Willem Sprong, technical executive, electrical engineering for GIBB Engineering and Architecture, writes that the maintenance of infrastructure is a vital role for engineers worldwide – to mitigate risks for industry and for society at large.
With recent advances in information technology (IT), possibilities have opened up to ensure that this maintenance and risk management happens effectively. It is in the interest of all of us, that these possibilities be explored.
Generally, the maintenance of key infrastructure has taken a preventative approach. A team would follow a schedule, and every month they would go out into the field to do maintenance.
Another common type is reactive maintenance, where one waits until something breaks, and then repairs it. Condition-based maintenance, is where one measures the condition of a particular part, then decides whether maintenance is required.
These legacy methods of maintenance are not an optimal use of resources. Parts need to be kept in inventory, and skilled staff kept on call.
The technology currently exists to follow a more scientific approach – predictive maintenance, which fuses engineering and IT for optimum awareness, timing and resourcing. Predictive maintenance compares the trend of measured physical parameters against known engineering limits to detect, analyse and correct problems before failures occur.
I have been fortunate to be involved in a predictive maintenance case study using railway traction substations. Here, we began by identifying the vital components in the substation. Then we determined what parameters could be easily monitored and measured, and began comparing them.
The case study compared the oil temperature inside a transformer with the current drawn through it. The relationship showed a gradual rise in temperature as operation continued. Where there was suddenly a significant change in that ratio, we could predict a possible failure and plan maintenance accordingly.
Fundamental to predictive maintenance is that conditions be monitored constantly. Monitoring looks at the trend of all previous measurements to gauge whether maintenance is required.
IT is involved not so much in the monitoring, but with the processing of the large amounts of data being generated. Once sensors have been installed, one collects data. This information must then be transmitted to a central point, and stored in a database.
Different sets of data can be compared and reports issued. These reports might just be a dashboard interface, or a more detailed database report. Once the report is generated, maintenance decisions can be taken and acted on.
In the past, condition-based maintenance required switching off a substation in order to do measurements, which rather undermined the goal of maintenance: to keep the system running.
Similar benefits could be realised for a water pump station or a railway line, for use in municipalities or water utilities.
Predictive maintenance has found application worldwide, including in automotive, rail and manufacturing sectors. As it gains currency, there will be a need for more partnerships between engineers and other professionals. We will need hardware experts to develop the microcontroller units and sensors to collect data. Communication experts will be needed to transfer data to offsite servers. Programmers must write the code to analyse this information and to set up dashboards.
Industry looking for a more efficient maintenance model should note that predictive maintenance technology is already available – it need only be combined in an innovative new system. We already have vibration sensors and smart technology on our cellphones to identify patterns.
This technology can integrate a machine learning, artificial intelligence approach component, so that the longer these systems are in place, the more effectively they would work.
Challenges we face in implementing such an approach are the perceived complexity of the systems, and the perceived cost. But in reality, costs are plunging. Where equipment to measure vibration cost R50 000 eight years ago, something similar now costs barely R1 000.
Predictive maintenance would require creating new teams with different expertise, which does increase the complexity of the system. However, the benefits outweigh the drawbacks.
An always-on monitoring approach, with easy digital reporting and a predictable maintenance schedule would mean a more effective use of resources for all stakeholders – benefiting the organisation’s bottom line, and society’s.