Quality-Related Fault Detection Based on Improved Independent Component Regression for Non-Gaussian Processes

Aljunaid, Majed; Shi, Hongbo; Yang, Tao

doi:10.1109/access.2019.2948756

Cited by 10 publications

(17 citation statements)

References 33 publications

(52 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this test, IDV (1) and IDV (17) were selected to analyze the monitoring features of the proposed AAECPLS. IDV (1) was a step fault caused by a sudden change in A and C feed ratios.…”

Section: Test Resultsmentioning

confidence: 99%

Adversarial autoencoder concurrent projection to latent structure and its application

2023

View full text Add to dashboard Cite

Multivariate monitoring plays an important role in process monitoring. Among the multivariate monitoring methods, the projection to latent structure method (PLS) has been most widely used in the fields of quality control and fault diagnosis. To improve the monitoring capability of traditional PLS methods in nonlinear and non‐Gaussian multivariate systems, this paper proposes an innovative multivariate monitoring strategy, which combines adversarial autoencoder (AAE) and the concurrent projection to latent structure method (CPLS). In the proposed strategy, the original data are mapped to the high‐dimensional space using the AAE method with the Gaussian prior distribution to realize data transformation. The mapped data are linearly divisible and approach the Gaussian distribution. Then, the projection with orthogonality is realized using the CPLS method. In addition, the reconstruction error and distribution properties are used as evaluation indexes of the high‐dimensional mapping performance. The corresponding monitoring strategy is established using the traditional statistical method based on the Gaussian distribution. Finally, the simulations are performed on the Tennessee‐Eastman process platform, and the results show that the proposed method could efficiently extract the principal components, especially in quality‐relevant fault monitoring.

show abstract

“…In this test, IDV (1) and IDV (17) were selected to analyze the monitoring features of the proposed AAECPLS. IDV (1) was a step fault caused by a sudden change in A and C feed ratios.…”

Section: Test Resultsmentioning

confidence: 99%

Adversarial autoencoder concurrent projection to latent structure and its application

2023

View full text Add to dashboard Cite

show abstract

“…Fig. 8 shows the theoretical failure rate line based on (5) and the example of artificial data randomly generated from the theoretical failure model (5). The failure rate increases as x increases, but the upper change point is unclear.…”

Section: Experiments With Dataset Generated From Non-linear Modesmentioning

confidence: 99%

“…In a real-world production line, QOSPC has two major challenges: extracting process variables that affect product quality and determining QCLs for each variable. To address the first challenge, multivariate statistical models, such as partial least squares regression, are often used (e.g., [5]), but it is difficult to construct a robust model with a small amount of data. For the second challenge, as described in Section II, an unclear boundary makes it difficult to determine QCLs.…”

Section: Introductionmentioning

confidence: 99%

Quality-Oriented Statistical Process Control Utilizing Bayesian Modeling

Date

Tanaka

2021

IEEE Trans. Semicond. Manufact.

View full text Add to dashboard Cite

Quality control is an important issue in semiconductor manufacturing. Statistical process control (SPC) is known as a powerful method for accomplishing process stability and reducing variability. In this paper, we adopt the quality-oriented statistical process control (QOSPC) method. In QOSPC, product quality test data, such as electrical performance and product reliability, are incorporated into the process control procedure. QOSPC has two major challenges: extracting process variables that affect product quality, and determining quality control limits (QCLs) for each variable. In this work, we fully exploit a Bayesian approach to resolve both of these challenges simultaneously. We introduced a linear bathtub model that contains parameters corresponding to QCLs as obvious change points and fit the model to the observed data by Bayesian inference (BI). In our experiments with artificial datasets, we demonstrated that the values of QCLs and their confidence, by which we can judge whether the measured process variable is related to product quality, are estimated successfully by BI. We verified the robustness of our method by testing it repeatedly. The proposed method reduced the human labor cost for extracting quality-related process variables and determining QCLs by 93%.

show abstract

“…In contrast, it may be observed that the detection rate of ICA-KD is very good and slightly better in comparison with conventional fault indicators of ICA strategy. The ICA-KD strategy shows good advantage over other FD strategies for IDV (10), IDV (11), IDV (16), IDV (19), IDV (20) and IDV(21) dom variation fault in C feed temperature and IDV (16) which is defined as unknown in the process flowsheet.…”

Section: Tennessee Eastman Processmentioning

confidence: 99%

“…The multi-variate method based on Independent Component Analysis (ICA) has received good attention vowing to its ability of extracting information from non-gaussian and non-linear industrial data. ICA disintegrates observed data into latent variables or independent components that are independent and non-gaussian in nature [11], [12]. The ICA method de-correlates the data by reducing higher order statistical dependencies and focuses on making distribution of the disintegrated data as independent as possible.…”

Section: Introductionmentioning

confidence: 99%

Improved Process Monitoring Strategy Using Kantorovich Distance-Independent Component Analysis: An Application to Tennessee Eastman Process

Kini

Madakyaru

2020

IEEE Access

View full text Add to dashboard Cite

Vowing to the increasing complexity in industrial processes, the need for safety is of highest priority and this has led to development of efficient fault detection (FD) methods. Also, with rapid development of data acquisition systems, process history based methods have gained importance as their dependency is on large volume of sensor data extracted from the process. The industrial data exhibits some degree of non-gaussianity for which Independent Component Analysis (ICA) technique has usually been applied in practice. Recently, a new fault indicator based on Kantorovich Distance (KD) has been proposed which computes distance between two distributions and uses the distance as an indicator of fault. The KD metric has found to provide good monitoring results for data in presence of noise and offers enhanced detection of small magnitude faults. Considering the benefits offered by KD metric, the objective of this work is to amalgamate KD metric with ICA modeling framework to have a fault detection strategy that can improve process monitoring in noisy environment. The proposed ICA-KD FD strategy is illustrated on four processes that includes Modified Continuous Stirred Tank Heater (CSTH), Tennessee Eastman (TE) process and Experimental Distillation Column Process. The simulation results indicate that the proposed FD strategy exhibits improved performance over conventional strategies while monitoring different sensor faults in noisy environment.INDEX TERMS Process monitoring, fault detection, independent component analysis, Kantorovich distance, small magnitude faults, Tennessee Eastman process, experimental distillation column process, modified continuous stirred tank heater process.

show abstract

Quality-Related Fault Detection Based on Improved Independent Component Regression for Non-Gaussian Processes

Cited by 10 publications

References 33 publications

Adversarial autoencoder concurrent projection to latent structure and its application

Adversarial autoencoder concurrent projection to latent structure and its application

Quality-Oriented Statistical Process Control Utilizing Bayesian Modeling

Improved Process Monitoring Strategy Using Kantorovich Distance-Independent Component Analysis: An Application to Tennessee Eastman Process

Contact Info

Product

Resources

About