Contribution Plots Research Articles

Processes monitoring using adaptive confidence limit based on T-S fuzzy model and Luenberger observer

In hazard-sensitive processes, the monitoring upsets and malfunctions correctly is an important challenge to operation safely and enhance the performance. Conventional process monitoring frequently assumes that process data follow only one Gaussian distribution, which generates a constant confidence limit and hence produces a high number of false alarms. However, in fact, industrial processes usually include various operating modes. To ovoid this drawback, the suggested approach employs an adaptive confidence limit (ACL) when a substantial number of false alarms are created. The fundamental concept underlying this study is to extract internally several local linear sub-modes of the monitored variables. In typical operating circumstances, the Gaussian mixture model (GMM) is utilized to extract several local linear sub-modes, followed by fuzzy linearization using the Takagi-Sugeno model, thereafter a bank of Luenberger observers to construct the residual spaces. An abnormal event is detected when the squared prediction error (SPE) is too great or exceeds the adaptive threshold designed to prevent the false alarms. Furthermore, an enhanced contribution plots is effectively used to identify the defective variable.

Fault root cause analysis using degree of change and mean variable threshold limit in non-linear dynamic distillation column

The chemical industrial processes are both dynamic and complex. The presence of instability and danger in the production process poses significant challenges to safety management. Early and accurate fault diagnosis is essential to ensure a sustainable and stable chemical process performance. Most studies in literature do not provide detailed information about diagnosing the root cause of faults in distillation columns using contribution plots. Using principal component analysis (PCA) in conventional contribution plots leads to a large number of contributing variables, making it difficult to identify the root cause of a specific fault. In this study, the authors suggest employing the Degree of Change of Contribution Values (DCC) and Mean Variable Threshold Limit (LMVT) techniques alongside the Wavelet Transform-Principal Component Analysis (WT-PCA) method to decrease the quantity of fault identification variables in a non-linear system (distillation column). The Aspen Plus software is utilized to perform a dynamic simulation of a distillation column, generating data for both normal and faulty operations. Two distinct types of faults are considered, which include loss of feed flow and feed temperature changes. This study aims to analyze the results obtained from the Hotelling T2 and SPE contribution plots by calculating the average difference between the normal and faulty conditions using the Degree of Change in Contribution Values. This procedure is performed to determine the extent of deviation from the standard state. Subsequently, the variables that have surpassed the threshold limit for means variables (LMVT) are designated as the variables that serve as the primary underlying cause of the fault. The technique was employed in a case study involving a nonlinear distillation column, and the results indicate that only one "dominant" variable intersects the LMVT by Hotelling T2, which represents the true underlying cause of the corresponding fault.

Wasserstein local slow feature analysis and its application to process monitoring

Abstract Complex industrial processes are commonly characterized by dynamics, which results from the fact that there exists compensation of the closed-loop control and complex reflux during process operations. In this work, we propose a new method termed Wasserstein local slow feature analysis approach (WLSFA) used for monitoring the dynamic process, which can learn slowly varying information to capture the trend of process variations and characterize dynamics. Specifically, Wasserstein graph embedding based on optimal transport is proposed to preserve the local geometrical structure of raw data so that the reduced-dimensional representations can more faithfully reveal the underlying information of data, enhancing monitoring performance. Moreover, the l_2 -norm orthogonal constraint is incorporated into the objective function to improve generalization ability and alleviate the false alarm rate. On the basis of WLSFA model, a contribution plot in the sense of reconstruction is developed to locate the fault variables, which is beneficial to regulate abnormal conditions towards normal levels. Finally, the efficiency of the proposed approach is illustrated by a benchmark process and the application to a real-world fractionation process.

Partial Cross Mapping Based on Sparse Variable Selection for Direct Fault Root Cause Diagnosis for Industrial Processes.

