In my post on the 2013 SME Annual Meeting, I promised to explain, in further detail, the work we presented at the conference on the effect of truck bunching on discrete event simulation (DES) analysis. The goal of this post is to expand on what I wrote in the previous post on this work.
As I explained in my series on Monte Carlo simulation, the main inputs of Monte Carlo (and DES) models are the statistical distributions that describe the input variables. In truck-shovel simulation analysis, this is usually done by collecting field data and doing statistical goodness-of-fit tests. Especially with today’s automated data acquisition (e.g. through dispatch systems), it is very difficult for a consultant to know when truck bunching occurs in the system (and is included in the data). Goodness-of-fit tests assume the data is independent and identically distributed (iid). This is not true when there is truck bunching.
The goal of this research was to: (i) account for truck bunching in DES modeling using Arena®; (ii) develop a statistical test to detect bunching in field data (e.g. VIMS or dispatch data); and (iii) use a simulation model to determine the effect of bunching on simulation results and conclusions. As you can see from the video below, we successfully accounted for bunching in our model. We also show that by simply running correlation tests between successive cycle times, it is possible to determine dependency in cycle times due to a slow truck(s).
We then conducted simulation experiments with our model to evaluate the effect of bunching on the simulation results. The main conclusions are:
- Ignoring bunching in truck-shovel modeling results in over-estimating productivity;
- Ignoring bunching results in over-estimating the uncertainty surrounding productivity estimates; and
- The two results above are exacerbated by pronounced bunching (i.e. the more bunching you have in the real system the more over-estimation of productivity and uncertainty you get by ignoring bunching).
The first conclusion is not necessarily surprising. But the second is significant because, as I have explained before in another post, uncertainty characterization is one of the main advantages of DES analysis over conventional fleet modeling software. The full presentation, as presented by Ms. Angelina Anani, a research assistant in the Sustainable Mining Modeling Group is below.
The work may be too technical for most mining engineers. However, the main take-away for those who are not trained in DES is to use consultants who know what they are doing. If you believe DES can provide value to your operation, talk to a reputable consultant like yours truly and Sphinx Mining Systems LLC.