Research And Development Of The Software User Interface For Production Index Prediction In Mineral Processing

Mineral processing with the characteristics of nonlinearity, multi-variables, time-varying, large lagging, strong coupling etc. will make the engineers out of the real-time adjustment. So researchers begin to do research on the prediction algorithm. However, the algorithm cannot be used in the real industrial process until it has been simulated. Besides, there are few researches on the production index prediction software in the world. Some researchers may develop the software according to their demands, but the human-computer interface usually cannot meet other researchers’needs. In order to provide prediction software with more powerful functions, better human-computer interface performance, and wider applicability, this paper begins to work. It is supported in part by the National Basic Research Program (973) of Program of China under Grant 2009CB320604. The software is shown to be useful, efficient, and provides a good human-computer interface from the experiments in a real production process. In general, the details of this paper can be listed as follows:(1) First of all, we concern about the particularity of the mineral process and researchers’ requirements. Then, we give a detailed demand analysis of the software function and its human-computer interface. After that, we begin to design the human-computer interface of each modules in the software, and the modules can be listed as index data management and analysis, prediction algorithm configuration and invocation, prediction result analysis and evaluation. Of cause, we must observe the rules of human-computer interface during the designing.(2) This paper puts forward an adaptive parameter online SVR (Support Vector Regression) method to predict the mineral production index in order to give a better verification of the software’s functions and human-computer interactive. This method not only can train the model online, but also predict the production index online. What’s more, the SVR algorithm based on the adaptive parameters will achieve more accurate forecast precision with the samples’ own attributes’ adding to the training model. At last, this method is developed as a separate module for the researchers to use directly.(3) In order to achieve each module’s function in this prediction software, we used several modern advanced programming techniques. The techniques can be listed as follows:Using the MVVM (Model View View-Model) pattern to design the software architecture, so we can separate the human-computer interface from the code in background. Using the subject-publish pattern to design the communication among different software models, so the models can be decoupled easily. Using the adapter pattern to develop the function of interface’s auto-matching when configuring the algorithm parameters. Using the command pattern to package each algorithm’s invoking command. And using the XML (eXtensible Markup Language) technique to package algorithm’s properties.(4) At last, this paper tests the performance of the software and the performance of the HCI (Human-Computer Interaction) interface what proved that the software has a high maintainability and extensibility performance, the HCI has a high usability and friendly performance. Then, some validations of software’s functions and HCI interface are given. After the validation, we found that:The model of index data management and analysis can provide a function of data collection, statistical analysis, and data pre-processing. The model of prediction algorithm configuration and invocation can provide a function of interface’s auto-matching when configuring the algorithm parameters and algorithm’s invoking command. The model of prediction result analysis and evaluation can provide a function of computing and saving performance index. This paper also gives a fully experiment of the adaptive parameter online SVR algorithm, and it proved that it has a high prediction accuracy.