Multistage Manufacturing Processes Research Articles

Multiresponse optimization of a multistage manufacturing process using a patient rule induction method

AbstractMost manufacturing industries produce products through a series of sequential stages, known as a multistage process. In a multistage process, each stage affects the stage that follows, and the process often has multiple response variables. In this paper, we suggest a new procedure for optimizing a multistage process with multiple response variables. Our method searches for an optimal setting of input variables directly from operational data according to a patient rule induction method (PRIM) to maximize a desirability function, to which multiple response variables are converted. The proposed method is explained by a step‐by‐step procedure using a steel manufacturing process as an example. The results of the steel manufacturing process optimization show that the proposed method finds the optimal settings of input variables and outperforms the other PRIM‐based methods.

Quality inspection planning within a multistage manufacturing process based on the added value criterion

This paper proposes an approach to the planning and optimization of quality inspections within a multistage manufacturing process based on quality costs and the value added to the production process by inspections. Inspection errors and the resulting costs of repair and scrapping are taken into account. In addition, the capability of the manufacturing process and of the inspection system is included. A mathematical model for calculating quality costs was developed with a focus on its practical application.

Error chain analysis—an effective method for tighter manufacturing process control

One of aims of manufacturing quality control is to ensure that products are made free from defects according to specifications without unnecessarily increasing time and cost of production. Over-control of a process can be as detrimental to a manufacturer as under-control. It is common in industry that operators use their personal knowhow and intuition to decide where to implement process verification, and where to tighten it when processes are not meeting specifications. This is partially because there is little scientific guidance that can assist operators in making a decision on levels of quality control of a process at varying stages. To remedy this, a new method for manufacturing quality control, namely Error Chain Analysis (ECA), is introduced and its application is illustrated in this article. ECA is capable of statistically analysing the quality of a multi-stage manufacturing process based on existing control measures, and it enables to indicate where added or tighter control may need to be effectively implemented. For testing its applicability, ECA was built into a user-friendly tool that was subsequently used to analyse data gathered from a large manufacturing company in the UK.

Fault detection and recognition of multivariate process based on feature learning of one-dimensional convolutional neural network and stacked denoised autoencoder

Multivariate process pattern recognition (MPPR) is essential towards continuous quality control task. A challenging problem is to extract effective features from complex process signals with high-dimensional and nonlinear characteristics. This affects effectiveness of various classifiers in process fault detection and diagnosis significantly. In this paper, we propose a hybrid deep learning model (i.e. 1-DCNN + SDAE) that integrates one-dimensional convolutional neural network (1-DCNN) and stacked denoising auto-encoders (SDAE) to extract high level features from complex process signals. In comparison with two-dimensional images, one-dimensional process signals allow not only to extract spatial features, but also reduce calculation cost. 1-DCNN is capable of extracting representative features from one-dimensional process signals and then improves MPPR performance of classifiers significantly. SDAE is embedded after fully connected layer of 1-DCNN for further dimension reduction and feature extraction. 1-DCNN + SDAE preserves advantages of 1-DCNN and SDAE for feature learning from high-dimensional data. This makes it be flexible for those process fault detection and diagnosis tasks. The effectiveness of 1-DCNN + SDAE is validated on a complex numerical process, two process benchmarks i.e. Tennessee Eastman process (TEP) and Fed-batch fermentation penicillin process (FBFP), and a real-life manufacturing case of industrial conveyor belt. The experimental results illustrate effectiveness of the proposed method for feature learning and fault diagnosis on multivariate manufacturing processes. The comparison between 1-DCNN + SDAE and other typical DNNs on these processes, indicates the effectiveness of the proposed method for process fault detection and diagnosis. This study will provide the guidance for development of hybrid deep learning-based multivariate control models.

Optimizing mean and variance of multiresponse in a multistage manufacturing process using operational data

A multistage process consists of sequential stages where each stage is affected by its preceding stage, and it in turn affects the stage that follows. The process described in this article also has several input and response variables whose relationships are complicated. These characteristics make it difficult to optimize all responses in the multistage process. We modify a data mining method called the patient rule induction method and combine it with desirability function methods to optimize the mean and variance of multiresponse in the multistage process. The proposed method is explained by a step-by-step procedure using a steel manufacturing process example.