Root cause diagnosis of process industry is of significance to ensure safe production and improve production efficiency. Conventional contribution plot methods have challenges in root cause diagnosis due to the smearing effect. Other traditional root cause diagnosis methods, such as Granger causality (GC) and transfer entropy, have unsatisfactory performance in root cause diagnosis for complex industrial processes due to the existence of indirect causality. In this work, a regularization and partial cross mapping (PCM)-based root cause diagnosis framework is proposed for efficient direct causality inference and fault propagation path tracing. First, generalized Lasso-based variable selection is performed. The Hotelling T2 statistic is formulated and the Lasso-based fault reconstruction is applied to select candidate root cause variables. Second, the root cause is diagnosed through the PCM and the propagation path is drawn out according to the diagnosis result. The proposed framework is studied in four cases to verify its rationality and effectiveness, including a numerical example, the Tennessee Eastman benchmark process, the wastewater treatment process (WWTP), and the decarburization process of high-speed wire rod spring steel.

Causality-embedded reconstruction network for high-resolution fault identification in chemical process

Fault identification is essential for analyzing the root causes and propagation of faults. Traditional identification based on contribution plots often suffer from the smearing effect, a phenomenon where the residuals of fault variables smearing over other non-fault process variables due to the complex interconnections within the processes, potentially leading to ambiguous or inaccurate diagnosis. To address this, we propose the Causality-Embedded Reconstruction Network (CERN) based high-resolution fault identification for chemical process. CERN is an advanced graph network model that embeds causality, featuring a dual-decoding scheme comprising a causal decoder and an attribute decoder. This design ensures that feature reconstruction is guided by variables with established causal relationships, effectively isolating the feature representations of normal variables from the influence of fault variables that lack a causal connection. By leveraging this causality-embedded reconstruction, the smearing effect can be alleviated. Experiments conducted on a Continuous Stirred-Tank Reactor system and a real-world Continuous Catalytic Reforming process to demonstrate that the proposed CERN-based fault identification method not only improves the accuracy in identifying fault variables but also offers explanatory insights and traceability for fault diagnosis, showing prospects for industrial process safety.

KPI-related monitoring approach for powertrain system in hybrid electric vehicles

This paper introduces a novel monitoring method related to key-performance-indicators (KPIs), specifically tailored for the hybrid electric vehicle (HEV) powertrain system. The proposed method establishes a new KPI that better reflects the performance of the HEV powertrain system. Through the application of partial least squares and contribution plot method, it excels in minimizing data scale and precisely monitoring HEV powertrain faults. Diverging from current methodologies, this method demands minimal prior knowledge and solely relies on previously observed HEV data. The simulation results demonstrate the method's capability to detect engine and motor overheating faults while accurately diagnosing their root causes. In summary, this innovative method marks a significant departure from existing approaches, providing a robust and powerful tool for state monitoring in the HEV powertrain system.

Multiscale Kernel Entropy Component Analysis With Application to Complex Industrial Process Monitoring

