Detecting Process Changes Research Articles

Defining Golden Batches in Biomanufacturing Processes From Internal Metabolic Activity to Detect Process Changes That May Affect Product Quality.

Cellular metabolism plays a role in the observed variability of a drug substance's Critical Quality Attributes (CQAs) made by biomanufacturing processes. Therefore, here we describe a new approach for monitoring biomanufacturing processes that measures a set of metabolic reaction rates (named Critical Metabolic Parameters (CMP) in addition to the macroscopic process conditions currently being used as Critical Process Parameters (CPP) for biomanufacturing. Constraint-based systems biology models like Flux Balance Analysis (FBA) are used to estimate metabolic reaction rates, and metabolic rates are used as inputs for multivariate Batch Evolution Models (BEM). Metabolic activity was reproducible among batches and could be monitored to detect a deliberately induced macroscopic process shift (i.e., temperature change). The CMP approach has the potential to enable "golden batches" in biomanufacturing processes to be defined from the internal metabolic activity and to aid in detecting process changes that may impact the quality of the product. Overall, the data suggested that monitoring of metabolic activity has promise for biomanufacturing process control.

Average Run Length Computations of Autoregressive and Moving Average Process using the Extended EWMA Procedure

In the past, the control chart served as a statistical tool for detecting process changes. The Exponentially Weighted Moving Average (EWMA) control chart is highly effective for detecting small changes. This research introduces the Extended Exponentially Weighted Moving Average (Extended EWMA) control chart for the Autoregressive and Moving average process with order p = 1 and q = 1 (ARMA(1,1)) The Extended EWMA control chart incorporates two smoothing parameters ( λ1 and λ2 ) derived from the EWMA control chart. A comparative analysis of the performance of the EWMA control chart. The Average Run Length (ARL) value as determined by the explicit formulas in this research, serves as a metric for evaluating the performance of the Extended EWMA control chart and the EWMA control chart. The Numerical Integral Equation (NIE) method is used to verify the accuracy of the ARL for the explicit formulas of the two control charts which has not been before discovered. The effectiveness of control charts can also be evaluated by analyzing SDRL, ARL, MRL, RMI, AEQL, and PCI values as metrics for various design parameter values. After analyzing the results of the ARL and all five performance meters, it was determined that the Extended EWMA control chart is better than the EWMA control chart at all shift sizes of process changes. Finally, the assessment of the ARMA process is being conducted to evaluate the ARL using a dataset on PM2.5 dust levels in Bangkok, Thailand during January and February of 2024.

Improved EWMA schemes for monitoring the ratio of two normal random variables

AbstractMonitoring the ratio of two normal random variables, which is usually denoted as RZ, is one of the most important directions in statistical process control (SPC). In production scenarios, when the product specification is related to the relative ratio of two components in a mixture, or when the ratio represents a property of the product, or the difference between a product's quality measurement before and after an operation (such as a chemical reaction), the product quality can be improved by applying monitoring schemes for RZ. To further enhance the sensitivity of traditional RZ schemes in detecting process changes, especially small and moderate ones, this paper proposes an improved exponentially weighted moving average (IEWMA) monitoring scheme. The run length (RL) distribution of the proposed IEWMA‐RZ scheme is obtained by using the Monte‐Carlo (MC) simulation. The superiority of the proposed monitoring scheme is shown by comparing with the existing EWMA‐RZ scheme.

A variable sampling interval one-sided CUSUM control chart for monitoring the multivariate coefficient of variation

Monitoring techniques based on the multivariate coefficients of variation (MCV) have received a great deal of attention in quality control. Numerous studies have shown that adaptively changing the charting parameters based on the past sample information can improve the performance of a chart in detecting process changes. In view of the performance benefits of adaptive strategies, two one-sided cumulative sum (CUSUM) charts are proposed to monitor the MCV and a variable sampling interval (VSI) strategy is incorporated into these charts. By using the Markov chain method, the formulas for calculating the average and standard deviation of time to signal measures of the VSI CUSUM MCV control charts are derived. Then for the known and unknown shift sizes, the optimization algorithms for obtaining the charting parameters are presented. The superiority of the proposed control charts is confirmed by comparing their performance with two existing MCV charts. Finally, an example using real investment data demonstrates the practical application of the proposed charts.