Variation Source Identification in Manufacturing Processes Using Bayesian Approach With Sparse Variance Components Prior

We present a Bayesian linear random effects’ model for variation source identification in multistage manufacturing processes with a prior for sparse variance components. A modified horseshoe+ (HS+) prior is used to tackle high-dimensional problems with low sample size and sparse variation sources. Furthermore, we introduce the informed HS+ (IHS+) prior that incorporates the likelihood information of possible variation sources. To estimate the variations from the IHS+ prior, a specially designed Gibbs sampler is established. Through a series of numerical experiments and case study, we showed that the proposed IHS+ outperforms the existing prior distributions when variation sources are sparse. Note to Practitioners —Economic globalization brings intense competition among manufacturing enterprises. The key to success in this competitive climate is a prompt response to rapidly changing market demands with high-quality products. Variation reduction is essential for improving process efficiency and product quality. Existing variation source identification approaches generally assume that the amount of measurement data is larger than that of the possible variation sources. We propose a method to identify the process variation sources with limited measurement capability. Moreover, we provide a tool to apply the domain knowledge on the possible variations in the manufacturing processes. The proposed method is flexible and can be adjusted according to the confidence level of the practitioners’ empirical knowledge. The experiments have shown that our method possesses advantageous features in detecting variation sources and estimating the size of variations.

Casting Manufacturing of Cylindrical Preforms Made of Low Alloy Steels

The manufacturing of complex drivetrain components made of steel requires multi-stage manufacturing processes, resulting in long processing times and high costs. Forged parts get their final geometry after multi-stage forming operations. One approach to shorten such process chains is close contour casting of steel preforms and a single subsequent forming step to achieve high strength properties similar to those of multi-stage forged parts. Since the application of precision forging using cast steel preforms has been studied insufficiently up to now, the manufacturing of cylindrical preforms made of the low-alloy heat-treatable steel G42CrMo4 by sand casting was investigated. Using casting simulations, suitable casting parameters for the finished casting systems were identified and nearly pore-free preforms were achieved. Depending on the casting system, metallographic investigations revealed different fractions of bainite in the microstructure of the casted preforms. Subsequently, the cylinders were deformed in upsetting experiments by which the remaining porosity could be eliminated. The numerical casting simulation was combined with a forming simulation to model the reduction in porosity and to allow the prediction of the densification behavior in further research employing more complex geometries.

Intelligent data-driven monitoring of high dimensional multistage manufacturing processes

Recent advances in cyber-physical systems and the Internet of things (IoT) have enabled the possible development of smart production systems. However, the complexity of such a system has posed significant challenges for traditional quality engineering methods, especially in monitoring and diagnosis of system performance. The traditional practices for monitoring or controlling multistage systems either treat each stage as an individual entity or model all stages as a whole. The formal approach mainly focuses on the most critical stages while ignores information from the other stages. In contrast, the latter approach attempts to build one model to account for all stages. In a complex production system, this latter approach is impractical, if not impossible. This research provides a control strategy by proposing an intelligent process monitoring system for high dimensional multistage processes using predictive models built from historical data. A repository dataset is used to demonstrate the implementation of the proposed framework.

Intelligent data-driven monitoring of high dimensional multistage manufacturing processes

Intelligent data-driven monitoring of high dimensional multistage manufacturing processes

Integrating Multivariate Statistical Analysis Into Six Sigma DMAIC Projects: A Case Study on AISI 52100 Hardened Steel Turning

DMAIC (define, measure, analyze, improve and control) is one of the most utilized methods for guiding practitioners in the decision-making process of quality improvement projects. Industrial processes commonly deal with multiple critical-to-quality (CTQ) characteristics. When these characteristics are correlated, multivariate statistical techniques should be applied. This paper aims to propose a domain-specific Six Sigma method, the MDMAIC (multivariate DMAIC). The new stepwise procedure helps practitioners not only to reduce problem dimension but also to take account of the correlation structure among CTQs during the decision-making process. Principal component analysis has been applied for assessing the measurement system, analyzing process stability and capability, as well as modeling and optimizing multivariate manufacturing processes. A hardened steel turning case has been presented for proposal validation. The result analysis has shown that the MDMAIC was very successful in leading the practitioner during the steps and phases of the quality improvement project. The multivariate capability index of the enhanced process emphasized the substantial economic improvement.