Modern industrial processes are characterized by numerous measurement points and wide operating ranges, resulting in extremely complex correlations among variables. Therefore, an effective monitoring system should balance diverse process characteristics such as nonlinearity, non-Gaussianity, and multiscale simultaneously. Moreover, it should have the ability to detect and diagnose faults in the incipient stage, thus avoiding accident escalation. With these goals in mind, this paper proposes an integrated monitoring solution based on multiscale kernel entropy component analysis (MSKECA). Specifically, process variables are first decomposed into approximations and details at different scales in real-time using the moving window-based wavelet, and contributions from each scale are collected in separate matrices. Then, KECA-based local-scale models are built to sift out important detail scales for reconstruction along with the approximate scale. Lastly, a KECA-based global-scale model is developed to monitor the reconstructed data. To improve the fault detection performance, a novel monitoring index based on the angle metric called angle variance index (AVI) is designed. In addition, to achieve effective diagnosis, MSKECA-based contribution plots are constructed, which depict the contributions of variables to faults at each scale, thus comprehensively revealing the root causes. Finally, the effectiveness and superiority of the proposed solution are validated by comparisons with other advanced counterparts in two industrial scenarios. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Note to Practitioners</i> —This paper proposes an integrated monitoring solution for complex industrial processes, i.e., MSKECA-based fault detection and diagnosis. The solution takes into account the diversity of process characteristics, the effectiveness of incipient fault detection, and the richness of diagnostic information. Specifically, 1) MSKECA can simultaneously handle the nonlinear, non-Gaussian, and multiscale characteristics that are prevalent in real process data. It enables on-line detection of significant events occurring at different scales and extraction of fault-sensitive features for monitoring; 2) based on the angular structure of KECA, the AVI statistic is designed, which exhibits low autocorrelation and is sensitive to faults. Leveraging the statistic, MSKECA allows reliable and prompt responses to faults; 3) MSKECA-based contribution plots not only convey diverse diagnostic information including fault variable, type, and grade but also are not susceptible to the smearing effect, which is helpful for practitioners to achieve fault repair. The solution has proven useful for a real hot rolling process. It can also be extended to other industrial processes.

Single and multiple fault diagnosis based on PCA with relative variation of contribution strategy

Data-driven methods have been recognized as an efficient tool for multivariate statistical process control. Contribution plots are also well known as a popular tool in principal component analysis, which is used for isolating sensor faults without the need for any prior information. However, studies carried out in the literature have unified contribution plots in three general approaches. Furthermore, they demonstrated that correct diagnosis based on contribution plots is not guaranteed for either single or multiple sensor faults. Therefore, to deal with this issue, the present paper highlights a new contribution formula called relative variation of contribution. Simulation results show that the proposed method of contribution can successfully perform the fault isolation task, in comparison with partial decomposition contribution and its relative version (rPDC) based on their fault isolation rate.

Fault detection and diagnosis in a non‐Gaussian process with modified kernel independent component regression

AbstractQuality‐related fault detection and diagnosis are crucial in the data‐driven process monitoring field. Most existing methods are based on principal component analysis (PCA) or partial least squares (PLS), which will miss high‐order statistical information when the industrial process does not satisfy a Gaussian distribution. Meanwhile, the traditional contribution plot is difficult to directly apply to nonlinear processes in some cases due to its limitation of convergence. As such, a modified kernel independent component regression (MKICR) model, which considers high‐order statistical information, is proposed for quality‐related fault detection and faulty variable identification. First, the relationship between the independent components and quality variables is established by kernel independent component regression, and the correlation matrix is obtained. Then, the kernel independent components can be suitably divided into quality‐related and quality‐unrelated parts. Finally, an analysis of the contribution of each variable to the statistics based on Lagrange's mean value theorem is presented. In addition, a numerical case and the Tennessee Eastman process (TEP) demonstrate the efficacy and superiority of the proposed method.

Voltage Correlation-Based Principal Component Analysis Method for Short Circuit Fault Diagnosis of Series Battery Pack

This paper concerns the issue of data-driven fault diagnosis for series lithium-ion (Li-ion) battery pack. A voltage correlation based statistical analysis method is proposed. First, the voltage of each cell within the battery pack is measured independently, and the correlation coefficient (CC) between the voltages of adjacent cells is calculated. Then, all the CC signals under normal conditions are used to train a principal component analysis (PCA) model, on the basis of which, a synthetic statistic and a kernel density estimation (KDE)-based threshold are designed for simultaneously monitoring all the measured CC signals at each sampling instant. Once a fault is detected, an accumulative relative contribution plot (ARCP) algorithm is immediately used to isolate which CC signal has a problem and locate the faulty cell through a cross-positioning strategy. The experimental results on a realistic test platform for series battery pack show that the new method provides accurate and reliable assessments for different fault specifics, and it performs better than the state-of-the-art CC methods in terms of parallel processing capability, sensitivity to weak short-circuit faults and robustness to window width.

