Dynamic belt tension inference method and related motor driven roller system

A dynamic belt tension inference method, comprising steps of:

• (A) performing a training process, comprising steps of:
• (A1) by a controller, using an Isolation Forest algorithm to train a tension inference model, wherein the tension inference model is configured to generate an anomaly score and a dynamic tension corresponding to the anomaly score;
• (A2) by the controller, using the Isolation Forest algorithm under same hyper-parameter set to perform multiple trainings to generate multiple tension inference models;
• (A3) by the controller, respectively computing multiple model performances of the multiple tension inference models according to the anomaly score and a pre-recorded data label;
• (A4) by the controller, computing an averaged model performance of the multiple model performances;
• (A5) by the controller, determining whether multiple averaged model performances have been acquired, wherein the multiple averaged model performances correspond to the multiple hyper-parameter sets;
• (A6) by the controller, selecting one of the multiple hyper-parameter sets that corresponds to an optimal averaged model performance as a final hyper-parameter set for training to output a final model, wherein the final model includes a framework and parameters;
• and
• (B) performing an inference process, comprising steps of:
• (B1) by the controller, using the final model to infer processed data to generate the anomaly score and the dynamic tension corresponding to the anomaly score, wherein the processed data is generated by preprocessing data of a motor that drives a belt to rotate, and the final model is used to infer the dynamic tension of the belt while the belt is rotating.