Surface mine design using intelligent computer techniques
Surface mine planning involves the results of algorithmic numerical calculations being used by engineers to make informed decisions relating to the design. The Department of Mining Engineering at the Unversity of Nottingham has in the past been involved in developing modular algorithmic packages. The emphasis of the computer research has now altered. Smaller specialised systems are now being developed to cover individual aspects of the design process. Artificial intelligence techniques are being introduced into the mining environment to solve the planning problems often associated with the large amounts of uncertain information needed by the engineer. This thesis is concerned with the development of MINDER, a decision support system capable of assisting the mine planner in the complex task of optimum surface mining equipment selection. An expert system shell has been used to create a series of individual application modules, each containing a multi-level knowledge base structure. An information handling system has been developed which is capable of storing consultation information and transfering it between knowledge bases and between application modules. Once an effective method of information handling had been achieved the flow of control between the system knowledge bases was rapid and followed complex inferencing routes. Most of the commercially available packages mathematically model a deposit, calculate volumes and simulate operations. One of the aims of the MINDER system was to integrate with other software, for example MINDER is capable of reading volumetric and material information from Surpac mine planning software. Geological data and manufacturer’s equipment specifications are stored in DbaseIV databases. The expert system is capable of writing macros based on the consultation and performing complex relation operations involved in the elimination and ranking of equipment. In a similar manner macros are written to control the simulation package GPSS which used to simulate operations using the selected equipment. A range of ‘in-house’ Pascal software is used for numerical calculations and matrix manipulation, an example of this is the fuzzy logic software used to handle uncertain information. Another aspect of the project is an investigation into the use of machine learning techniques in the field of equipment selection. Knowledge induction software has been used to induce new rules and check those produced in the MINDER system. Various experiments have been carried out using neural network software to produce equipment selection models. Training data taken from the mining industry was used on both these systems and the results were tested against MINDER consultation results.