Unsupervised Fault Detection of Pharmaceutical Processes Using Long Short-Term Memory Autoencoders

Unsupervised multilayer long short-term memory autoencoder (LSTM-AE) models are proposed for monitoring nonlinear batch processes. The methodology is demonstrated for a simulation-based study of an industrial-scale penicillin process and for an industrial vaccine manufacturing process, using production data. The LSTM-AE model was trained with two different loss functions: minimizing mean square error (MSE) between the input and reconstructed data and maximizing the average fault detection rate (FDR¯) in the training data set. Two algorithms are also proposed for obtaining contribution plots for the diagnosis of faults. For the industrial case study, where the faults are not known a priori, the contribution plots are found to be a valuable tool for identifying possible sources of faults. Furthermore, a semisupervised procedure has been proposed to select the normal process region for training the model. Two metrics are also presented to evaluate the performance of the proposed methodology: one for the simulator case study in which fault knowledge is available and one for the industrial case study in which fault knowledge is not available a priori. The proposed unsupervised algorithms exhibit a clear improvement in accuracy over linear methods or nonlinear techniques that do not explicitly account for dynamic behavior.

Multimode system condition monitoring using sparsity reconstruction for quality control

&lt;p&gt;In this paper, we introduce an improved multivariate statistical monitoring method based on the stacked sparse autoencoder (SSAE). Our contribution focuses on the choice of the SSAE model based on neural networks to solve diagnostic problems of complex systems. In order to monitor the process performance, the squared prediction error (SPE) chart is linked with nonparametric adaptive confidence bounds which arise from the kernel density estimation to minimize erroneous alerts. Then, faults are localized using two methods: contribution plots and sensor validity index (SVI). The results are obtained from experiments and real data from a drinkable water processing plant, demonstrating how the applied technique is performed. The simulation results of the SSAE model show a better ability to detect and identify sensor failures.&lt;/p&gt;

Efficient Sitting Posture Recognition for Wheelchair Users: An Unsupervised Data-Driven Framework

Automatic and reliable detection of a person's posture when sitting in a wheelchair is necessary to prevent major health issues. This study introduces an unsupervised anomaly detection and isolation approach to automatically recognize unbalanced sitting posture in a wheelchair using data from pressure sensors embedded in the wheelchair. Importantly, the advantages of independent component analysis (ICA) will be integrated with those of a Kantorovich Distance (KD)-driven anomaly detector by developing an ICA-driven KD methodology that can handle non-Gaussianity in the data and ameliorates the quality of anomaly detection. Due to pressure data displaying a non-Gaussian behavior, this work adopts ICA, which is well suited to handle this type of data. At the same time, the KD scheme is an effective anomaly detection indicator to evaluate the ICA residuals. Furthermore, the contribution plot strategy, which does not need a priori knowledge of anomalies, is employed for discriminating the type of the detected abnormal posture if it is caused due to higher pressure on the right side, on the left side, or higher forward pressure. The ICA-KD approach only employs normal events data to train the detection model, making them more attractive for identifying a person's posture in practice. The overall detection system provides a promising performance with an F1-score around 99.41%, outperforming some commonly used monitoring methods.

Development of a Machine Learning Model to Predict Recurrence of Oral Tongue Squamous Cell Carcinoma.

