Agriculture Rotary Tedder Fault Diagnosis using Multiple Input - Multiple Type - Parallel - Convolutional Neural Network Optimized by Genetic Algorithm

For several years, fault diagnosis methods have been increasingly using methods based on neural networks. These methods have many advantages, but the process of creating their architecture, training and testing requires a lot of expert knowledge and is time-consuming, especially when building a neural network by trial and error, as is usually done with new types of machines. For this reason, we propose an innovative method to solve this problem by automating the previously mentioned steps of building a neural network, adapting it to the available data and selecting the best training method. Automation is carried out using an appropriately adapted genetic algorithm that generates the entire architecture of the neural network, and this neural network can have multiple inputs, have parallel input data processing paths and handle different types of input data at the same time. Our algorithm was tested using highly noisy data from a field agricultural rotary tedder, and the resulting neural networks were able to detect the type of machine fault with an accuracy of over 80%.