基于改进随机森林算法的工业过程运行状态评价

运行状态评价是指在过程正常生产的前提下, 进一步判断生产过程运行状态的优劣. 针对复杂工业过程定量信息与定性信息共存的情况, 本文提出了一种基于随机森林的工业过程运行状态评价方法. 针对随机森林中决策树信息存在冗余的问题, 基于互信息将传统随机森林中的决策树进行分组, 并选出每组中最优的决策树组成新的随机森林. 同时为了强化评价精度高的决策树和弱化评价精度低的决策树对最终评价结果的影响, 使用加权投票机制取代传统众数投票方法, 最终构成一种基于互信息的加权随机森林算法(Mutual information weighted random forest, MIWRF). 对于在线评价, 本文通过计算在线数据处于各个等级的概率, 并且结合提出的在线评价策略, 判定当前样本运行状态等级. 为了验证所提算法的有效性, 将所提方法应用于湿法冶金浸出过程, 实验结果表明, 相对于传统随机森林算法, MIWRF 降低了模型的复杂度, 同时提高了运行状态评价精度.

Industrial Operation Performance Evaluation of Industrial Processes Based on Modified Random Forest

Operation performance evaluation refers to further judging the operation performance of process on the premise of normal production. In the view of coexistence of qualitative information and quantitation information during the industrial processes, a method of industrial operation performance evaluation of industrial processes based on modified random forest is proposed. In order to solve the problem of redundancy of decision trees information in random forest, decision trees are grouped based on mutual information, and the optimal decision tree in each group is selected to form a new random forest. Meanwhile, in order to strengthen the decision tree with high evaluation accuracy and weaken the decision tree with low evaluation accuracy, weighted voting mechanism are proposed to replace the traditional mode voting, and finally a mutual information weighted random forest (MIWRF) based on mutual information is formed. To verify the proposed method, the method is applied to hydrometallurgical leaching process. The result shows that MIWRF reduces the complexity of the model and improves the accuracy of operation performance evaluation compared with the traditional random forest algorithm.