Bauxite Grinding Ore Grading Process Hierarchical Data Correction Method

The raw measurements of bauxite grinding-classification plants directly affect mineral processing engineers to determine the condition accurately because of the measurement errors and dis-integrity, which leads to optimal control errors and decision-making mistakes in operation. Therefore, according to the characteristics of the bauxite grinding-classification process to study the process data rectification has important theoretical and practical significance, which will provide accurate, stable and integrated measurements for the optimal control and management of the process.Based on the characteristics of bauxite grinding-classification process data and constraints, a multi-layer data reconciliation model is proposed to avoid low efficiency of the steady-state data reconciliation model in dealing with high dimensional and strong nonlinear problems. The multi-layer model consists of global material balance layer, particle size distribution/grade layer and chemical compositions of different particle size layer (the metal distribution ratio layer). The particle swarm optimization (PSO) is then introduced to solve the model from the bottom to the top layer, which improves the efficiency of the reconciliation process.Normal test shows that the raw measurement data of grinding-classification process do not strictly follow a normal distribution by the impact of gross error. Therefore, based on iterative estimation a hierarchical test method is proposed for the gross error detection of raw measurements. The test method makes full use of the multi-layer reconciliation results to construct mathematical statistics hierarchically, and substitute reconciliation results for the raw measurements, which overcomes the defects of inaccurate data statistics and false alarm of classical test methods, and achieves gross error detection and identification.Finally application of the multi-layer reconciliation model and hierarchical test method for gross error detection to the actual data of a bauxite grinding-classification process is discussed. The accuracy and stability of the application results show that the process data after rectification is satisfied with the demands of mass balance, and provide a accurate and stable process data for optimal control, decision-making and management of the bauxite grinding-classification process. Moreover, more information can be obtained during the reconciliation process.