Application of Genetic Neural Network for Diagnosis of Anode Anomaly and Metal Wave in Aluminum Electrolysis

This paper presents a neural network model based on the cell resistance signal. In the model, the anode current spectral energy is set as the input vector; the normal production, anode anomaly and metal wave are set as the output sample. By using genetic algorithm to optimize the initial weights and thresholds of the network, the model realized the diagnosis of the anode anomaly and metal wave in the production of aluminum electrolysis. The results show that the non-optimized neutral network needs to be trained 3131 times to achieve the specified precision and running time is 388s. Then the one optimized by genetic algorithm needs to be trained 2571 times to achieve the specified precision and running time is 222s. The results of the diagnosis system applied to the 350kA aluminum electrolysis production show that the diagnostic accuracy is as high as 80%, basically meet the needs of the production process. Keywords-neutral network; genetic algorithm; aluminum electrolysis; fault diagnosis component