Permanent Production Loss Research Articles

Dynamic modeling and real-time performance analysis of multi-product batch manufacturing systems with perishable products

Production lines handling perishable goods in industries such as petrochemical and food processing demand meticulous precision in timing. These systems are governed by stringent timeframes, where products must strictly adhere to defined residence times. Exceeding these time constraints leads to product scrapping, posing a significant threat to overall productivity. Modeling such intricate systems presents a formidable challenge, and success hinges upon a deep understanding of the complex interplay between production and batch scrapping dynamics. In response to this challenge, this paper presents a novel model for a multiproduct batch production line with perishable products. Furthermore, the real-time performance metric of Permanent Production Loss (PPL) is extended to this system using a method of PPL back-adjustment in time. To validate the fidelity of the proposed mathematical model, numerical case studies are conducted, comparing the developed model to a discrete event simulation. The results demonstrate that the model accurately represents the system at hand. The expressions for PPL evaluation and attribution are also validated using case studies.

Multi-agent reinforcement learning for integrated manufacturing system-process control

The increasing complexity, adaptability, and interconnections inherent in modern manufacturing systems have spurred a demand for integrated methodologies to boost productivity, improve quality, and streamline operations across the entire system. This paper introduces a holistic system-process modeling and control approach, utilizing a Multi-Agent Reinforcement Learning (MARL) based integrated control scheme to optimize system yields. The key innovation of this work lies in integrating the theoretical development of manufacturing system-process property understanding with enhanced MARL-based control strategies, thereby improving system dynamics comprehension. This, in turn, enhances informed decision-making and contributes to overall efficiency improvements. In addition, we present two innovative MARL algorithms: the credit-assigned multi-agent actor-attention-critic (C-MAAC) and the physics-guided multi-agent actor-attention-critic (P-MAAC), each designed to capture the individual contributions of agents within the system. C-MAAC extracts global information via parallel-trained attention blocks, whereas P-MAAC embeds system dynamics through permanent production loss (PPL) attribution. Numerical experiments underscore the efficacy of our MARL-based control scheme, particularly highlighting the superior training and execution performance of C-MAAC and P-MAAC. Notably, P-MAAC achieves rapid convergence and exhibits remarkable robustness against environmental variations, validating the proposed approach’s practical relevance and effectiveness.

From Nash Q-learning to nash-MADDPG: Advancements in multiagent control for multiproduct flexible manufacturing systems

The emergence of flexible manufacturing systems (FMS) capable of processing multiple product types is a result of the growing demand for product customization and personalization. Such multiproduct systems are characterized by a higher level of uncertainty and variability when compared to traditional manufacturing systems. This paper proposes a Nash integrated multiagent deep deterministic policy gradient method (Nash-MADDPG) to control the mobile robots’ assignment in a multiproduct FMS to enable intelligent decision-making, interaction, and dynamic learning capabilities. A mathematical model of a multiproduct FMS from a previous study is described, and system dynamic property is characterized by permanent production loss (PPL). Then, by observing PPL of the system and market demand for each product type, the multi-agent control scheme is developed to assign mobile robots to load/unload various product types at various machines. First, a Nash game is developed among the mobile robots and to improve the cooperation, a collaboration cost is defined. This collaboration cost is then used in the reward function of the multiagent deep deterministic policy gradient (MADDPG) algorithm. Second, the actions are jointly defined based on the action values of MADDPG, and the mobile robots’ strategies in the Nash game, which update the environment to a new state. The performance of the proposed method is verified by comparing it with conventional Nash Q-learning, vanilla MADDPG, Q-learning based single agent reinforcement learning (SARL) and a first-come-first-serve (FCFS) based control. The results demonstrate that the multi-agent control scheme under the proposed Nash-MADDPG is effective in dealing with cooperative FMS environment that involves complicated dynamics and uncertainties.

