A variable mosquito flying optimization-based hybrid artificial neural network model for the alarm tuning of process fault detection systems

Chemical process systems are becoming extremely complex due to increased automation, heat and mass intensification, and expectation of higher efficiency. Many fault detection and diagnostic methods have been proposed for processing facilities. However, managing the missed alarm rate and the false alarm rate (FAR) in the detection and isolation of the fault is crucial in the complex process systems. This work presents a new data-driven fault detection model using an artificial neural network (ANN) and variable mosquito flying optimization (V-MFO) technique. The model is based on the optimization of the number of neurons in the hidden layer of the neural network. Subsequently, the model parameters have been tuned using the V-MFO algorithm for maximizing the fault detection rate (FDR) while minimizing the FAR. The proposed fault detection method has been implemented on the Tennessee Eastman benchmark process. The performance of the proposed model has been evaluated in terms of accuracy, FDR and FAR against well-known statistical-based methods such as principal component analysis (PCA), kernel PCA, semiparametric PCA, modified independent component analysis, k nearest neighbors, linear discriminant analysis, support vector machine, and the ANN. The model is observed to be competitive for fault detection among the test algorithms. It recorded slightly improved accuracy and FDR. The proposed model also resulted in 0.6% improvement in the FAR and 8% improvement in missed detection rate compared to the simple ANN. This method provides an efficient fault detection tool for complex process systems.