Multi-stage Assembly Systems Research Articles

A Sparse Bayesian Learning for Diagnosis of Nonstationary and Spatially Correlated Faults With Application to Multistation Assembly Systems

A Sparse Bayesian Learning for Diagnosis of Nonstationary and Spatially Correlated Faults With Application to Multistation Assembly Systems

A novel sparse Bayesian learning and its application to fault diagnosis for multistation assembly systems

This article addresses the problem of fault diagnosis in multistation assembly systems. Fault diagnosis is to identify process faults that cause excessive dimensional variation of the product using dimensional measurements. For such problems, the challenge is solving an underdetermined system caused by a common phenomenon in practice; namely, the number of measurements is less than that of the process errors. To address this challenge, this article attempts to solve the following two problems: (i) how to utilize the temporal correlation in the time series data of each process error and (ii) how to apply prior knowledge regarding which process errors are more likely to be process faults. A novel sparse Bayesian learning method is proposed to achieve the above objectives. The method consists of three hierarchical layers. The first layer has parameterized prior distribution that exploits the temporal correlation of each process error. Furthermore, the second and third layers achieve the prior distribution representing the prior knowledge of process faults. Since posterior distributions of process faults are intractable, this article derives approximate posterior distributions via Variational Bayes inference. Numerical and simulation case studies using an actual autobody assembly process are performed to demonstrate the effectiveness of the proposed method.

Control strategies comparison for a multi-stage assembly system using simulation

The goal of Zero Defect Manufacturing (ZDM) is to reduce part scrap by compensating manufacturing deviations and avoiding their propagation on multi-stage production systems. In particular, the joint use of in-process measuring and flexible locating systems enables the application of different control strategies to proactively improve the geometrical quality of the product and to achieve the productivity targets. In this work, focused on the production quality paradigm, two compensation strategies based on the adjustment of part locators are simulated to compare how different control loops can reduce the out-of-control parts in multi-stage assembly systems. For this purpose, a simulation tool developed on OpenModelica is used taking into account the logistic flow and the quality characteristics flow based on the Stream of Variation model.

Building a Scalable and Interpretable Bayesian Deep Learning Framework for Quality Control of Free Form Surfaces

Deep learning has demonstrated high accuracy for 3D object shape error modeling necessary to estimate dimensional and geometric quality defects in multi-station assembly systems (MAS). Increasingly, deep learning-driven Root Cause Analysis (RCA) is used for decision-making when planning corrective action of quality defects. However, given the current absence of scalability enabling models, training deep learning models for each individual MAS is exceedingly time-consuming as it requires large amounts of labelled data and multiple computational cycles. Additionally, understanding and interpreting how deep learning produces final predictions while quantifying various uncertainties also remains a fundamental challenge. In an effort to address these gaps, a novel closed-loop in-process (CLIP) diagnostic framework underpinned algorithm portfolio is proposed which simultaneously enhances scalability and interpretability of the current Bayesian deep learning approach, Object Shape Error Response (OSER), to isolate root cause(s) of quality defects in MAS. The OSER-MAS leverages a Bayesian 3D U-Net architecture integrated with Computer-Aided Engineering simulations to estimate root causes. The CLIP diagnostic framework shortens OSER-MAS model training time by developing: (i) closed-loop training to enable faster convergence for a single MAS by leveraging uncertainty estimates of the Bayesian 3D U-net model; and, (ii) transfer/continual learning-based scalability model to transmit meta-knowledge from the trained model to a new MAS resulting in convergence using comparatively less training samples. Additionally, CLIP increases the transparency for quality-related root cause predictions by developing interpretability model which is based on 3D Gradient-based Class Activation Maps (3D Grad-CAMs) and entails: (a) linking elements of MAS model with functional elements of the U-Net architecture; and, (b) relating features extracted by the architecture with elements of the MAS model and further with the object shape error patterns for root cause(s) that occur in MAS. Benchmarking studies are conducted using six automotive-MAS with varying complexities. Results highlight a reduction in training samples of up to 56% with a loss in performance of up to 2.1%.

