A back propagation neural network based low frequency clustering man-hour prediction method for digital manufacturing

With the transition to intelligent and digital manufacturing, traditional work hour prediction methods are often unable to effectively handle complex production environments and sparse data patterns. In this paper, a low-frequency clustering work hour prediction method based on backpropagation neural network (BP neural network) is proposed to provide accurate work hour prediction tools for the next generation manufacturing industry. First, due to the problems such as data missing, data exception and so on in man-hour feature data, it is necessary to conduct integrity verification and data quality optimization towards the man-hour data, in order to ensure the accuracy and diversity of the input data for man-hour prediction. Second, regarding the characteristics of high dimensionality and redundancy of man-hour feature data, a feature integration method based on Spearman coefficient is proposed to reduce data dimensionality. Finally, the firefly algorithm is used to improve the K-Means clustering algorithm, dividing the integrated features to construct man-hour prediction models, which can reduce the model complexity and improve the prediction accuracy. By comparing the prediction errors before and after the feature integration and clustering, the accuracy of the man-hour prediction method is verified, implying a positive significance for improving the machining man-hour prediction.