CNN-LSTM-Attention with PSO optimization for temperature and fault prediction in meat grinder motors

Meat grinders are essential in meat production, and predicting the working parameters of the meat grinder motors can help prevent potential failures by providing timely warnings. In this paper, a deep learning model, CNN-LSTM-AP, is developed, combining convolutional neural network (CNN), long short-term memory network (LSTM), attention mechanism (Attention), and particle swarm optimization (PSO). The Attention mechanism is used to assign weights to input features, enhancing the model focus on important data. The CNN and LSTM networks are used to extract features and capture long-term dependencies in time series, respectively. Additionally, the PSO algorithm optimized the number of neurons and the learning rate in the model. For early fault detection, the Mahalanobis distance-based function mapping method is established, along with the PMT monitoring index, early warning, and alarm threshold. Experimental results show that the mean absolute error (MAE) is reduced by 0.5656 and the coefficient of determination (R2) is increased by 0.0619 in the CNN-LSTM-AP compared with other 5 prediction models. In fault scenarios, the PMT index drops below the warning threshold five consecutive timepoints before failure, enabling proactive alerts. The deep learning model and warning method proposed in this paper can accurately predict the condition of the meat grinder and issue a warning of potential faults based on the PMT data, offering a novel approach to the health management of meat processing equipment.