RETRACTED: Data-driven model predictive control for real-time planned lead time optimization in a reconfigurable flow line

Mass customization has enabled smart manufacturing systems to develop reconfigurable machines and fixtures to fulfill customized products with high process variety. Coupled with these reconfigurable components, diversified design and operational problems should be dealt with such that the efficiency of reconfigurations can be maximized. This paper investigates a planned lead time (PLT) optimization problem after addressing the pre-determined reconfigurations and real-time disruptions in a flow line. A data-driven model predictive control (d-MPC) is developed for this problem such that the total expected cost for work-in-process inventory, final-product inventory, and tardiness penalty can be minimized. Considering the predictive multi-stage transitions and reconfigurable behaviors, d-MPC utilizes max-plus algebra to generate a time-varying state-space equation for a flow line. In this regard, a systematic method for permanent production loss identification can be formulated for real-time disruption diagnosis. This identification is useful in providing feedback signals for d-MPC and thus a closed-loop feedback control for PLT optimization can be achieved. A case study of an industrial welding line is reported to illustrate the operational benefits of d-MPC. The results demonstrate that d-MPC can efficiently cooperate with PLT optimization with machine reconfiguration such that the total expected cost can be minimized. Compared with continuous re-optimization, d-MPC adopts an event-driven mechanism to respond to real-time disruptions and can decrease the nervousness of flow lines with lower operational costs.

Adaptive Mobile Robot Scheduling in Multiproduct Flexible Manufacturing Systems Using Reinforcement Learning

Abstract The integration of mobile robots in material handling in flexible manufacturing systems (FMS) is made possible by the recent advancements in Industry 4.0 and industrial artificial intelligence. However, effectively scheduling these robots in real-time remains a challenge due to the constantly changing, complex, and uncertain nature of the shop floor environment. Therefore, this paper studies the robot scheduling problem for a multiproduct FMS using a mobile robot for loading/unloading parts among machines and buffers. The problem is formulated as a Markov Decision Process, and the Q-learning algorithm is used to find an optimal policy for the robot's movements in handling different product types. The performance of the system is evaluated using a reward function based on permanent production loss and the cost of demand dissatisfaction. The proposed approach is validated through a numerical case study that compares the proposed policy to a similar policy with different reward function and the first-come-first-served policy, showing a significant improvement in production throughput of approximately 23%.

Dynamic modeling and analysis of multi-product flexible production line

ABSTRACT Mass customization has been a major challenge in the manufacturing sector. For this purpose, Flexible Manufacturing Systems (FMS) are usually designed to accommodate variations and manufacture different types of products in batches, but the products follow a single rigid path using a conveyor and in case of a failure on a single machine, the whole production line is affected. However, a multi-product dynamic production system can operate under an elongated failure on a single machine. A novel dynamic modeling method for a multi-product production line is developed to investigate dynamic properties of the system under random disruptions such as machine failures and market demand changes. Permanent production loss (PPL) and demand dissatisfaction for multi-production lines are formally defined as real-time performance metrics. A real-time analysis method is developed to evaluate the PPL and demand dissatisfaction. Numerical case studies are presented to validate the fidelity of the real-time multi-product production line model and the effectiveness of the real-time analysis method.

Production Loss Analysis in Mobile Multi-skilled Robot Operated Flexible Serial Production Systems

In a world where demand and supply can change instantaneously as demonstrated by the recent coronavirus pandemic, it is very important for our production lines to be able to handle abrupt changes very effectively. Flexible manufacturing systems (FMS) are designed to be able to ramp production up and down quickly to make this possible. In this paper, we propose a novel data-enabled method of modelling an FMS using mobile multi-skilled robots and evaluate its dynamic performance. We use permanent production loss as the performance metric and derive expressions for its evaluation and attribution and validate them using two case studies.

