Dynamic Process Safety Assessment Using Adaptive Bayesian Network with Loss Function

Abstract

Fault detection and diagnosis (FDD) is crucial for dynamic process safety analysis. Integrated with failure prediction models, it enables us to realize how a deviation in process variable(s) can affect system safety (measured as risk). This work aims to overcome the challenges of nonlinear, non-Gaussian, and multimodal behavior of the processing systems to detect abnormal process operations, predict dynamic operational risk, and diagnose root cause of the abnormal situation. A methodology is proposed here by integrating different techniques. The artificial neural network (ANN) is used to identify process modes, while the Bayesian network (BN) is used for fault detection. How a fault will lead to a process failure is modeled using the event tree (ET), whereas time-dependent losses associated with the failure scenarios are assessed using the inverted normal loss function (INLF). A probability adaption mechanism is used to estimate the conditional probabilities in each time slice. The complexity of estimating conditional probabilities is handled using the copula theory. The proposed framework is validated using numerical, simulated, and industrial datasets. The results suggest that the developed framework can provide greater flexibility and wider applications.

A robust neural network model for fault detection in the presence of mislabelled data

“…Twenty distinct known and unknown faults (IDV-1 to IDV20) have been studied in several investigations. [66,[68][69][70] The proposed robust ANN model has been examined for detecting the step fault (IDV-1) and the sticking fault (IDV-14) of the TE process using 500 normal and 500 faulty samples. The real and mislabelled data of IDV-1 and IDV-14 faults of the TE process have been presented in Figures 7 and 8, respectively.…”

Process safety analysis using operational data and Bayesian network

“…Generally speaking, these cannot proactively predict failure unless the initiating event happens. Efforts are ongoing on how to integrate data‐driven FDD models with these logical failure prediction models for dynamic risk monitoring and early failure prognosis (FP) 20,21 …”

“…Efforts are ongoing on how to integrate data-driven FDD models with these logical failure prediction models for dynamic risk monitoring and early failure prognosis (FP). 20,21 Although data-driven methods are widely used to describe and validate anecdotes of accident prevention, there is a lack of understanding among early Chemical Engineering students (many of whom will be future process safety torchbearers) about how these methods can be integrated with the commonly used software (e.g., Aspen HYSYS). Dynamic process simulation can improve safety and reliability.…”

Data-Knowledge-Driven Modeling and Operational Adjustment for the Pharmaceutical Tablet Manufacturing Process via Wet Granulation

“…In recent years, data-knowledge-driven approaches have attracted increasing attention and interest in academia and have achieved the broad applications in solving various engineering problems such as system modeling and control, process monitoring, fault diagnosis, and so on. − Specifically, as an important artificial intelligence (AI) technique, the Bayesian network (BN) is a type of probabilistic graphical model that is capable of effectively integrating data and knowledge to simulate human reasoning. It represents the causal relations of variables by a directed acyclic graph, and so it has better interpretability than other AI methods. , Owing to the advantages in interpretability, probabilistic modeling, and dealing with data uncertainties, BNs have been widely applied to a variety of industrial systems and processes in different areas to successfully solve problems such as process monitoring, fault diagnosis, prognosis, risk assessment, decision making, etc. ,, However, the BN-based PQC studies in the pharmaceutical field are rarely reported. Because of its significant advantages and application potential, we attempt to utilize BN for PQC (or operational adjustment) tasks for the first time.…”