In this paper, the motion characteristics of a cam-linkage modulator, a cam unit and a motion transmission unit were analyzed in order to establish a quasi static model, explored their motion characteristics, and identified the factors affecting the stability and reliability of the heald frame motion. Based on the mapping points, the conjugate cam profile was established using non-uniform rational B-splines and a mathematical model of the motion process from the uniform rotation motion input of the dobby to the vertical variable output of the heald frame was constructed. Additionally, an analytical method for each motion transmission process was established and a numerical programming method was developed. The displacement, velocity and acceleration of the modulator, cam unit and heald frame were simulated and measured using a motions simulation and test bench, and the analysis, simulation and verification results were compared. The results show that the motion characteristics of the heald frame can be obtained through the cam-linkage modulator, cam unit and the motion transmission unit. The accuracy of the constructed model was verified which lays the foundation for further exploration of more stable, reliable and high-speed dobby mechanisms that will meet the requirements of a new generation of looms.

In this paper, we propose a convolutional autoencoder with a new structure for unsupervised learning when the purity of the training data is not guaranteed. This autoencoder has two unique features: the target area is reconstructed from the surrounding areas and the L2 loss is predicted simultaneously. The superiority of this model was verified using SEM images of defective nanofibrous materials by calculating the AUC value. The results of our experiments with the training data contaminated by defective data show that the former feature improves the robustness against contamination of the training data and the latter improves the accuracy. Although this approach did not achieve the highest accuracy, it could reduce the cost of annotation for practical use. Furthermore, we applied our method to images of NISHIJIN textiles and found that it worked well for some types of textiles.