Dynamic Robot Assignment for Flexible Serial Production Systems

This letter aims at modeling and real-time control of a flexible manufacturing system (FMS) that is operated by multi-skilled mobile robots. By introducing the idea of a unique ideal clean configuration and the effective disruption event, the concepts of Opportunity Window (OW) and Permanent Production Loss (PPL) are extended to the FMS. The control considered in this letter is a robot assignment problem where individual robots are dynamically assigned to each workstation in real time to improve performance. The problem is formulated as a Markov Decision Process (MDP) and solved using the Double Deep Q-Network (DDQN) reinforcement learning algorithm. PPL is used as the reward setting for training and a space discretization is used between each workstation to capture real time movement of the robots on the plant floor. The effectiveness of the control strategy is then demonstrated by comparing the results to several heuristic control schemes.

Integrated Process-System Modeling and Performance Analysis for Serial Production Lines

The performance of a smart manufacturing system is affected by not only the constituent processes but also their system-level interactions. However, in most current studies, individual process modeling and system-level performance evaluation are independent. This can substantially impact production efficiency. In this paper, utilizing available sensor data, an integrated data-enabled model is developed to seamlessly fuse two conventionally separated system-level and process-level models and analysis. A fast recursive method is developed to evaluate the system yield. The permanent production loss (PPL) concept is defined and evaluated based on the proposed integrated model. Furthermore, PPL attributions due to random downtime and quality issues have been identified. Case studies have shown that the integrated model is of high fidelity, and the PPL analysis can effectively identify the root cause of production yield loss.

Data-Enabled Permanent Production Loss Analysis for Serial Production Systems With Variable Cycle Time Machines

Real time production performance evaluation plays a vital role in diagnosing manufacturing system health status and achieving productivity improvements. However, most existing studies on system performance evaluation are based on steady state analysis and focused on the production system with fixed cycle time machines. The real-time performance evaluation for a manufacturing system with variable cycle time machines, although typical for a large number of realistic scenarios, has been mostly ignored. The development of smart manufacturing and increasingly available sensor data have provided unprecedented opportunities to carry out thorough analysis on the real-time performance of such complex systems. In this letter, we developed a data-enabled methodology to efficiently identify and predict the real-time permanent production loss for a serial production line with variable cycle time machines. The concept and evaluation method of opportunity window are introduced to facilitate the permanent production loss estimation. Numerical case studies are presented to demonstrate the effectiveness of the proposed methods for opportunity window evaluation and production loss identification.

Data-Enabled Modeling and Analysis of Multistage Manufacturing Systems with Quality Rework Loops

In a multistage serial production line, multiple inspection stations and repair processes are typically involved to ensure high product quality. Quality rework is the activity to repair or repeat the work on the defect parts during manufacturing processes. The rework process after each inspection can add cost and cycle time to the normal process and impose negative impact on the throughput. This paper studies real-time performance of multistage serial manufacturing systems with quality rework loops and machine random failures. A production line with multiple quality rework loops is first unify by segmenting it into a set of serially connected quality rework loops. An event-based data-enabled mathematical model is developed to evaluate real-time production rate of each machine for such a system. In addition, the system properties are analyzed and permanent production loss due to quality rework loops and random machine failures are identified respectively. The permanent production loss attribution to each disruption event and machine can be used as real-time performance indicators to diagnose production system inefficiency. The mathematical model and system performance identification methodology are studied analytically and validated through numerical case studies.

Event-Based Modeling and Analysis of Sensor Enabled Networked Manufacturing Systems

Mathematical models are often required in process modeling, control, and evaluation. To meet the increasing demand of production system real-time performance evaluation and improvement, in this paper: 1) an innovative event-based, data-driven mathematical model is established for network structured manufacturing systems; 2) important properties of network structured manufacturing systems are obtained through the concept development of a virtual ideal clean system; and 3) a system performance diagnostic method is developed based on the mathematical model and system properties, as well as available sensor data. The mathematical model and system performance identification methodology are studied analytically and validated by simulation studies. This data-driven mathematical framework and system real-time performance diagnostic methodology are invaluable for real-time production control to improve system responsiveness and efficiency. Note to Practitioners —For production systems with complex networked structure layouts, the dependence relationship between machines and buffers is more complicated compared with serial production lines. Therefore, it is more difficult to analyze and evaluate their real-time production performance such as permanent production loss. Such information is critical for real-time production management and control to achieve higher system responsiveness and efficiency. This paper establishes an event-based data-driven mathematical model to describe the real-time dynamic behavior of manufacturing systems with complex networked structures. Furthermore, an analysis method for networked manufacturing system properties and a data-driven system performance diagnostic method are proposed. The methods provide a quantitative solution to evaluate production capacity and production constraints of complex networked manufacturing systems, and to evaluate the system real-time performance such as permanent production loss and its attribution to each disruption event and machine.

