Abstract: A testing-and-evaluating system based on BP-neural network is designed to deal with the difficulty of evaluating sewing performance of sewing machine. A neural network with three levels composed of four sub-networks which are used to evaluate the four sewing performance indexes respectively is modeled between mechanical properties and sewing performance, and the Quantization relationship among sewing performance, sewability performance and sewing quality is researched. The sample collecting method for BP-learn is presented and a set of application software base on VB and Matlab and testing system based on USB-DAQ card are designed, then experiments are performed on the designed system. Testing results show that BP-neural network achieves satisfactory convergence performance, and the evaluating results correspond with the actual fact.

Key words: sewing performance, testing-and-evaluating system, BP-neural network, sewing machine