Long-term process stability in additive manufacturing

Laser powder bed fusion (L-PBF), also known as selective laser sintering or direct laser melting, is an additive manufacturing process in which part geometries are formed simultaneously with the underlying material. The microstructure, defect content, and surface quality are all synthesized conjointly with the part shape. While the geometric design freedom allowed by this process enables new complex features and parts with small (∼1 mm) features, challenges associated with process qualification can deter wider adoption. Furthermore, a lack of historical performance data for statistical process control of witness coupons, for either bulk material or for small features, makes the barrier to entry more difficult. Here, we demonstrate long-term, property-based process monitoring and variability assessment using both small-featured (1 mm) and larger, bulk-representative material witness coupons. Over a one-year period, more than 550 tensile bars and 80 Charpy impact bars were printed alongside 316 L stainless steel parts built using L-PBF and tested to detect shifts in the process over time. Miniature tensile bars with a 1 mm2 gage area were tested using a high throughput mechanical testing system. In parallel, a larger test coupon was used to monitor density, hardness, and Charpy impact toughness. This collection of measurements was used to determine detectable property shifts correlated to L-PBF process changes including powder feedstock, machine hardware, software versioning, and machine parameter settings. The benefits of using small featured, high-throughput samples are discussed based on process sensitivity and the number of repeat tests possible for each build. This study not only reveals the utility of property-based process monitoring but illustrates the sensitivity of these measurements to detect process changes and provides further evidence for property stability in modern L-PBF.

Control charts for profile monitoring of within-profile correlations using the Tweedie exponential dispersion process model

In some manufacturing processes, the quality of a product can be characterized by the functional relationship between the response variable and the explanatory variables. Profile monitoring is a tool used to examine the stability of the functional relationship over time. Many studies have pointed out that the measurements within a profile are uncorrelated. In this research, to take advantage of the ability to provide a comprehensive description of a wide variety of data, the Tweedie exponential dispersion process model is used to account for the within-profile correlation. Then, two exponentially weighted moving average control charts are developed to detect process changes during a Phase II application. In the simulation studies, we showed that our proposed control charts outperform the existing methods for monitoring linear and nonlinear profiles. Finally, two numerical examples are given to demonstrate the usefulness of our proposed control charts.

Median‐rate control chart for simultaneous monitoring of frequency and magnitude of events

AbstractControl charts are one of the most important tools of statistical process control. Many control charts have been developed for the monitoring of time between events. However, there are situations for which magnitude of events is also important and needs to be monitored along with the time between them. In this case, simultaneous monitoring of both time between and magnitude of events can be very efficient in detecting process changes. Therefore, a single control chart, referred to as the median‐rate chart, is developed in this paper for simultaneous monitoring of magnitude and frequency of events. It is supposed that the magnitude and frequency of events are independent, and follow gamma and exponential distributions, respectively. The proposed chart is based on the ratio of magnitude and time between events and the median of these variables. The results of the performance evaluation represent that the median‐rate chart mostly outperforms the existing charts in detecting both small and large shifts of the process. We use the average time to signal for comparing the power of the control charts in detecting out‐of‐control conditions. Finally, some examples are illustrated to show the applicability of the proposed chart.

Using feedback control of thermal history to improve quality consistency of parts fabricated via large-scale powder bed fusion