Production planning and control in multi-stage assembly systems: an assessment of Kanban, MRP, OPT (DBR) and DDMRP by simulation

Multi-stage assembly systems where the demand for components depends on the market-driven demand for end products, are commonly encountered in practice. Production Planning and Control (PPC) systems for this production context include Kanban, Materials Requirement Planning (MRP), Optimised Production Technology (OPT), and Demand Driven MRP (DDMRP). All four of these PPC systems are widely applied in practice and literature abounds on each of these systems. Yet, studies comparing these systems are scarce and remain largely inconclusive. In response, this study uses simulation to assess the performance of all four PPC systems under different levels of bottleneck severity and due date tightness. Results show that MRP performs the worst, which can be explained by the enforcement of production start dates. Meanwhile, Kanban and DDMRP perform the best if there is no bottleneck. If there is a bottleneck then DDMRP and OPT perform the best, with DDMRP realising lower inventory levels. If there is a severe bottleneck, then the performance results for DDMRP and OPT converge. This identification of contingency factors not only resolves some of the inconsistencies in the literature but also has important implications for the applicability of these four PPC systems in practice.

Multidomain Simulation Model for Analysis of Geometric Variation and Productivity in Multi-Stage Assembly Systems

Nowadays, the new era of industry 4.0 is forcing manufacturers to develop models and methods for managing the geometric variation of a final product in complex manufacturing environments, such as multistage manufacturing systems. The stream of variation model has been successfully applied to manage product geometric variation in these systems, but there is a lack of research studying its application together with the material and order flow in the system. In this work, which is focused on the production quality paradigm in a model-based system engineering context, a digital prototype is proposed to integrate productivity and part quality based on the stream of variation analysis in multistage assembly systems. The prototype was modelled and simulated with OpenModelica tool exploiting the Modelica language capabilities for multidomain simulations and its synergy with SysML. A case study is presented to validate the potential applicability of the approach. The proposed model and the results show a promising potential for future developments aligned with the production quality paradigm.

Integration of in-plane and out-of-plane dimensional variation in multi-station assembly process for automotive body assembly

Automotive body assembly systems contain multiple operations in multi-station processes. One of the most critical challenges for such manufacturing systems is dimensional quality, which is affected by the accumulation and propagation of variation caused by manufacturing imperfections. However, sheet metal part compliancy behavior makes the variation modeling method extremely intricate when both rigid (in-plane) and compliant (out-of-plane) variations are considered. This paper develops a more accurate variation propagation model to describe dimensional variation of sheet metal assembly in multi-station assembly system through involving both the variation types simultaneously as well as the impacts of assembly operations on each other. In this methodology, three sources of deviations—non-ideal parts, fixture errors, and assembly operations effects—are taken into account. The variation generated in every assembly operation (placing, clamping, fastening, and releasing steps) and the variation propagation through station-to-station interaction (repositioning) are analyzed by the transfer function mechanism. In the in-plane direction, the stream of variation analysis is adopted to obtain the rigid transfer function to describe the position and orientation relationships between part and assembly element errors. For the simulation of part deformation during the assembly process in the out-of-plane direction, the compliant transfer functions are extracted by variation response methodology. A nested state space model is used to integrate the overall assembly variation by updating part geometry after each assembly operation. The capability of proposed method is illustrated through a case study on an automotive body side assembly process.

A quality-driven assembly sequence planning and line configuration selection for non-ideal compliant structures assemblies

In automotive body assembly systems, an optimum assembly sequence planning (ASP) not only increases production efficiency and product quality, but also decreases cost and process cycle time. Typically, ASP evaluation approaches are focused on design for assembly criteria, and very few studies have considered the impact of ASP on dimensional accuracy. The major challenges involving quality-driven ASP evaluation can be enumerated into three categories: (1) batch of compliant non-ideal parts to consider real part defects; (2) variation propagation modeling in multi-station assembly (MSA) system in the presence of stochastic manufacturing errors both at product and process levels; and, (3) the development of dimensional quality criteria for quantitative ASP comparisons. This paper proposes a methodology based on the modeling of dimensional errors propagation in MSA with a batch of compliant non-ideal parts to improve product dimensional quality through optimizing ASP and assembly line configuration. It entails three main steps: (i) assembly sequence generation by k-ary assembly operation method for a predetermined assembly line configuration; (ii) variation propagation simulation taking into account a batch of non-ideal parts, station-to-station repositioning errors, and spring-back phenomenon in MSA system; and, (iii) robust optimization of ASP based on developed quality criteria which contains two quantitative indices. The potential benefits of the proposed methodology are successfully demonstrated on automotive front-rail assembly process.

