An Integrated Distributed Fault Diagnosis Framework for Large-Scale Industrial Processes Based on Spatio–Temporal Causal Analysis

Spatio-temporal information analytics based performance-driven industrial process monitoring framework with cloud–edge-device collaboration

Multimode Process Monitoring and Fault Diagnosis Based on Tensor Decomposition

Large-scale industrial processes are characterized by complex reaction mechanisms and strong correlation among operational units, which makes it more difficult to extract the temporal correlation of process data, thereby affecting the process monitoring performance. Thus, a novel temporal variational autoencoder network based process monitoring method for large-scale industrial processes is proposed. First, the large-scale industrial processes are divided into multiple sub-blocks based on the spatial distribution of operational units. Second, the consistent trend feature analysis method is introduced to quantify the degree of consistency between adjacent trend features by calculating the differences among trend vectors. Then, the differences between trend features are integrated into the variational autoencoder to extract temporal features from each sub-block. Finally, Bayesian fusion strategy is employed to fuse the temporal features from all sub-blocks, constructing a comprehensive global monitoring indicators for process monitoring. Experimental results from an actual hot strip mill process demonstrate that the proposed method outperforms traditional approaches in monitoring performance.

Focusing on quality-related complex industrial process performance monitoring, a novel multimode process monitoring method is proposed in this paper. Firstly, principal component space clustering is implemented under the guidance of quality variables. Through extraction of model tags, clustering information of original training data can be acquired. Secondly, according to multimode characteristics of process data, the monitoring model integrated Gaussian mixture model with total projection to latent structures is effective after building the covariance description form. The multimode total projection to latent structures (MTPLS) model is the foundation of problem solving about quality-related monitoring for multimode processes. Then, a comprehensive statistics index is defined which is based on the posterior probability of the monitored samples belonging to each Gaussian component in the Bayesian theory. After that, a combined index is constructed for process monitoring. Finally, motivated by the application of traditional contribution plot in fault diagnosis, a gradient contribution rate is applied for analyzing the variation of variable contribution rate along samples. Our method can ensure the implementation of online fault monitoring and diagnosis for multimode processes. Performances of the whole proposed scheme are verified in a real industrial, hot strip mill process (HSMP) compared with some existing methods.

Permanent magnet synchronous machines (PMSM) represent the core of many industrial processes. Therefore, the monitoring of electrical and mechanical parameters and variables allows to obtain important information for a predictive maintenance and fault detection. However, some parameters are not directly measurable and need to be estimated. For this purpose, it is good to define a parameter estimation approach relevant in large-scale industrial processes. In this paper a parameter identification of IoT-Embedded PMSMs approach is defined. This approach combines the benefits of IoT, edge computing and cloud computing technologies. AWS services are used in the implementation of the parameter identification algorithm in cloud environment. So, a scalable, reliable, and secure solution is obtained to be applied in contexts related to facility management.

Shape setup model (SSM) plays a critical role to achieve satisfactory precision of strip shape in hot strip mill process (HSMP). In this paper, the process of SSM is introduced, taking 1700mm HSMP line in Ansteel as the research object. Special attention is focused on the gap crown model, which is the foundation of SSM. And the allocation strategy of reference values for mechanical actuators is presented. Then an off-line simulation method is proposed to obtain several key parameters of the model. SSM has been put into practical use and proved to be effective to control strip flatness and profile by high accuracy.

Variable selection constitutes an essential step to reduce dimensionality and improve performance of fault detection and diagnosis in large scale industrial processes. For this reason, in this paper, variable selection approaches based on causality are proposed and compared, in terms of model adjustment of available data and fault detection performance, with several other filter-based, wrapper-based, and embedded-based variable selection methods. These approaches are applied in a simulated benchmark case and an actual oil and gas industrial case considering four different learning models. The experimental results show that obtained models presented better performance during the fault detection stage when variable selection procedures based on causality were used for purpose of model building.

Decentralized adaptively weighted stacked autoencoder-based incipient fault detection for nonlinear industrial processes

As a result of slow perturbations, the mathematical model of an actual system is difficult to be accurate. So when optimizing large-scale industrial process, the mathematical model and the actual system does not match, that is model-actual difference. Large-scale industrial process optimization based on fuzzy model is an effective way of this issue. However, the optimization model is the process of establishing a non-linear programming model. So, differential evolution algorithm is studied this paper to solve the problems of large-scale industrial processes optimization based on fuzzy models. To solve model-actual differences is mainly to solve fuzzy nonlinear problems: firstly, differential evolution algorithm is studied for solving fuzzy nonlinear problems in this paper. Then the combination of fuzzy nonlinear problems and the interaction balance method coordinated approach of large-scale industrial processes is proposed. Last, simulation results show the validity of the method which this paper studies.

With the continuous evolution of modern industrial technology, industrial production has grown progressively complex, necessitating the use of various sensors to measure multiple process variables. However, intricate temporal dependencies and nonlinear relationships between data presented by multivariate sequences pose significant challenges to process fault detection. In response to these challenges, this paper proposes an attention-based gated recurrent unit autoencoder with skip connection (SAGRU-AE) model for monitoring large-scale, nonlinear, and multivariate industrial process faults. SAGRU-AE combines gated recurrent units, multi-head self-attention, and autoencoder to extract features from multivariable time series data efficiently. Concurrently, feature reuse is achieved through the skip connection structure, which improves the accuracy of data reconstruction. Based on the implementation of process data feature extraction and input reconstruction in SAGRU-AE, two statistics have been developed, namely the H2 statistic and square prediction error (SPE) statistic, for fault detection tasks. Ultimately, the feasibility and effectiveness of the proposed algorithms are validated through experimentation on the TE process.