Process inconsistency in additive manufacturing (AM) leads to irregular quality of the final parts and obstructs its broader adoption in critical structural parts manufacturing. Especially in the manufacture of sizable components, detecting process changes in a real-time and accurate manner for potential corrective operations is crucial. This study aims to develop a feedback control system to reduce inconsistent part quality caused by heat accumulation differences. There are two significant challenges. 1) Thermal images suffer from a low signal-to-noise ratio. 2) Discrete local sintering information should be able to indicate the adjustments to the global temperature field. To tackle these challenges, a feedback model based on a multi-input neural network is proposed to evaluate the sintering status accurately by integrating the thermal history and process features. Subsequently, a layerwise feedback control strategy is proposed to process the discrete sintering status into the variation trend of part quality and ensure that the material has the desired thermal history during sintering. A controlled experiment is used to demonstrate the effectiveness of the proposed approach compared with its traditional counterparts, and the result illustrates the elimination of differences in heat accumulation by the proposed method. • The feedback control method for the thermal history of materials effectively improved the consistency of part quality. • The method realized accurate control of the sintering status by integrating the thermal history and process features. • The method was validated using a 2 × 2 m large-scale powder bed fusion equipment.

Monitoring and Diagnosis of Multistage Manufacturing Processes Using Hierarchical Bayesian Networks

In recent years, manufacturing systems have given rise to manufacturing big data due to the rapid developments in sensor and information technology and that has fueled data-driven research techniques towards addressing the issues in multistage quality control and diagnosis. In this paper, a unified framework with dual Hierarchical Bayesian Networks (HBNs) has been presented for simultaneous online process monitoring and fault diagnosis of a multistage manufacturing system. To achieve this, a novel AMDS (Absolute Mean Deviation of States) control chart has been developed for monitoring the unobserved inputs. The AMDS control chart is built on the AMDS statistic, which is calculated using the inferred states distribution generated utilizing the HBNs of the unobserved inputs. Discrete event simulation results of the two-stage process demonstrate that the methodology can successfully detect process changes and diagnose the root causes of the change. In addition, it can also identify the time at which the fault has occurred and the type (mean shift or variance change) and nature (step faults or slow drifts) of the change. The robustness of the proposed methodology is extensively tested against multiple randomly generated non-linear quadratic process models for two-stage systems.

A Statistical Method for Sequential Images–based Process Monitoring

Today, with the growth of technology, monitoring processes by the use of video and satellite sensors have been more expanded, due to their rich and valuable information. Recently, some researchers have used sequential images for defect detection because a single image is not sufficient for process monitoring. In this paper, by adding the time dimension to the image-based process monitoring problem, we detect process changes (such as the changes in the size, location, speed, color, etc.). The temporal correlation between the images and the high dimensionality of the data make this a complex problem. To address this, using the sequential images, a statistical approach with RIDGE regression and a Q control chart is proposed to monitor the process. This method can be applied to color and gray images. To validate the proposed method, it was applied to a real case study and was compared to the best methods in literature. The obtained results showed that it was more effective in finding the changes.

Manufacturing Process Monitoring With Nonparametric Change-Point Detection in Automotive Industry

Automatic sensing devices and computer systems have been widely adopted by the automotive manufacturing industry, which are capable to record machine status and process parameters nonstop. While a manufacturing process always has natural variations, it is crucial to detect significant changes to the process for quality control, as such changes may be the early signs of machine faults. This motivates our study on change-point detection methods for automotive manufacturing. We aim at developing a systematic approach for detecting process changes retrospectively in complex, nonstationary data. The proposed approach consists of nonparametric change-point detection, alarm generation based on change-point estimations, and performance evaluation against historical maintenance records. For change-point detection, three nonparametric methods are suggested—least absolute shrinkage and selection operator (LASSO), thresholded LASSO, and wild binary segmentation (WBS). Multiple decision rules are proposed to determine how to generate alarms from change-point estimations. Numerical studies are conducted to demonstrate the performance of the proposed systematic approach. The different change-point detection methods and different decision rules are evaluated and compared, with scenarios for choosing one set of change-point detection method and decision rule over another combination identified. It is shown that LASSO and thresholded-LASSO outperform WBS when the shift size is small, but WBS produces a smaller false alarm rate and handles the clustering of changes better than LASSO or thresholded LASSO. Data from an automotive manufacturing plant are used in the case study to demonstrate the proposed approach. Guidelines for implementation are also provided.

