Remaining useful life prediction using hybrid neural network and genetic algorithm approaches

Recently, numerous approaches have been applied for predicting the RUL of machinery based on condition information. Artificial Intelligence (AI) methods such as Long Short-Term Memory (LSTM), Feed Forward Neural Network (FFNN), Convoluted Neural Network (CNN), Recurrent Neural Network (RNN), and many more have been applied successfully in detecting the faults and predicting the RUL of machines. But these methods involve uncertainties in RUL prediction due to the inability to select the best input and suboptimal Artificial Neural Network (ANN) structures. The manual method of optimizing the ANN structure is time taking preprocessing to formulate the prediction model. To sort out these issues, this paper proposes a hybrid ANN and Genetic Algorithm approach to select the best input and optimize the ANN structure for higher accuracy. The open-source simulated data sets of realistic large commercial turbofan engines have been used in the proposed network of ANN with GA. GA selects the case to trim down the data dimensionality. The application of GA intends to optimize the hyperparameters of ANN to make more accurate networks. This hybrid method has been implemented in a Jupyter Notebook Anaconda software environment and the language used is python. The outcomes of simple ANN and hybrid method of ANN and GA are compared and found that the latter approach provides better RUL than the former.

Automated summary

This paper proposes a hybrid ANN and Genetic Algorithm approach to improve the accuracy of predicting the Remaining Useful Life (RUL) of machinery. By using real data sets, the ANN structure is optimized and the best input is selected, leading to improved accuracy in predicting RUL.