In view of the large-scale and high-dimensional problems of industrial data and fault-tracing problems, a fault detection and diagnosis method based on multi-block probabilistic kernel partial least squares (MBPKPLS) is proposed. First, the process variables are divided into several blocks in a decentralized manner to address the large-scale and high-dimensional problems. The probabilistic characteristics and relationship between the corresponding process variables and the quality variables of each block are analyzed using latent variables, and the PKPLS model of each block is established separately. Second, the MBPKPLS model is applied to process monitoring, statistics of each block are established in a high-dimensional space, and the monitoring indicators in each block are used to detect faults. Third, based on fault detection, the multi-block concept is further used to locate the cause of fault, thereby solving the problem of fault tracing. Finally, a numerical example and the penicillin fermentation process (PFP) are used to test the effectiveness of the MBPKPLS method. The results demonstrate that the proposed method is suitable for processing large-scale, high-dimensional data with strong nonlinear characteristics, and the MBPKPLS process monitoring method is effective for improving the performance of fault detection and diagnosis.

SummaryA fault diagnosis and compensation problem for two‐dimensional discrete time systems with time‐varying state delays is studied in this paper. The concerned two‐dimensional systems are described by the Fornasisi–Marchesini second model and are subject to unknown disturbances. First, a fault detection and diagnosis module is designed to obtain the information on sensor faults; a new fault detection and diagnosis integrated design, using the observer based on descriptor system approach, is proposed to detect and estimate the sensor faults. The integrated design can maximize the fault detection rate for a given false alarm rate. Sufficient conditions for the existence of the integrated fault detection and diagnosis design are derived in the context of norm evaluation and provided in terms of matrix inequalities. Second, a fault‐tolerant control module is proposed upon an existing output feedback controller. When the sensor fault occurs, the faulty measurement can be identified and corrected by the proposed fault detection and diagnosis module. In this case, the feedback controller can guarantee the performance of the closed‐loop system even when encountering sensor faults. Finally, the proposed method is applied to a thermal process to illustrate its effectiveness. Copyright © 2017 John Wiley & Sons, Ltd.

Fault detection and diagnosis in process data using one-class support vector machines

Thousands of steel strips are processed through the Port Talbot Steelworks’ Hot Strip Mill every year. Hundreds of these, however, are either scrapped, sold at a concession, or require further processing due to a variety of defects, a significant percentage of which are width-related. Tata Steel has identified this process as being abundant with underutilised data and is currently investing in innovative research and development which aims to address this issue as well as other industry-wide challenges including decarbonisation and digitisation of legacy processes. Currently, Artificial Intelligence and Machine Learning has become a popular technology for identifying and analysing defects in manufacturing processes. However, the scope of applications which utilise these technologies within the steel industry are currently limited and are often contained to specific subprocesses. Simultaneously, gathering expert knowledge from employees with first-hand experience is crucial to understanding defects and how they can be analysed. This thesis aims to expand the scope of Machine Learning tools for defect analysis in steel-making processes and showcase their potential when combined with expert knowledge on a process-wide scale. This is achieved by understanding the challenges of collecting and utilising industrial data, exploring the technologies available for various types of analyses, and reviewing the history and development of these technologies in manufacturing and in the steel industry. From this, technologies and analyses specific to the Port Talbot Hot Strip Mill process are selected, providing a platform on which tools for the analysis of width-related defects are developed. An in-depth understanding of these defects and their root causes is gained through continual communication with employees at the Port Talbot Steelworks and, using this, several Machine Learning models are created for the purpose of identifying them in several Hot Strip Mill subprocesses. Further to this, a new, data-driven decision-making process is developed for the purpose of identifying width-related defects and determining their root causes across the entire Hot Strip Mill process as they occur.

Industrial equipment fault diagnosis faces dual challenges: significant data distribution discrepancies caused by diverse operating conditions impair generalization capabilities, while underutilized spatio-temporal information from multi-source data hinders feature extraction. To address this, we propose a spatio-temporal collaborative perception-driven feature graph construction and topology mining methodology for variable-condition diagnosis. First, leveraging the operational condition invariance and cross-condition consistency of fault features, we construct fault feature graphs using single-source data and similarity clustering, validating topological similarity and representational consistency under varying conditions. Second, we reveal spatio-temporal correlations within multi-source feature topologies. By embedding multi-source spatio-temporal information into fault feature graphs via spatio-temporal collaborative perception, we establish high-dimensional spatio-temporal feature topology graphs based on spectral similarity, extending generalized feature representations into the spatio-temporal domain. Finally, we develop a graph residual convolutional network to mine topological information from multi-source spatio-temporal features under complex operating conditions. Experiments on variable/multi-condition datasets demonstrate the following: feature graphs seamlessly integrate multi-source information with operational variations; the methodology precisely captures spatio-temporal delays induced by vibrational direction/path discrepancies; and the proposed model maintains both high diagnostic accuracy and strong generalization capacity under complex operating conditions, delivering a highly reliable framework for rotating machinery fault diagnosis.