Optimal design of the variable sampling interval np control chart for monitoring service process

As customer satisfaction is a key success factor for a service system, the main purpose of this paper is to design a control chart for monitoring the service process. The variable sampling interval (VSI) control charts are useful and have important applications in service industries, which sample at a higher rate when there is evidence of a change in the process, and are thus able to detect process changes faster than the traditional fixed sampling interval (FSI) control charts. This paper evaluates and compares the performance of the VSI np control chart with the average time to signal (ATS) properties, when the process parameters are known and estimated, the results demonstrate that these performances of the VSI np chart are quite different when the number of samples used during Phase I is small. Then we suggest new dedicated chart parameters for the VSI np chart with estimated parameter, to have approximately the same in-control ATS values as in the known parameter case.

랜덤포레스트 기반 다 범주 분류기를 이용한 RTC(Real-time Contrast) 관리도

Abnormality detection and causal variables isolation are very important in the manufacturing process. However traditional multivariate statistical process control charts should assume the distribution and are challenged by high dimensional and non-linear data. To overcome these limitations, random forest based real-time contrast (RTC) control chart that transform test procedures to sequential classifications was proposed. Although RTC control chart has the advantage to isolate causal variables, monitoring statistics of the RTC control chart is the probability limited between 0.5 and 1; this could deteriorate abnormality detection ability. Features that use the sliding window can also reduce the sensitivity of detecting process changes. Therefore, we propose improved RTC control chart using random forest based multi-class classifier. This improved RTC control chart has the wider range of monitoring statistics and can detect process changes more quickly. In addition, the causal variable can be detected in the same way as the existing RTC control chart.

Dual process monitoring of metal-based additive manufacturing using tensor decomposition of thermal image streams

Additive manufacturing (AM) processes are subject to lower stability compared to their traditional counterparts. The process inconsistency leads to anomalies in the build, which hinders AM’s broader adoption to critical structural component manufacturing. Therefore, it is crucial to detect any process change/anomaly in a timely and accurate manner for potential corrective operations. Real-time thermal image streams captured from AM processes are regarded as most informative signatures of the process stability. Existing state-of-the-art studies on thermal image streams focus merely on in situ sensing, feature extraction, and their relationship with process setup parameters and material properties. The objective of this paper is to develop a statistical process control (SPC) approach to detect process changes as soon as it occurs based on predefined distribution of the monitoring statistics. There are two major challenges: 1) complex spatial interdependence exists in the thermal images and current engineering knowledge is not sufficient to describe all the variability, and 2) the thermal images suffer from a large data volume, a low signal-to-noise ratio, and an ill structure with missing data. To tackle these challenges, multilinear principal component analysis (MPCA) approach is used to extract low dimensional features and residuals. Subsequently, an online dual control charting system is proposed by leveraging multivariate T2 and Q control charts to detect changes in extracted low dimensional features and residuals, respectively. A real-world case study of thin wall fabrication using a Laser Engineered Net Shaping (LENS) process is used to illustrate the effectiveness of the proposed approach, and the accuracy of process anomaly detection is validated based on X-ray computed tomography information collected from the final build offline.

Evaluating the ANSS and ATS Values of the Multivariate EWMA Control Charts with Markov Chain Method

AbstractAverage number of samples to signal (ANSS) and average time to signal (ATS) are the most widely used criterion for comparing the efficiencies of the quality control charts. In this study the method of evaluating ANSS and ATS values of the multivariate exponentially weighted moving average (EWMA) control charts with Markov chain approach was presented when the production process is in control state or out of control state. Through numerical results, it is found that when the number oftransient state
 is less than 50, the calculated ANSS and ATS values are unstable; and š­¬Þ
s tends to be stabilized when
is greater than 100; in addition, when the properties of multivariate EWMA control chart is evaluated using Markov chain method, the number of transient state
 requires bigger values when the smoothing constatnt
A becomes smaller. Key words: Asymptotic ATS, Control Chart, Markov Chain Method 1. Introduction 1 The main purpose of statistical process control (SPC) is to improve the quality and productivity. One of the efficient methods for checking shifts or variations in the production process is control chart. Average run length (ARL), ANSS and ATS are the most widely used criterion for comparing the efficiency of the quality control charts. The ability of a control chart detecting process changes is determined by the length of time required for the chart to signal. Thus, a good control chart detects any changes quickly in the process while producing few false alarms. In traditional control chart, the run length (RL) is defined as the random number of samples required for the chart to signal and the ARL is the expected value of the RL. Therefore, the expected time to signal is simply the product of the ARL and the length of the fixed sampling interval in fixed sampling interval (FSI) chart, so the ARL can be thought of as the expected time to signal. If there are no changes in the process then the ARL should be large so that the frequency of false alarm

