This paper attempts to solve the insufficient machining precision of computer numerically controlled (CNC) machine tools, which is induced by the thermal error of the spindle. Firstly, the relationship between machining error and thermal sensitive points was analyzed through experiments. On this basis, the backpropagation neural network (BPNN) was improved by particle swarm optimization (PSO). Next, the improved network (PSO-BPNN) was used to build a thermal error compensation (TCE) model for the spindle of machine tools. Taking VM-500T precision machine tool as the object, the temperature data were grouped through the optimization based on thermal imaging, grey relational analysis (GRA), and fuzzy clustering, to determine the temperature sensitive items that causes the thermal error. To speed up network convergence, the PSO algorithm was introduced to optimize the number of hidden layers and the number of hidden layer nodes of the BPNN, lifting the network from the local optimum trap. To enhance the generalization ability, the weights and thresholds of the BPNN were also improved by the PSO. After that, two TCE models were established for the spindle of the machine tool, respectively based on the original BPNN and PSO-BPNN. Contrastive experiments show that the PSO-BPNN TCE model achieved the better generalization ability, and improved the prediction accuracy of the machining error of the CNC machine tool.
Thermal error compensation of a 5-axis machine tool using indigenous temperature sensors and CNC integrated Python code validated with a machined test piece
