Multi-channel Mechanical Signal Analytical Estimation and Optimal Combination for Mill Load Parameter Forecasting

The sensitivity of multi-source (multi-channel) mechanical signals of ball mill system at different working conditions is different. Moreover, the valuable information for constructing mill load parameters (MLP) in them is also diverse. Aim at these problems, an optimal combination method for MLP forecasting (MLPF) and an analytical estimation of multi-channel mechanical signal at multiple working conditions were proposed. First, the data-driven MLPF models under multi-condition based on multi-channel mechanism signals were constructed, whose structure parameters and validation errors are calculated based on validation dataset. Then, a new combined evaluation index was defined to analyze and measure the prediction contribution ratio and hidden information contribution ratio of the multi-channel mechanical signals. By using the combined evaluation index and pre-set threshold, the initial selection of the multi-channel signals was achieved. Finally, these selected multi-channel mechanical signals were optimally combined and the best combination was used to construct MLPF model. The simulation results based on multi-channel mechanical signals of a laboratory scale ball mill verified the effectiveness of the proposed method.