The Design of the ARL-Unbiased S 2 Chart When the In-Control Variance Is Estimated

The S2 chart has been known as a powerful tool to monitor the variability of the normal process. When the variance of the process is unknown, it needs to be estimated by Phase I samples. It is well known that there are serious effects of parameter estimation on the performance of the S2 chart based on known parameter assumption. If the effects of parameter estimation are not considered, it can lead to an increase in the number of false alarms and a reduction in the ability of the chart to detect process changes except for very small shifts in the variance. Based on the criterion of average run length (ARL) unbiased, a S2 control chart is developed when the in-control variance is estimated. The performance of the proposed control chart is also evaluated in terms of the ARL and standard deviation of the run length. Finally, an example is used to illustrate the proposed control chart. Copyright © 2013 John Wiley & Sons, Ltd.

Review on Measurement Techniques for Drop Size Distribution in a Stirred Vessel

A literature review on measurements of drop size distribution in liquid–liquid dispersion produced in a stirred vessel is presented in this work. The methods of measurement can be classified into in situ and external measurement. Two main groups of measurement techniques, namely, a laser system and image analysis, are reviewed. Several issues regarding the applications of the techniques and possible ways to overcome the problems are discussed. The suitability of different techniques depends on the operating conditions and properties of the drops. Laser systems provide fast in situ measurements which are useful for online monitoring and detecting process changes but unable to deliver reliable drop size and distribution values. In situ image analysis techniques could give accurate measurement of drop size, but a long time is required to analyze drops from a large number of images. However with development of automated image analysis, analysis time can be reduced. Therefore real-time monitoring and process control by image analysis techniques can be possible.

Cumulative Conformance Count Charts with VariableSample Sizes

Cumulative count of conforming (CCC) chart has been shown effective for high-yield processes with very low fraction of nonconforming items. Traditional CCC chart assumes the products from a process are inspected item by item. Recently, the CCC chart was generalized to conform to the industrial practice where items are inspected sample by sample without according to the production order. Different from the traditional chart, the generalized CCC chart monitors the cumulative count of samples until a nonconforming sample is encountered. In order to improve the power of the generalized CCC chart to detect process changes, this study applied the control scheme of variable sample size (VSS) to the generalized CCC chart. The average time to signal (ATS) process change of the proposed chart was derived by Markov chain approach and taken as performance measure to evaluate its statistical efficiency. Performance comparisons between the proposed chart and generalized CCC chart are made to evaluate its usefulness.

A new adaptive control chart for monitoring process mean and variability

Traditionally, an \(\bar{X}\) chart is used to control the process mean, and an R chart is used to control the variance. However, these charts are not sensitive to the small shifts in the processes. The adaptive charts might be considered if the aim is to detect process changes quickly. In this paper, we propose a new adaptive single control chart which integrates the exponentially weighted moving average procedure with the generalized likelihood ratio test statistics for jointly monitoring both the process mean and variability. This new chart is effective in detecting the disturbances that shift the process mean, increase or decrease the process variance, or lead to a combination of both effects.

A New Chart for Detecting the Process Mean and Variability

Since the earliest days of statistical process control, it has been recognized that it is essential to monitor both mean and variability and to trigger an alarm if either characteristic shows sign of a special cause. It is, however, possible to monitor both mean and variance in a single chart responsive to shifts in either. Traditionally, this has been done by using separate and S or R charts. The purpose of this paper is to propose a new single chart with the Shiryaev–Roberts (SR) procedure to monitor both mean and variance. It is shown that, in addition to the simplicity of a single chart rather than two, the new chart has some advantages over many other charts in detecting process changes.