We consider the problem of fault detection and isolation for the penicillin fermentation process. A penicillin fermentation process is a highly complex and nonlinear dynamic process with batch processing. A data-driven approach is utilized for fault diagnostics due to the availability of huge batch processing data and the unavailability of an analytical model. To address the non-linearity, a subspace-aided parity-based residual generation technique is proposed by using a just-in-time learning approach. For the just-in-time learning approach, the most similar data samples are selected from the database for incoming test samples and a subspace aided parity-based residual is generated using these samples. The designed fault detection technique is implemented for the penicillin fermentation process to demonstrate real-time health monitoring of the process under sensor noise and process disturbances. Two sensor faults and an actuator fault are considered and successfully detected using the proposed technique. Further, a fault isolation approach is developed to isolate these faults and their location has been identified. A case study is given to show the improvement offered by the proposed technique for the fault detection rate and minimization of the false alarm rate as compared to the existing techniques for the penicillin fermentation process.

This paper discusses a framework for fault detection in nonlinear processes. A novel subspace aided parity-based data-driven technique is proposed using the so-called just-in-time learning (JITL) approach. The idea of the proposed technique is to choose a set of the most similar data samples from the database for each test data sample using the JITL approach to address the nonlinearity problem. The parity vector is constructed for each test sample using the selected data samples to generate the residual signal. The salient features of the proposed technique include easy and simple implementation together with effectiveness in fault detection in the presence of disturbances and measurement noise. The fault detection scheme is so designed that it exhibits robustness against sensor noise and disturbances and sensitivity to faults. A case study for the fault detection of a nuclear research reactor (NRR) is presented to demonstrate the efficacy of the technique. The NRR is a highly nonlinear and complex process. Two faults of NRR, namely external reactivity insertion and control rod withdrawal, are successfully detected using the proposed approach.

Fault diagnosis and tolerance are crucial for monitoring system health and ensuring stability in industrial processes. Challenges arise in designing fault diagnostic solutions for real-time industrial processes with inherent nonlinear dynamic behaviors, particularly when dealing with multiple operating regions characterized by varying dynamics. This article addresses this challenge and proposes a fault diagnostic and tolerant control scheme for industrial systems. The proposed approach integrates a fuzzy-based realization technique with a subspace-aided methodology to effectively handle the nonlinear dynamic behavior observed across different operational scenarios. A practical solution is presented, significantly reducing the computational burden associated with online diagnostics, as the parity vectors are computed offline using available input–output data for different operating regions. During online diagnostics, only computed parity spaces are used with fuzzy realizations for residual generation, leading to a significant reduction in online computation. Numerical examples demonstrate the effectiveness of the proposed method, achieving a high precision rate in fault diagnostics. Furthermore, the diagnostic methodology is integrated with fault-tolerant control for practical applications, as demonstrated in the application of a continuous stirred tank reactor. This integration enables the system to effectively tolerate faults and ensure sub-optimal operation of the industrial process.

The article introduces a hybrid fault detection and prognosis system design for motor bearings, which is crucial for the reliability and efficiency of rotating machinery. The focus is on the concern of bearing fault detection with limited data quantity. Traditional approaches struggle with fault detection depending on the quantity and quality of historical data samples, especially in dynamic conditions, often failing to capture long-term dependencies and uncertainties in bearing health states, thereby limiting precise fault detection and remaining useful life predictions. To overcome these challenges, the research proposes a novel hybrid framework that combines data-driven and model-driven learning. This framework employs the Siamese neural network (SNN) containing two identical subnetwork models for fault detection and Bayesian estimation for prognosis purposes, bridging the gap between data-driven and model-based approaches. SNNs excel in learning from limited labeled data and are suitable for real-time fault detection. Experimental validation with the PRONOSTIA-FEMTO bearings dataset confirms the framework’s superior performance in fault detection, promising improved maintenance practices and equipment reliability in industrial processes.