneously in real time owing to their limited computing ability and working memory. To address the problem, we design a human-in-the-loop control method in which decision-makers only need to supervise the control actions generated by a controller and can change these actions based on their real-time observations and judgements. The designed method contains four major parts which are the data processing and checking component, the controller, the exception warning component, and the controller and decision-maker interaction component. We develop a decision support tool based on the proposed method and apply it in a real-life application of controlling the truck arriving sequences at the quay cranes of a container terminal during the loading operation.

An approach to Quantum Annealing-based workforce assignment in manufacturing systems

Workforce assignment (WA) is a crucial aspect of production planning and control. However, current global crises have led to workforce absence and skill shortages. To tackle these challenges, a short-term WA approach is necessary for manufacturing systems flexibility and workforce health. Conventional methods for WA face complexity issues due to complex manufacturing systems, requiring substantial computation time. A Quantum Annealing-based WA approach is proposed in this paper to allocate tasks and shifts to workers. The proposed method is validated in a use case, showing improved solution quality compared to traditional methods.

A digital twin based framework for detection, diagnosis, and improvement of throughput bottlenecks

Digitalization through Industry 4.0 technologies is one of the essential steps for the complete collaboration, communication, and integration of heterogeneous resources in a manufacturing organization towards improving manufacturing performance. One of the ways is to measure the effective utilization of critical resources, also known as bottlenecks. Finding such critical resources in a manufacturing system has been a significant focus of manufacturing research for several decades. However, finding a bottleneck in a complex manufacturing system is difficult due to the interdependencies and interactions of many resources. In this work, a digital twin framework is developed to detect, diagnose, and improve bottleneck resources using utilization-based bottleneck analysis, process mining, and diagnostic analytics. Unlike existing bottleneck detection methods, this novel approach is capable of directly utilizing enterprise data from multiple levels, namely production planning, process execution, and asset monitoring, to generate event-log which can be fed into a digital twin. This enables not only the detection and diagnosis of bottleneck resources, but also validation of various what-if improvement scenarios. The digital twin itself is generated through process mining techniques, which can extract the main process map from a complex system. The results show that the utilization can detect both sole and shifting bottlenecks in a complex manufacturing system. Diagnosing and managing bottleneck resources through the proposed approach yielded a minimum throughput improvement of 10% in a real factory setting. The concept of a custom digital twin for a specific context and goal opens many new possibilities for studying the strong interaction of multi-source data and decision-making in a manufacturing system. This methodology also has the potential to be exploited for multi-objective optimization of bottleneck resources.

Digital Twin-Based Analysis and Optimization for Design and Planning of Production Lines

With the increasing dynamic nature of customer demand, production, product, and manufacturing design changes have become more frequent. Moreover, inadequate validation during the manufacturing design phase may result in additional issues, such as process redesign and layout reallocation, during the operation phase. Therefore, systems that can pre-validate and allow accurate and reliable analysis in the manufacturing design phase, as well as apply and optimize variations in production lines in real time, are required. Previously, digital twin (DT) has been studied a lot in product design and facility prognostics and management fields. Research on the system framework leading to DT utilization and optimization and analysis through DT in complex manufacturing systems with continuous processes such as production lines is insufficient. In this study, a system based on a DT and simulation results is developed; this system can reflect, analyze, and optimize dynamic changes in the design of processes and production lines in real time. First, the framework and application of the proposed system are designed. Subsequently, optimization methodologies based on heuristics and reinforcement learning (RL) are developed. Finally, the effectiveness and applicability of the proposed system are verified by implementing an actual DT application at a real manufacturing site.

Enhancing Antifragile Performance of Manufacturing Systems through Predictive Maintenance

Antifragility was introduced as a term no later than 10 years ago. As presented by Taleb, antifragility means that a system becomes more resilient and more robust with every harmful and/or stressful action it is confronted with. This paper is based on a study which aimed to use the concept of antifragility during the design stage of a self-improving system. This way, it is expected to obtain a fast adaptive system capable of functioning at optimal parameters even when it works under adverse conditions or faces unforeseen changes in the environment. Assuming that an antifragile system not only maintains its robust behavior when faced with stressful and harmful events but even benefits from them to optimize its performance, the paper offers a detailed description of the features that must be ensured when designing a self-improving antifragile manufacturing system. By ensuring the property of antifragility, complex manufacturing systems are much safer to exploit under uncertain conditions, which brings major benefits to the process management. Starting from consecrated solutions such as preventive maintenance (PvM) and predictive maintenance (PdM) and using techniques of artificial intelligence, we present the concept of antifragile maintenance (AfM).

Self-organized criticality in manufacturing system unscheduled downtime series and its application to major failure prediction

In this study, self-organized criticality theory was applied to analyze the unscheduled downtime (USDT) characteristic of a complex manufacturing system. The long-range correlation of the USDT series was verified through the Hurst exponent value. Then, we developed a model to depict the relationship between the frequency and USDT duration with the extreme value theory. The parameters of the model were estimated with the maximum likelihood method and updated with Bayesian inference when new data was available. Finally, the effectiveness of the proposed method was validated by the USDT data of the bottleneck system from a semiconductor assembly and test factory.