Despite diagnostic advancements, the development of reliable prognostic systems for assessing the risk of cancer recurrence still remains a challenge. In this study, we developed a novel framework to generate highly representative machine-learning prediction models for oral tongue squamous cell carcinoma (OTSCC) cancer recurrence. We identified cases of 5- and 10-year OTSCC recurrence from the SEER database. Four classification models were trained using the H2O ai platform, whose performances were assessed according to their accuracy, recall, precision, and the area under the curve (AUC) of their receiver operating characteristic (ROC) curves. By evaluating Shapley additive explanation contribution plots, feature importance was studied. Of the 130,979 patients studied, 36,042 (27.5%) were female, and the mean (SD) age was 58.2 (13.7) years. The Gradient Boosting Machine model performed the best, achieving 81.8% accuracy and 97.7% precision for 5-year prediction. Moreover, 10-year predictions demonstrated 80.0% accuracy and 94.0% precision. The number of prior tumors, patient age, the site of cancer recurrence, and tumor histology were the most significant predictors. The implementation of our novel SEER framework enabled the successful identification of patients with OTSCC recurrence, with which highly accurate and sensitive prediction models were generated. Thus, we demonstrate our framework's potential for application in various cancers to build generalizable screening tools to predict tumor recurrence.

Operating performance assessment method for industrial process with slowness principle-based LSTM network

Timely and accurate performance assessment and non-optimal regulation of industrial processes can effectively guarantee product quality. Most industrial processes are highly nonlinear and dynamic, so long short-term memory (LSTM) network is suitable for industrial performance assessment. However, in the network learning, the typical LSTM network focuses on the representation learning of input variables, lacks the representation of comprehensive economic indexes (CEI), and cannot selectively learn essential features, which increases the computational burden and easily mixes redundant information. Thus, a supervised slow feature analysis (SFA)-based LSTM (SSFALSTM) network is proposed for industrial operating performance assessment. By utilizing CEI information and SFA constraints, the network is guided to simultaneously learn features related to CEI and slow-changing features that reflect the inherent dynamics of the process. Further, cascade performance recognition model to construct the complete performance assessment framework. For the non-optimal performance, a reconstruction-based contribution plot method is proposed to identify the main cause variables and guide adjustment operations. Finally, the effectiveness of the proposed method is validated on the dense medium coal preparation process.

An Early Minor-Fault Diagnosis Method for Lithium-Ion Battery Packs Based on Unsupervised Learning

Dear Editor, Any fault of a battery system that is not handled timely can cause catastrophic consequences. Therefore, it is significant to diagnose battery faults early and accurately. Due to the complex nonlinear features and inconsistency of lithium batteries, traditional fault diagnosis methods usually fail to detect battery minor faults in the early stages. Therefore, this letter proposes a real-time unsupervised learning diagnosis approach for early battery faults based on improved principal component analysis. The technique rotates the battery pack voltage sequence into a new coordinate space through linear combination, while the detection metrics of square prediction errors and modified contribution plots are employed to achieve minor fault traceability. In addition, the training sample relies on the voltage sequence of the battery health state instead of the fault data, which is difficult to collect. Moreover, this approach can not only locate the battery cell where the fault occurs but also diagnose battery open-circuit and short-circuit faults as well as the occurrence and duration of the fault in real-time. Furthermore, the feasibility and stability of the proposed method are verified by applying different experimental data. In summary, the presented approach provides an easy-to-implement option that does not require accurate mathematical modeling, expert understanding, and complex computational processes.

Genetic algorithm-aided ensemble model for sensor fault detection and diagnosis of air-cooled chiller system

Due to the change of environment and operation conditions, sensor data in chiller system show the coexistence of gaussianity, non-gaussianity, non-linearity, and dynamic characteristic. Most of the existing methods can only be used for fault diagnosis based on a specific assumption, such as linearity or gaussianity. Therefore, this paper proposes a genetic algorithm (GA)-aided ensemble model without any characteristic assumption to detect and diagnose sensor fault in air-cooled chiller. Different from conventional techniques, the proposed method improves the detection and diagnosis performances by considering the Gaussian, non-Gaussian, non-linear, and dynamic characteristics, simultaneously. Firstly, several statistical models are analyzed and selected as the sub-models in our ensemble framework. The underlying assumptions for these models are different so that all the above characteristics can be considered. Then, an ensemble detection framework and an ensemble diagnosis framework based on GA are proposed to adapt to the complex data characteristics. To maximize the fault detection rate and diagnosis accuracy, two fitness functions of GA are proposed for detection and diagnosis, respectively. Finally, an integrated statistic and an integrated contribution plot are proposed based on the GA-aided ensemble framework. The detection and diagnosis performances of the ensemble statistic and the ensemble contribution plot are validated by using the data collected from a real chiller system. Results show that the proposed method delivers superior performance compared with other methods.

