On Intelligent Fault Detection in Industrial Processes

Abstract

The evolution of sensor and monitoring technologies has opened the door for developing advanced methods in order to enhance fault detection and diagnosis of different industrial systems. The usage of real-time sensor data to capture the health status of a system using data-driven techniques has been intensely investigated in both literature and practice. However, several open research issues remain to be unresolved. This thesis develops and employs advanced data-driven approaches to tackle the fault detection problem from different perspectives. In addition, since the condition of any system deteriorates over time due to aging and environmental effects, this dissertation proposes a new reliability monitoring scheme that is able to detect any system deterioration timely.

The thesis consists of four studies. In essence, each study represents a new perspective motivated by a reliability-related problem that is regularly encountered in the industry.

The first study focuses on the abnormal Control Chart Patterns (CCP) inevitably encountered while monitoring a process. These patterns point out manufacturing faults that can lead to significant internal and external failure costs unless treated promptly. Thus, detecting such abnormalities is of utmost importance. Machine learning algorithms have been widely applied to this problem. Nevertheless, the existing CPP recognition methods can only deal with a fixed input size rather than different input sizes according to the actual production needs. In order to tackle this problem, an original CPP recognition method relying on Convolutional Neural Network (CNN) named Variable Input Size (VIS)-CNN is proposed. Signal resizing is performed using resampling methods, and then CNN is used to extract the abnormal patterns in the datasets. Five different input sizes are generated for model training and testing. The optimal hyperparameters as well as the best structure of the used CNN are optimized using Bayesian optimization. Simulation results show that the correct recognition rate of the VIS-CNN is 99.78%, based on different window size control charts. Furthermore, we address the issue of the mixed CCP and provide a modified scheme to achieve a high recognition ratio for 8 mixed patterns, on top of 6 standard patterns. The modified scheme includes wavelet noise reduction and Adaptive Boosting. An actual case study on metal galvanization process is presented to show that the method has potential applications in industrial environments.

The second work proposes a Variational Autoencoder-Long Short Term Memory (VAE-LSTM) based chart for monitoring time-dependent high dimensional processes. Appropriate monitoring of such processes allows for efficient decision-making that can improve the baseline of manufacturing companies, either through decreasing production costs or enhancing production efficiency. Various latent variable-based methods have been proposed for addressing high dimensional data; however, many of these methods assume that the data are independent and normally distributed. The violation of these assumptions results in an increased false alarm rate, in addition to the deterioration in the performance of such methods for detecting anomalies. We propose a Variational Autoencoder-Long Short Term Memory (VAE-LSTM) deep learning based T² chart that integrates the unique features of both VAE and LSTM for intelligent fault detection of time-dependent high dimensional processes. The effectiveness and applicability of the proposed model are demonstrated through extensive simulations, an open-source dataset, and a real case study.

The third work focuses on uncertainty utilization in fault detection using Bayesian deep learning networks. Up to now, the actual usage of deep learning in manufacturing sites is somehow restrained by the quality of the obtained data, especially for machine failure cases. This research proposes an approach for utilizing the prediction uncertainty information generated by Bayesian deep learning models to improve decision-making in fault detection. Inference is carried out using Automatic Differentiation Variational Inference, and the resultant prediction uncertainty information is utilized to enhance fault detection. The proposed approach is applied to an open-source dataset and a real case study on vertical continuous plating of printed circuit boards. The experiments show that the performance of the proposed scheme is considerably beneficial compared to classical deep learning models.

The fourth and final work deals with monitoring event-based data for imperfect maintained systems. The state of any system deteriorates with time; hence preventive maintenance is often conducted to restore the system’s condition and prevent failures. Preventive maintenance activities are usually imperfect, meaning that the system failure rate increases with the increase in the number of repairs being carried out. Monitoring the time between failures can help in deciding whether further maintenance is beneficial for a system or if it has deteriorated to a state where preventive maintenance is no longer effective. Therefore, we introduce a time between events monitoring scheme for imperfect maintained Weibull distributed systems. We study the effect of changes in both the shape and scale parameters of the Weibull distribution in terms of the Average Run Length (ARL), Standard Deviation of the Run Length (SDRL) and provide a case study on a painting shop where a robotic system is operated. The proposed scheme showed a better performance for monitoring imperfect maintained systems than its counterparts.

With a series of new and practical perspectives related to fault detection and reliability analysis of industrial systems, this thesis aims to bridge existing research gaps that should be of interest to academicians as well as practitioners.

Date of Award	30 Dec 2022
Original language	English
Awarding Institution	City University of Hong Kong
Supervisor	Min XIE (Supervisor)