Open Cut » Overburden Removal
This is a report of the outcomes of an ACARP sponsored research project that investigates data analysis techniques suitable for analysing the large volumes of data being generated by equipment performance monitors that have become widespread within the mining industry. The monitors are often sold on the premise of improving productivity but the monitor does not directly improve productivity, instead it is people that utilise the data who improve the equipment's productivity. This is achieved by reacting to information gained from analysing the data. The objective of this project is to provide the necessary tools to enable this to be achieved. It focuses on revealing hidden information within the data.
The project addresses three broad analysis categories and in each one, subtleties that would be missed by conventional analysis techniques.
- Summary statistics - includes such things as mean, mode and standard deviation. Most analyses will ultimately use summary statistics since they provide a means to compare one result with another. The focus here is to make these values more precise and meaningful with appropriate use of filtering and other techniques to sub-sample the data.
- Quantifying differences between equipment or practices - for example, comparing buckets, dragline rigging configurations, trucks, etc.
- Identifying when some aspect of the system changes - this differs from item 2 (above) in that the change is not deliberate or planned and identifying it quickly could prevent a costly period of substandard productivity. For example, a reduction in a dragline's swing motor performance could potentially be picked up directly from the data as soon as it occurs.
A major deliverable of the project is to produce a software package that utilises cluster analysis to perform the detailed comparison of equipment changes. JKCluster Plus facilitates this in three easy steps.
- Data Acquisition - connect directly to the Tritronics dragline monitor database to retrieve data or alternatively, import it via the Windows clipboard or text file. Cluster Analysis - novice users only need to select the range of clusters to be fitted to the data while advanced users can experiment with the subtleties of various parameters.
- Cluster Comparison - an important step in the process is to identify clusters from the two data sets being compared, which represent similar activities; effectively sub-sampling the data. A valid comparison can then be made between these pairs of clusters.
JKCluster Plus has been designed to be modular so that any future analysis techniques can be incorporated and exploit the components already present. It is planned that all of the analysis techniques investigated in this project will ultimately be part of JKCluster Plus.
It is recognised that identifying when a change occurs to a system could eliminate periods of costly underperformance. The objective is to achieve this directly from the data being collected by the equipment performance monitors, (eg. Tritronics TMS9000 dragline monitor) rather than having to retrofit additional sensors and equipment.
The report details several techniques that are suitable for analysing data from mining equipment performance monitors but these are by no means the only ones applicable. The information that is sought from the analysis will have a big influence on the type of analysis that is ultimately done. The following points can be used as a guide to selecting an appropriate analysis techniques.
What information is sought?
- Equipment operation summary - Summary Statistics, Graphs
- Identify trends - Graphs, CuSum, SPC
- Compare two "systems" - Cluster Analysis, Summary Statistics, Graphs
- Identify when a change occurs - CuSum, SPC
- Determine what has caused a change - Cause & Effect Relationships
- Is a "change" simply a random event or is it real? - Significance Test, SPC
Arguably the most important lessons to emerge from this research is that while these advanced analysis techniques are able to reveal details that would otherwise go undetected, they require people to interpret and react to the results - it is the tool not the solution. Just as any piece of mining equipment must by adequately resourced and managed, the same applies for analysing the data produced by the equipment.