Mine Digital Twin For Value Chain Optimization

点击翻译按钮获取中文摘要

The overall goal of this program is to develop a digital twin platform that will bring together data across the mining value chain and enable data-driven decision making. The flow of information and the decision-making along the chain of mining typically occurs in a discontinuous fashion over long time spans. Also, due to the uncertain nature of the knowledge about the deposit and its inherently heterogeneous spatial distribution of material characteristics, actual mine production performance (produced ore grades and quantity) and extraction process efficiency often deviate from expectations. Embedding digital technologies and real-time data capture and analytics into mining operations can create an opportunity for transformation across the mining value chain. A digital model of the physical environment, constructed using geological, engineering, and asset information, can be continuously updated with data from sensors and location-aware systems. Recent advances in automation and data analytics lead us to believe that this is possible and this research program aims to address this. Motivated by advances in machine learning, the objective is to develop a suite of algorithms for data assimilation and analysis, which will enable integration of diverse types of data from all phases of a mine lifecycle. The proposed research program will be developed around two interrelated themes: 1) Quantification of contextual relationships for available geological, mining, and processing data across multiple sources and various data types and structures; and 2) Development of a mine digital twin; an integrated framework that manages data, performs signature analysis to extract the salient features and uncertainty from the raw data, finds correspondences across datasets, and aids the user in visualizing extreme-scale digital mine data. This integrated data assimilation-machine learning approach will allow for uneven spatial and temporal data distributions and redundancies to be addressed so that models can ingest massive, heterogeneous, and potentially incomplete data sets. Newly developed digital twin will be trained to analyze historical production data and discover what operational parameters have the most significant impact on the operational performance. This will allow us to simulate the future by experimenting with `levers'. It will enable us to pull `levers' in a virtual resource-constrained environment to simulate possible future scenarios and define optimal production strategies. The impact of this program would be mostly methodological, allowing others to build on our findings and tackle previously unreachable problems. Also, the program will provide extensive training in digital technologies and data science for three Ph.D. students and one PDF. Students trained in this program will be positioned to capitalize upon new ideas and technologies learned through this project and be able to use them for the benefit of Canada and its mining industry.