Modeling and Performance Diagnostics of Composite Work Cells With Gantries

In this paper, the performance of a typical composite work cell structure that includes multiple machines and a material handling gantry is studied. A mathematic model is established to analyze the performance of such gantry systems. Due to the unique features of the gantry system, cycle time and waiting time of the gantry and machines are discussed and formulated based on two basic scenarios. To measure the impact of disruption events, permanent production loss is evaluated by using the sensor data collected from plant floors. Data-driven diagnostic methods are developed to identify the bottleneck machine and to evaluate permanent production loss attribution of each machine. The results provide a solid base to efficiently and effectively improve the performance of a composite work system. Performance diagnostics of gantry systems provides a theoretical and practical basis for production improvement. A gantry system for composite lay-up processes is used for numerical simulation case studies, by which the performance diagnostic methodologies and improvements are demonstrated. Note to Practitioners —Accurate performance diagnostics and effective improvement methods are critical to manufacturing operation and management. Most existing efforts are devoted to analyzing the performance of a whole production line rather than considering the complexity of each station/work cell in the line. This paper zooms into a composite work cell with a material handling gantry and focuses on the performance diagnostics of the lay-up station in a composite work center. The methods presented in this paper can be extended and applied to other manufacturing work cells comprising the similar gantry structure. We build a mathematical model to analyze the characteristics of the gantry system such as gantry cycle and waiting time. The evaluation of system production loss, which directly adopts sensor data from the factory floors, provides a quantitative tool for production engineers to monitor and control the system in real-time operation. In addition, the performance diagnostic methods based on production loss attribution and bottleneck identification are developed to guide the decision makers to improve the production system under different resource conditions. The methods are developed based on a work cell with only one gantry under a fixed gantry moving sequence. In the future research, we will extend the study to the systems with more gantries and more complex gantry moving policies.

Production System Performance Identification Using Sensor Data

Advanced manufacturing systems are characterized by their complex dynamics, which are subject to constant changes caused by technology insertion, engineering modification, as well as disruption events. To support daily operation, distributed sensors are utilized to monitor the status of each process. The advantages of the sensor data are not fully realized in current production system analysis due to a lack of data-driven system level modeling. Motivated by this need, we propose a data-driven manufacturing system model to describe production dynamics and develop a systematic method to identify the causes of permanent production loss. This research enables real-time production system performance diagnosis, which is invaluable in increasing system responsiveness and improving real-time production control to effectively enhance overall system efficiency.

Gantry Work Cell Scheduling through Reinforcement Learning with Knowledge-guided Reward Setting

In this paper, a manufacturing work cell utilizing gantries to move between machines for loading and unloading materials/parts is considered. The production performance of the gantry work cell highly depends on the gantry movements in real operation. This paper formulates the gantry scheduling problem as a reinforcement learning problem, in which an optimal gantry moving policy is solved to maximize the system output. The problem is carried out by the Q-learning algorithm. The gantry system is analyzed and its real-time performance is evaluated by permanent production loss and production loss risk, which provide a theoretical base for defining reward function in the Q-learning algorithm. A numerical study is performed to demonstrate the effectiveness of the proposed policy by comparing with the first-come-first-served policy.