A survey on data-driven process monitoring and diagnostic methods for variation reduction in multi-station assembly systems

PurposeDriven by the development in sensing techniques and information and communications technology, and their applications in the manufacturing system, data-driven quality control methods play an essential role in the quality improvement of assembly products. This paper aims to review the development of data-driven modeling methods for process monitoring and fault diagnosis in multi-station assembly systems. Furthermore, the authors discuss the applications of the methods proposed and present suggestions for future studies in data mining for quality control in product assembly.Design/methodology/approachThis paper provides an outline of data-driven process monitoring and fault diagnosis methods for reduction in variation. The development of statistical process monitoring techniques and diagnosis methods, such as pattern matching, estimation-based analysis and artificial intelligence-based diagnostics, is introduced.FindingsA classification structure for data-driven process control techniques and the limitations of their applications in multi-station assembly processes are discussed. From the perspective of the engineering requirements of real, dynamic, nonlinear and uncertain assembly systems, future trends in sensing system location, data mining and data fusion techniques for variation reduction are suggested.Originality/valueThis paper reveals the development of process monitoring and fault diagnosis techniques, and their applications in variation reduction in multi-station assembly.

Safety Times for Multistage Assembly System

Nowadays, a wide class of problems can be solved by using the classical newsboy model. However, in problems where uncertainty of events and randomness are omnipresent, there is a necessity to adapt the existing solutions and/or find new extensions that will properly answer all requirements. This paper considers a multistage assembly system where interrelated assembly operations with independent stochastic operation times should be planned in an optimal way. Delivery of items in a requested time implies that either delay costs or holding costs appear. The goal is to find optimal safety times. We propose an approximate technique based on successive application of the solution of simpler one-stage problem. The generalized mathematical model suggested is built up on the relaxed hypothesis and can be used in multistage assembly networks. The existence and uniqueness of the solution are proven. The preliminary tests are performed and our approximate technique is compared to exact results.

Quality-driven Optimization of Assembly Line Configuration for Multi-Station Assembly Systems with Compliant Non-ideal Sheet Metal Parts

Final product quality is affected by both defects generation as well as defects propagation through the multi-station assembly system. Less quantifiable understood is quality defects propagation which depends on assembly line configuration. This paper focuses on modelling and optimization of multi-station assembly line configuration for systems with compliant non-ideal sheet-metal parts. The proposed methodology is based on: (1) assembly sequence generation; (2) multi-station assembly variation propagation model and simulations considering batch of non-ideal parts, station-to-station repositioning error and spring-back phenomenon; (3) development of assembly configuration index; and, (4) calculation of optimal configuration based on stochastic robust optimization. The methodology is demonstrated using automotive front-rail subassembly process.

Fault pattern identification in multi-stage assembly processes with non-ideal sheet-metal parts based on reinforcement learning architecture

A reinforcement learning-based architecture to address the fault detection on body in white assembly processes is introduced in this paper. During the research were addressed: (i) generation of a random defect pattern database using a multi-physics variation modeling for multi-stage assembly systems; (ii) design and implementation of a fault pattern identification reinforcement learning-based architecture, combining neural network, genetic algorithm and Q-learning algorithms; and (iii) validation based on non-ideal sheet-metal parts case study generated by the Variation Response Method toolkit. Finally, a comparative study between the different topologies is done, highlighting the influence of the Q-learning in the default identification process.