Design and optimization of multi-stage manufacturing process of stirling engine crankshaft

Crankshafts are among the most important parts in internal combustion engines, of which stirling engine is a useful example. Manufacturing process of a crankshaft, is considered as a three-step forging process using preform, due to the complexity in geometry. The most challenging step of the multistage forging process is to avoid stress concentration and to create uniformity of strain by controlling metal flow. In the present study, the final part was achieved under three manufacturing processes namely: upsetting, hot and cold forging. The models used in each manufacturing process are designed by CATIA software. A finite element simulation on the basis of Cockcraft–Latham damage criterion was developed in DEFORM software. Using experiment design by Taguchi method, The optimization of manufacturing processes were carried out by MINITAB software in two steps, in which the optimization objectives are considered as force, damage and strain uniformity, and; input variables are taken as part-mold friction, pressing velocity and process temperature. In order to find the most effective parameter of each manufacturing process, analysis of variance was conducted on the results, in which, the most effective parameters in the upsetting, hot and cold forging processes were temperature, friction and temperature, respectively.

Modeling multivariate dependencies for manufacturing single-mode I/O structures of integrated MMI-based splitters in glass sheets

For the realization of integrated optical interconnect systems, optical waveguide components are needed. A promising approach to manufacturing graded-index waveguide components is their embedding in thin glass sheets by ion-exchange processes. Conventionally, the modeling of integrated waveguide components manufactured by thermal diffusion processes involves the numerical solution of differential equations. In this work, a semi-analytical approach for the modeling of the concentration profile is extended by the Gauss–Newton optimization algorithm. This is done with the specific purpose of accurately modeling single-mode input and output waveguide structures of multimode-interference-based (MMI-based) optical splitters. The constraints imposed by the multivariate manufacturing process for the realization of these MMI-based components are elaborated. Corresponding correlations between these constraints and the according modeling of the devices are established.

Controlling autocorrelated data in multistage manufacturing processes with an application to textile industry

AbstractIn recent years, much attention has been given to monitoring multistage processes in order to effectively improve the product reliability. To this end, the output of the process is investigated under special circumstances, and the values corresponding to reliability‐related quality characteristic are measured. However, analyzing reliability data is quite complicated because of their unique features such as being censored and obeying nonnormal distributions. A more sophisticated picture arises when the observations of the process are autocorrelated in some cases, which makes the application of previous control procedures futile. In this paper, the accelerated failure time (AFT) regression models have been modified in order to account for autocorrelated data. Then, a cumulative sum (CUSUM) control chart and an exponentially weighted moving average (EWMA) control chart based on conditional expected values have been proposed to monitor the quality variable with Weibull distribution while taking the effective covariates into consideration. Extensive simulation studies reveal that the CUSUM control chart outperforms its counterpart in detecting out‐of‐control conditions. Finally, a real case study in a textile industry has been provided to investigate the application of the CUSUM control scheme.

Data-driven modeling and simulation of complex multistation manufacturing process for dimensional variation analysis

Dimensional variation analysis in multistation manufacturing processes (MMPs) is a challenging research topic with great practical significance. Researchers have been focused on constructing various mathematical models to identify the correlations among the huge amounts of collected production data. However, current models have achieved insufficient insights into the variation correlation laws due to the complexity of the data’s mutual relations. In this study, a data-driven modeling method is developed for deep data-mining and dimensional variation analysis. The proposed initial mathematical expression originates from practical engineering knowledge. Through a mathematical treatment, the mathematical expression is transformed into a first-order AR(1) model format, which contains multiple dimensional variations’ interstation and temporal correlating information. To obtain this information, the estimation of the proposed model is discussed in detail. A simulation case involving two key product characteristics of a grinding process is used to demonstrate the effectiveness and accuracy of the proposed method for dimensional variation analysis in MMPs.

