Abstract:

Objective: In order to achieve more efficient and low-carbon intelligent manufacturing of mass customization furniture, it is necessary to implement more accurate intelligent production decisions on the basis of automated and continuous production, promote the production management from the traditional batch level to the part level, and carry out production scheduling based on the differentiated process time. By analyzing the key influencing factors of the NC (numerical control) drilling operation time, this paper constructs a mathematical model that can predict the NC drilling operation time of each part, so as to obtain more accurate operation time data before production, solve the practical problem of the missing operation time of custom furniture, and lay a solid data foundation for the part level production scheduling of custom furniture. Method: By conducting in-depth research around the actual manufacturing of custom furniture, starting from the operation logic of the NC drilling machining center, the paper analyzes the key influencing factors of the NC drilling operation time under different data conditions, collects more than 60 000 historical operation time data from the bottom of the equipment through MES (manufacturing execution system), constructs an analytic program to extract the key influencing factors, builds a three-layer artificial neural network model using the Mish activation function, and introduces momentum and learning rate attenuation algorithms to improve the model accuracy and operation efficiency. The prediction of the part level operation time in the NC drilling process of custom furniture is realized. Result: Under different data conditions, good prediction results are achieved based on artificial neural network, among which the prediction error rate of the average working time of the 9-dimensional variables extracted based on the NC program is 2.68%, which has the ability to predict the overall efficiency of the process. The mean absolute error of single part prediction results is 2.87 s, which is significantly better than the conventional linear regression model under the same data conditions, and has practical significance for guiding the part level accurate scheduling. Conclusion: By collecting a large amount of historical data through MES, constructing an analytic program, extracting the corresponding influencing factors, and building a three-layer artificial neural network based on the manufacturing big data, it is possible to accurately predict the part level operation time of NC drilling, thus laying a good data foundation for the optimization of equipment and process in the custom furniture manufacturing process, and promoting the practice of intelligent manufacturing in the field of custom furniture.

Key words: custom furniture, NC (numerical control) drilling, operation time, artificial neural network, furniture manufacturing, intelligent manufacturing