Fault Diagnosis in Multistation Assembly Systems Using Spatially Correlated Bayesian Learning Algorithm

The problem of fault diagnosis for dimensional integrity in multistation assembly systems is addressed in this paper. Fault diagnosis under this context is to identify the process errors which significantly contribute to the large product dimensional variation based on sensor data. The main challenges to be resolved in this paper include (1) the number of measurements is less than the process errors, which is typical in practice, but results in an ill-posed estimation problem, and (2) there exists spatial correlation among the dimensional variation of process errors, which has not been addressed yet by existing literature. A spatially correlated Bayesian learning (SCBL) algorithm to address these challenges is developed. The SCBL algorithm is based on the relevance vector machine (RVM) by exploiting the spatial correlation of dimensional variation from various process errors, which occurs in some circumstances of assembled parts and is well defined in GD&T standards. The proposed algorithm relies on a parametrized prior including the spatial correlation, and eventually leads sparsity in fault diagnosis; hence, the issues with ill-posedness and structured process errors will be addressed. A number of simulation studies are performed to illustrate the superiority of SCBL algorithm over state-of-the-art algorithms in sparse estimation problems when spatial correlation exists among the nonzero elements. A real autobody assembly process is also used to demonstrate the effectiveness of proposed SCBL algorithm.

Rapid Response Diagnosis of Multi-stage Assembly Process with Compliant non-ideal Parts using Self-evolving Measurement System

This paper introduces the novel concept of self-evolving measurement system with the aim of rapidly identifying and localising defect patterns in multi-stage assembly systems with compliant non-ideal parts. This allows to enhance the level of diagnosability which cannot be achieved using fixed and static pre-determined measurement systems. The proposed methodology helps to identify and select new measurement points to increase the likelihood of isolating root causes of defects. This happens by automatically classifying defect patterns and associating them to critical key control characteristics. The methodology integrates supervised machine learning tools with first principle engineering simulations. It is based on the principle of pattern similarity, taking into account data generated by the self-evolving measurement system. The methodology is demonstrated and validated using the results of an automotive door assembly system.

Lead time control in multi-class multi-stage assembly systems with finite capacity

In this paper, we model multi-class multi-stage assembly systems with finite capacity as queueing networks. It is assumed that different classes (types) of products are produced by the production system and products’ orders for different classes are received according to independent Poisson processes. Each service station of the queueing network specifies a manufacturing or assembly operation, in that processing times for different types of products are independent and exponentially distributed random variables with service rates, which are controllable, and the queueing discipline is First Come First Served (FCFS). Different types of products may be different in their routing sequences of manufacturing and assembly operations. For modeling multi-class multi-stage assembly systems, we first consider every class separately and convert the queueing network of each class into an appropriate stochastic network. Then, by using the concept of continuous-time Markov processes, a system of differential equations is created to obtain the distribution function of manufacturing lead time for any type of product, which is actually the time between receiving the order and the delivery of finished product. Furthermore, we develop a multi-objective model with three conflicting objectives to optimally control the service rates, and use goal attainment method to solve a discrete-time approximation of the original multi-objective continuous-time problem.

Due-date assignment for multi-server multi-stage assembly systems

In this paper, we attempt to present a constant due-date assignment policy in a multi-server multi-stage assembly system. This system is modelled as a queuing network, where new product orders are entered into the system according to a Poisson process. It is assumed that only one type of product is produced by the production system and multi-servers can be settled in each service station. Each operation of every work is operated at a devoted service station with only one of the servers located at a node of the network based on first come, first served (FCFS) discipline, while the processing times are independent random variables with exponential distributions. It is also assumed that the transport times between each pair of service stations are independent random variables with generalised Erlang distributions. Each product's end result has a penalty cost that is some linear function of its due date and its actual lead time. The due date is calculated by adding a constant to the time that the order enters into the system. Indeed, this constant value is decided at the beginning of the time horizon and is the constant lead time that a product might expect between the time of placing the order and the time of delivery. For computing the due date, we first convert the queuing network into a stochastic network with exponentially distributed arc lengths. Then, by constructing an appropriate finite-state continuous-time Markov model, a system of differential equations is created to find the manufacturing lead-time distribution for any particular product, analytically. Finally, the constant due date for delivery time is obtained by using a linear function of its due date and minimising the expected aggregate cost per product.

