Data Quality : Effect on Interpretation of Simulation Model Results

Digital Foundations


Mélanie, Kahle

The capacity of mines, mineral processing and metallurgical facilities and logistical chains can be assessed using discrete-event simulation models. The primary parameter establishing confidence in model results is the confidence interval, which reflects the variability of results with relation to variability in the inputs, random events, and dynamic interactions within the system. The quality of input data is one of the key factors which determine the quality of a given model.  Various data issues result in poor-quality data being a common issue in the mining industry.  There is a trade-off between the effort to put into data production, collection and processing, and the resulting improvement in quality of results, or confidence in results. To illustrate the impact of data quality on simulation results and their interpretation, a case study was conducted using a simple mine-crusher-mill system.  Three data sets of increasing quality (“typical”, “experience-based” and “data-based”) were defined and used to predict the system’s capacity through a simulation model.  Each data set provided a different estimate of the system’s capacity.  All data sets provided results that were directionally consistent when used to compare different stockpiling options. Data gaps for the specific system were identified, and sensitivity analysis performed to quantify the impact of inputs of lesser quality. Recognizing a model’s input data’s accuracy and scalability informs interpretation of results and planning of future simulation work.  While lower quality data may be acceptable for certain purposes, in other cases a higher confidence in results is required.  Sensitivity analysis determines the potential impact of changes in input data and helps determine which data gaps should be addressed.  In the context of capital projects in the mining industry, it is suggested that data production and collection activities should be integrated in the project’s life cycle and schedule, such that predictions with an appropriate level of confidence be available on time to make decisions.
Keywords: CIMBC22