Cape Town, 21 May 2014
The terms predictive maintenance (PdM) and predictive asset management (PAM) are thrown about with feted enthusiasm nowadays. It is no secret that the marriage of predictive analytics and engineering systems is rife with potential.
As modern machinery has become ever more automated on a mass scale, the potential for intelligent everyday systems seems an imminent reality. Perhaps the most obvious, and economically enticing, first step in this evolution is intelligent performance and maintenance.
Traditionally, PdM has found its roots in industrial scale machinery. The type of process-critical machinery found in power plants and production lines, where the value of reducing downtime is noteworthy.
Live sensor measurements are compared against a self-learning statistical model of the machine to pick up irregularities and wear, which are forecast into the future. PAM is one level up and encompasses management of the entire machine life cycle.
While these applications make cut and dry business sense, one might be forgiven for thinking on a far grander scale. The inner workings of a modern car more closely resemble that of a robot, with upwards of 100 sensors, and, thanks to recent EPA (Environmental Protection Agency) requirements requiring on-board diagnostics, it is already possible to introduce some third party PdM tools.
The next step, it would seem, is to run integrated PAM software that determines smart maintenance schedules and picks up mechanical issues before they become a problem. At a manufacturer level, this abundant wealth of data would be the holy grail of prognosis tools for technicians in a rapidly evolving industry. The concept of integrated PAM need not be limited to the automotive industry and newer machinery however. Generic PAM modules can be retro-fitted onto older machinery.
The true potential lies in shifting the focus from specialised industrial machinery to applications that favour big data. Here, the automotive industry is an analytical goldmine. With enough data, the world is your oyster. Beyond these large scale industries and big data, however, lies a grey area.
At what point does it make sense to invest in PdM?
The value here lies in the ability to minimise and effectively schedule downtime. In order to understand the value of PdM for a medium or small scale setup, it is necessary to consider (1) the overall cost per unit of downtime, (2) the percentage downtime experienced and (3) the PdM project cost. One need also have a realistic expectation of the percentage reduction of downtime that PdM can achieve; a value that differs across industry and process. Lastly, with no historic store of maintenance and sensory data, a certain learning period is necessary for the PdM model that is directly dependent on the rate of data capture and machine wear.
Once set up, PdM is a low cost process that opens a number of doors as a powerful diagnostic tool for technicians and engineers. As such, it is critical to involve technicians and engineers in the design of the PdM process in order to realise its full potential. Ideally, an individual with significant experience in electronics, mechanics and predictive analytics should be on the project team to communicate effectively between these three integral facets.
Predictive maintenance is more than just preventative maintenance with foresight, it is a stepping stone to greater tools to come. As electro-mechanics melds with predictive analytics, driving toward smarter machinery, it is not only the immediate value of PdM that begs attention, but also the basis it creates for technologies to come.
Share