Detecting the process changes for multivariate nonlinear profile data

AbstractIn profile monitoring for a multivariate manufacturing process, the functional relationship of the multivariate profiles rarely occurs in linear form, and the real data usually do not follow a multivariate normal distribution. Thus, in this paper, the functional relationship of multivariate nonlinear profile data is described via a nonparametric regression model. We first fit the multivariate nonlinear profile data and obtain the reference profiles through support vector regression (SVR) model. The differences between the observed multivariate nonlinear profiles and the reference profiles are used to calculate the vector of metrics. Then, a nonparametric revised spatial rank exponential weighted moving average (RSREWMA) control chart is proposed in the phase II monitoring.Moreover, a simulation study is conducted to evaluate the detecting performance of our proposed nonparametric RSREWMA control chart under various process shifts using out‐of‐control average run length (ARL1). The simulation results indicate that the SREWMA control chart coupled with the metric of mean absolute deviation (MAD) can be used to monitor the multivariate nonlinear profile data when a common fixed design (CFD) is not applicable in the phase II study. Finally, a realistic multivariate nonlinear profile example is used to demonstrate the usefulness of our proposed RSREWMA control chart and its monitoring schemes.

Estimation of Nonstationary Process Variance in Multistage Manufacturing Processes Using a Model-Based Observer

In this paper, we propose a recursive algorithm to estimate the process variance in multistage manufacturing or assembly processes. We use a replicated model that includes the process variance to be estimated as a time-varying state that changes slowly. For this model, we develop an estimation strategy including tuning parameters that play a direct role in the tradeoff between the estimation accuracy and the adaptation to changes. We also develop a statistical confidence interval for the estimations which enhances the decision of whether the process variances have changed. Unlike other batch methods in the literature, our proposal is computed recursively, and it allows us to tune the tradeoff between the convergence speed and the accuracy without modifying the sample size, which only contains the data of the last manufactured piece. Note to Practitioners —Variation reduction appears as the key to succeed in improving product quality, which is a major asset in manufacturing. To achieve variation reduction, it is necessary to quickly identify the root causes, which are responsible for these variations. The state of the art in root cause analysis is based on a stationary point of view, where the estimations regarding the mean and variance of the variation sources are based on the sample provided by a batch of multiple manufactured pieces. This paper focuses on the development of an online estimation strategy, which recursively updates the estimation of the variance of the variation sources with the data provided by the last manufactured piece in order to enhance a rapid decision of whether any fault has occurred in the process at the same time of reducing the computational cost and the data storage needs. We present a design strategy of the estimation algorithm, which sets the tradeoff between the accuracy of the estimations and their capability of tracking changes through two well-known physically meaningful engineering parameters: the cumulative squared error under abrupt changes and the variance of the estimations. Two detailed cases of study illustrate the effectiveness of this method. The strategy presented in this paper represents a novel approach for dealing with quality-control online algorithms and it can be used for the quality improvement of multistage manufacturing processes.

Autonomous supervision and optimization of product quality in a multi-stage manufacturing process based on self-adaptive prediction models

