Fault Diagnosis of the Dynamic Chemical Process Based on the Optimized CNN-LSTM Network

Abstract

Deep learning provides new ideas for chemical process fault diagnosis, reducing potential risks and ensuring safe process operation in recent years. To address the problem that existing methods have difficulty extracting the dynamic fault features of a chemical process, a fusion model (CS-IMLSTM) based on a convolutional neural network (CNN), squeeze-and-excitation (SE) attention mechanism, and improved long short-term memory network (IMLSTM) is developed for chemical process fault diagnosis in this paper. First, an extended sliding window is utilized to transform data into augmented dynamic data to enhance the dynamic features. Second, the SE is utilized to optimize the key fault features of augmented dynamic data extracted by CNN. Then, IMLSTM is used to balance fault information and further mine the dynamic features of time series data. Finally, the feasibility of the proposed method is verified in the Tennessee-Eastman process (TEP). The average accuracies of this method in two subdata sets of TEP are 98.29% and 97.74%, respectively. Compared with the traditional CNN-LSTM model, the proposed method improves the average accuracies by 5.18% and 2.10%, respectively. Experimental results confirm that the method developed in this paper is suitable for chemical process fault diagnosis.

Odor-Cued Grab Air Sampling for Improved Investigative Odorant Prioritization Assessment of Transient Downwind Environmental Odor Events

“…To date, conventional efforts to assess transient odor events have primarily utilized the analytical approach (i.e., direct sampling from the air) to begin the priority odorant assessment. However, the transient event characteristic also magnifies the challenge associated with follow-up investigation of citizen odor complaints utilizing human "sensors" and dynamic dilution olfactometry (DDO) (ASTM E-679; 27 ASTM E-1432; 28 and CEN 29 ). Typically, agency officials receive a complaint from downwind citizens and possibly assign an investigator to travel to the complaint site.…”

Research into Prediction Method for Pressure Pulsations in a Centrifugal Pump Based on Variational Mode Decomposition–Particle Swarm Optimization and Hybrid Deep Learning Models

“…The analysis process of 1DCNN is illustrated in Figure 1, which shows a multidimensional matrix of the input time series data. The red and blue colors represent different filters [25]. After the filters perform convolutional operations with the input data, the extracted feature dimension becomes N × 1, where N depends on the input data dimension, filter size, and convolution stride.…”

An Improved Ship Classification Method Based on YOLOv7 Model with Attention Mechanism

“…Recently, with the rapid development of DL in various fields [3][4][5], it has made significant breakthroughs in image detection, gradually solving the problems of slow training speed and low detection accuracy of object detection. Detection algorithms based on DL are divided into two categories:…”