Study on Performance Evaluation of Multi-Station Assembly System Based on Stream of Variation

Based on stream of variation theory, parametric state space model of multi-station assembly process was built up. For problems of system-level evaluation index, the sensitivity evaluation index, soundness index as three methods of assembly system performance evaluation, the comprehensive stability index was presented. Parametric state space model was as a tool and the comprehensive stability index was as evaluation index to optimize the assembly lines, reduce the geometric deviation of product, improve the quality of product. Experiment verified comprehensive stability index can accurately evaluate the performance of assembly systems, improve the efficiency of optimizing assembly lines.

Lead time control in multi-server multi-stage assembly system

In this paper, two multi-objective models to optimally control the lead time in multi-server multi-stage assembly system by considering server allocation problem (SAP) and also service rate control problem (SCP) are presented, where new product orders, including all their operations, are entered to the system according to a Poisson process, only one type of products also is produced by the production system, and multi-servers can be settled in each service station. Each operation of any work is operated at a devoted service station with only one of the servers located at a node of the network based on a first-come-first-serve discipline, while the processing times are independent random variables with exponential distributions. Furthermore, it is also assumed that the transport times between each pair of service stations are independent random variable with generalized Erlang distributions. Such system can be modeled as a queueing network, where the system is in the steady state and the lead time is controllable. For modeling of multi-server multi-stage assembly system, initially the network of queues is transformed into an appropriate stochastic network with exponentially distributed arc lengths. A differential equations system is organized to solve and obtain approximate lead time distribution for any particular wok by applying a proper finite-state continuous-time Markov model. Then, two multi-objective models including four conflicted objectives are presented to optimally control the servers allocated to the service stations in SAP and also service rate of service stations in SCP. For solving a discrete-time approximation of the primary multi-objective problems, the goal attainment technique is employed. In this research, reactive controlling in a multi-server multi-stage assembly system also is discussed.

Enhanced piecewise least squares approach for diagnosis of ill-conditioned multistation assembly with compliant parts

Assembly process-induced dimensional variation has a significant impact on product quality and functionality. The complexity of modern products coupled with part compliance/flexibility frequently results in ill-conditioned assembly systems further adding to the challenges of process control and fault failure diagnosis. This paper proposes an enhanced piecewise least squares (EPLS)-based approach for dimensional fault failure diagnosis of ill-conditioned multistation assembly systems. In this approach, predetermined fault patterns derived from an inverse stiffness matrix of assembly structures and fault patterns obtained from product measurement data are used to detect and isolate dimensional failures caused by fixturing error(s). The EPLS-based diagnostic methodology searches for a set of components called latent vectors with thesearch constrained by assembly response function, the stream of variation analysis (SOVA) model of an assembly system, which performs decomposition of response based on the end-of-line measurements. The verification of the proposed methodology is conducted based on a beam-based model of a multistation assembly process with compliant parts and includes diagnosis of both single and multiple fault scenarios.

Robust Design for Fixture Layout in Multistation Assembly Systems Using Sequential Space Filling Methods

Fixture layout robust design of multistation manufacturing systems aims for an optimal design that enables the dimensional variation of a product insensitive to the variations of process variables in the manufacturing process. The robust design involves a high dimension and complex global optimization problem. Recent advances in stream of variation modeling techniques enable effective formulation of the optimization problem at the system level. However, there is a challenge of computation complexity in terms of searching optimal design parameters in a high dimension, nonconvex, and discontinuous design space. This makes many available algorithms ineffective or even invalid. In this paper, an alternative sequential space filling strategy is proposed, which adopts sampling approaches to search optimal designs. To improve computation efficiency, the search space is sequentially reduced to generate a series of subspaces, and a method is designed to ensure a complete coverage of these subspaces in the original feasible space. In order to validate the proposed method, a floor pan assembly from an automotive body assembly process is modeled, and then the fixture robust design is conducted with the developed methods. To show the effectiveness of the proposed method, genetic algorithm and sequential quadratic programming are also applied in the case study for comparison.