Industrial Application of Data-Driven Process Monitoring with an Automatic Selection Strategy for Modeling Data

The increasing scale of industrial processes has significantly motivated the development of data-driven fault detection and diagnosis techniques. The selection of representative fault-free modeling data from operation history is an important prerequisite to establishing a long-term effective process monitoring model. However, industrial data are characterized by a high dimension and multimode, and are also contaminated with both outliers and frequent random disturbances, making automatic modeling data selection a great challenge in industrial applications. In this work, an information entropy-based automatic selection strategy for modeling data is proposed, based on which a general real-time process monitoring framework is developed for a large-scale industrial methanol to olefin unit with multiple operating conditions. Modeling data representing normal operating conditions are automatically selected with only a few manually defined normal samples. A long-term effective process monitoring model is then established based on a multi-layer autoencoder, through which unexpected disturbances in real-time operation can be detected early and the root cause can be preliminarily diagnosed by contribution plots. The adjustment of operating conditions has also been considered through a model update strategy. Details of the proposed data selection strategy and modeling process have been provided to facilitate the industrial application of process monitoring systems by other researchers or companies.

Hierarchical Causal Graph-Based Fault Root Cause Diagnosis and Propagation Path Identification for Complex Industrial Process Monitoring

Modern complex industrial processes are increasingly integrated and automated. There are numerous control loops in the process system and the closed-loop feedback makes units strongly coupled. This makes the faults propagate and evolve with the material and energy flow networks. Thus, it poses a great challenge to the fault root cause diagnosis and propagation path identification for industrial processes. This paper proposes a fault root cause diagnosis and propagation path identification strategy for complex industrial process based on hierarchical causal graph (HCG). Firstly, the process is divided into several sub-blocks according to process knowledge. The monitoring model is built for fault detection. Secondly, the fault-related sub-blocks and variables are selected based on contribution plot, and the causal analysis based on transfer entropy is performed on them to construct HCG. In this way, the causal connections can be significantly reduced. Subsequently, a propagation path identification method based on causal feature long short-term memory (CF-LSTM) is designed. Finally, real data from hot strip mill process (HSMP) and their comparative experiments are adopted to validate the fault diagnosis performance of the proposed framework.

Dynamic monitoring and anomaly tracing of the quality in tobacco strip processing based on improved canonical variable analysis and transfer entropy.

Multivariate statistical monitoring methods are proven to be effective for the dynamic tobacco strip manufacturing process. However, the traditional methods are not sensitive enough to small faults and the practical tobacco processing monitoring requires further root cause of quality issues. In this regard, this study proposed a unified framework of detection-identification-tracing. This approach developed a dissimilarity canonical variable analysis (CVA), namely, it integrated the dissimilarity analysis concept into CVA, enabling the description of incipient relationship among the process variables and quality variables. We also adopted the reconstruction-based contribution to separate the potential abnormal variable and form the candidate set. The transfer entropy method was used to identify the causal relationship between variables and establish the matrix and topology diagram of causal relationships for root cause diagnosis. We applied this unified framework to the practical operation data of tobacco strip processing from a tobacco factory. The results showed that, compared with traditional contribution plot of anomaly detection, the proposed approach cannot only accurately separate abnormal variables but also locate the position of the root cause. The dissimilarity CVA proposed in this study outperformed traditional CVA in terms of sensitiveness to faults. This method would provide theoretical support for the reliable abnormal detection and diagnosis in the tobacco production process.