Genetic Algorithm-Optimized CNN-BiLSTM Framework for Predicting the Remaining Useful Life of IGBT Modules

Area of Science:

• Electrical Engineering
• Materials Science
• Artificial Intelligence

Background:

• Insulated Gate Bipolar Transistors (IGBTs) face aging and failure issues during long-term operation.
• Accurate Remaining Useful Life (RUL) prediction is crucial for maintaining system reliability and preventing failures.

Purpose of the Study:

• To develop an advanced method for predicting the RUL of IGBTs.
• To enhance the accuracy and efficiency of RUL prediction models.

Main Methods:

• A hybrid model combining Convolutional Neural Network (CNN) for spatial feature extraction and Bidirectional Long Short-Term Memory (BiLSTM) for temporal dependency analysis was developed.
• A Genetic Algorithm (GA) was employed to optimize the hyperparameters of the CNN-BiLSTM model, including the initial learning rate.
• The proposed method was benchmarked against existing RUL prediction techniques.

Main Results:

• The proposed CNN-BiLSTM model demonstrated superior performance across all evaluation metrics compared to benchmark algorithms.
• The genetic algorithm significantly accelerated the parameter optimization process.
• Overall training efficiency of the RUL prediction model was enhanced.

Conclusions:

• The hybrid CNN-BiLSTM model optimized with a genetic algorithm offers a robust and efficient solution for IGBT RUL prediction.
• This approach effectively addresses the challenges of IGBT aging and failure, contributing to improved device reliability and predictive maintenance strategies.