In the past, fleet failures have presented a significant risk to global mining projects. With the implementation of telematics, however, these problems have not only become manageable, but largely avoidable.
Recently, an Indian company suggested telematics could be enhanced with the use of fault codes — or diagnostic troubleshooting codes — to help monitor the health and performance of on-site vehicles in order to improve efficiencies and reduce downtime through planned maintenance.
At its heart, telematics is a communication system that is based on information flowing to, and generated from, vehicles via wireless networks.
It involves data management, and in-fleet computing, to allow miners to exchange and convey information while providing drivers and passengers with personalised information services.
Telematics gather details via global positioning system (GPS) technology to help fleet operators manage their resources.
The GPS provides live visibility into location, speed and movement within points of interest through geo-fencing.
Meanwhile, sensors enable the capture of data on driver activity — including aggressive acceleration, harsh braking and erratic cornering.
They can also be used to monitor any in-vehicle activity such as a door opening, tail lift raising and trailer temperature.
The sensors extract information directly from the vehicle and are primarily used to gather real-time data on fuel efficiency and odometer readings to help power digital fleet maintenance.
This data enables the fleet management software to provide easy-to-understand visualisations that help miners optimise their operations.
According to research by Cummings Incorporated, one traditional method used by service representatives to address the fault codes was to follow “recommended troubleshooting trees”.
With the electronic engine and subsystem performance interdependencies, however, this was becoming a challenge.
Furthermore, the grouping of the fault codes was unknown if they were related to a specific cause.
“We implemented unsupervised machine learning and data mining techniques to address such issues,” the company said.
“First, we enforced the co-occurrence theory that helps us understand the fault codes that occur together and exhibit dependencies and relations.
“Second, we implemented clustering algorithms to known groups and categories of fault codes based on their functional states.
“These studies provide insights into the failures of the components and their conditions. (They) also help in resolving the problems experienced by the engines and subsystems.
“Moreover, these methods address the issues in the early stage and help technicians improve uptime (via early repairs and diagnostics).”
In Cummins’ analysis, the clustering method was used to find the group of fault codes with similar behaviour, while multi-label classification was utilised to both predict the codes for a given record and indicate the sets of co-occurring ones for the given form.
These proposed approaches, the company said, were helpful for fault code diagnostics in the telematics data.
The clustering approach also assisted in identifying the group of fault codes, which in turn allowed the domain experts to identify and name the clusters and compare one occurrence of fault codes with the other.
This methodology also looked at the common parameter values belonging to a specific cluster.
Thus, an in-depth analysis of fault codes and the associated parameters was possible by grouping them and scrutinising the chain of occurrences.
Furthermore, a supervised multi-label machine learning approach helped in identifying troubleshooting codes.
It also assisted miners to understand fault code behaviour, their dependencies and the significance of co-occurrence.
“Additionally, identification of implicit and functional relationships between co-occurring fault codes can be achieved,” Cummins said.
“For the given set of telematics parameters, the approach helps identify the fault codes that are likely to co-occur.”
One example, the company said, involved the creation of a troubleshooting tree whereby primary fault codes, such as possible engine speed sensor malfunctioning, were reviewed.
This involved checking whether the turbocharger oil seal was leaking oil into the air intake and, or, the exhaust system.
“Co-occurrences of fault codes help identify relationships within the data, illustrating not just which codes appear together, but how often they appear, providing a means of assessing the prominence of the combination,” Cummins said.
“Also, such an analysis helps subject matter experts to be informed of multiple possibilities that are inherently not apparent.”