In modern manufacturing facilities, there are basically two essential phases for assuring high production quality with low (or even zero) defects and waste in order to save costs for companies. The first phase concerns the early recognition of potentially arising problems in product quality, the second phase concerns proper reactions upon the recognition of such problems. In this paper, we address a holistic approach for handling both issues consecutively within a predictive maintenance framework at an on-line production system. Thereby, we address multi-stage functionality based on (i) data-driven forecast models for (measure-able) product quality criteria (QCs) at a latter stage, which are established and executed through process values (and their time series trends) recorded at an early stage of production (describing its progress), and (ii) process optimization cycles whose outputs are suggestions for proper reactions at an earlier stage in the case of forecasted downtrends or exceeds of allowed boundaries in product quality. The data-driven forecast models are established through a high-dimensional batch time-series modeling problem. In this, we employ a non-linear version of PLSR (partial least squares regression) by coupling PLS with generalized Takagi–Sugeno fuzzy systems (termed as PLS-fuzzy). The models are able to self-adapt over time based on recursive parameters adaptation and rule evolution functionalities. Two concepts for increased flexibility during model updates are proposed, (i) a dynamic outweighing strategy of older samples with an adaptive update of the forgetting factor (steering forgetting intensity) and (ii) an incremental update of the latent variable space spanned by the directions (loading vectors) achieved through PLS; the whole model update approach is termed as SAFM-IF (self-adaptive forecast models with increased flexibility). Process optimization is achieved through multi-objective optimization using evolutionary techniques, where the (trained and updated) forecast models serve as surrogate models to guide the optimization process to Pareto fronts (containing solution candidates) with high quality. A new influence analysis between process values and QCs is suggested based on the PLS-fuzzy forecast models in order to reduce the dimensionality of the optimization space and thus to guarantee high(er) quality of solutions within a reasonable amount of time (→ better usage in on-line mode). The methodologies have been comprehensively evaluated on real on-line process data from a (micro-fluidic) chip production system, where the early stage comprises the injection molding process and the latter stage the bonding process. The results show remarkable performance in terms of low prediction errors of the PLS-fuzzy forecast models (showing mostly lower errors than achieved by other model architectures) as well as in terms of Pareto fronts with individuals (solutions) whose fitness was close to the optimal values of three most important target QCs (being used for supervision): flatness, void events and RMSEs of the chips. Suggestions could thus be provided to experts/operators how to best change process values and associated machining parameters at the injection molding process in order to achieve significantly higher product quality for the final chips at the end of the bonding process.

A Bayesian framework to estimate part quality and associated uncertainties in multistage manufacturing

Manufacturing is usually performed as a sequence of operations such as forming, machining, inspection, and assembly. A new challenge in manufacturing is to move towards Industry 4.0 (the fourth Industrial revolution) concerning the full integration of machines and production systems with machine learning methods to enable for intelligent multistage manufacturing. This paper discusses Multistage Manufacturing Processes (MMPs) and develops a probabilistic model based on Bayesian linear regression to estimate the results of final inspection associated with comparative coordinate measurement given in-process measured coordinates. The results of two case studies for flatness tolerance evaluation demonstrate the effectiveness of the probabilistic model which aims at being part of a larger metrology informatics system to be developed for predictive analytics and agent-based advanced control in multistage manufacturing. This solution relying on accurate models can minimise post-process inspection in mass production with independent measurements.

An Intelligent Metrology Informatics System based on Neural Networks for Multistage Manufacturing Processes

The ability to gather manufacturing data from various workstations has been explored for several decades and the advances in sensory and data acquisition techniques have led to the increasing availability of high-dimensional data. This paper presents an intelligent metrology informatics system to extract useful information from Multistage Manufacturing Process (MMP) data and predict part quality characteristics such as true position and circularity using neural networks. The input data include the tempering temperature, material conditions, force and vibration while the output data include comparative coordinate measurements. The effectiveness of the proposed method is demonstrated using experimental data from a MMP.

Optimizing Mean and Variance of Multiresponse in a Multistage Manufacturing Process Using a Patient Rule Induction Method

Most manufacturing industries produce products through a series of sequential stages, known as a multistage process. In a multistage process, each stage is affected by its preceding stage, at the same time, it affects its following stage. Also, each stage often includes several response variables to be optimized. In this paper, we attempt to optimize the several response variables of the multistage process simultaneously considering the relationships among the stages. For this purpose, we use a particular data mining method, called a patient rule induction method. Because the relationships among the stages are often complicated, using a data mining method is a good approach for analyzing the relationships. According to the procedure of the patient rule induction method, the proposed method searches for an optimal setting of input variables directly from operational data at which mean and variance of the several response variables of the multistage process are optimized. The proposed method is explained by a step-by-step procedure using a steel manufacturing process example.