Gantry Scheduling for Multi-Gantry Production System by Online Task Allocation Method

For a multi-gantry production system, in which several gantries move among the machines and load/unload parts, the assignment of gantries directly impacts the system performance. Gantry scheduling problem is extremely important, involving when and where the gantries are assigned to the machines. This letter formulates the gantry scheduling problem as an online task allocation (OTA) problem, and solves for an optimal gantry assignment policy to maximize the system throughput. The multi-gantry production system is analyzed and important system properties are derived. The production system performance is evaluated through permanent production loss and opportunity window. Such knowledge is used to define the key parameters in the OTA problem. Based on the analysis of the system production performance, the optimization problem is solved and consequently a gantry assignment policy is proposed. A numerical study is performed to demonstrate the effectiveness of the proposed policy by comparing with first-come-first-served policy.

Dynamic production system diagnosis and prognosis using model-based data-driven method

Advanced manufacturing systems are becoming increasingly complex, subjecting to constant changes driven by fluctuating market demands, new technology insertion, as well as random disruption events. While information about production processes has been becoming increasingly transparent, detailed, and real-time, the utilization of this information for real-time manufacturing analysis and decision-making has been lagging behind largely due to the limitation of the traditional methodologies for production system analysis, and a lack of real-time manufacturing processes modeling approach and real-time performance identification method. In this paper, a novel data-driven stochastic manufacturing system model is proposed to describe production dynamics and a systematic method is developed to identify the causes of permanent production loss in both deterministic and stochastic scenarios. The proposed methods integrate available sensor data with the knowledge of production system physical properties. Such methods can be transferred to a computer for system self-diagnosis/prognosis to provide users with deeper understanding of the underlying relationships between system status and performance, and to facilitate real-time production control and decision making. This effort is a step forward to smart manufacturing for system real-time performance identification in achieving improved system responsiveness and efficiency.

An event-based analysis of condition-based maintenance decision-making in multistage production systems

Condition-based maintenance (CBM) is becoming increasingly prevalent because of its capability to continuously track equipment health degradation and accurately predict unscheduled equipment failure. CBM helps to improve the business bottom line by preventing costly station failure. However, it is not uncommon that CBM needs to stop stations for maintenance during operation, which can severely impede the normal production. The objective of this paper is to develop a systematic method to predict the negative impact of CBM stoppage events on production in a multistage manufacturing system. The research helps to predict the real expense of applying CBM, which is the foundation to establish a comprehensive real-time CBM decision-making model. We start from the event-based analysis of system dynamics and develop a stochastic estimation method to predict the permanent production loss caused by a CBM stoppage event. The monotonicity property of permanent production loss is investigated. Simulation case studies are performed to illustrate the theoretical results and demonstrate their potential in facilitating CBM decision-making.

Production Loss Diagnosis and Prognosis Using Model-based Data-driven Method

Advanced sensing and control are strategically as important to manufacturing as are design and new product. Through the application of advanced sensor technologies, more detailed information about manufacturing plants is obtained. Although evaluation and improvement have been made on individual machine performance based on sensor data, there is still no data-driven system model for the identification of system inefficiencies and process optimization. In this paper, we build a mathematic model based on sensor data to describe the production dynamics. Furthermore, a systematic methodology is further developed to perform real-time diagnosis and prognosis on permanent production loss and conduct analysis on system robustness to disruption events, which can enhance overall system efficiency by increasing system responsiveness.

Event-based modelling of distributed sensor networks in battery manufacturing

Battery manufacturing systems are characterised by their complex dynamics subject to constant changes caused by technology insertion, engineering modifications, as well as disruption events. To support daily operation, distributed sensors are used to provide real-time data describing the status of each process. Despite the big potential in improving productivity, the advantages of distributed sensor networks are not fully realised for overall system efficiency due to a lack of system-level modelling. Motivated by this need, we develop an event-based modelling (EBM) approach to quantify the systematic impacts of stations and supporting activities with an index called permanent production loss. EBM instantaneously captures the system dynamics using distributed sensor information and provides a severity ranking of stations and supporting activities. We also study the system dynamics of serial production lines with multiple slowest stations and quantify the impacts of disruption events to the production system. A case study is conducted to demonstrate the application of EBM in a battery production system and its ability to facilitate decision-making at plant floor on resources and budget allocation.