Predictive maintenance systems are no guarantee against major losses
• At the industrial insurance conference, AGCS Expert Days 2019, experts discuss the opportunities and risks associated with predictive maintenance.
• Analysis by Allianz Global Corporate & Specialty demonstrates that predictive maintenance systems contribute in preventing loss events at industrial facilities. Yet they do not offer a guarantee in avoiding the worst loss events.
• Outages and malfunctions as a result of misinterpreted data or inadequate data quality emerge as new risks.
As part of ongoing industrial networking (Industry 4.0), predictive maintenance is beginning to make inroads in many factories and industrial facilities. In an ideal world, this would allow faults to be predicted, and maintenance and repair work to be initiated before they result in outages. "Predictive maintenance will help to reduce the numerous instances of more minor physical loss or damage to machines. However, it is no guarantee against major losses and harbors its own risks," explains Hartmut Mai, Member of the Board of Management at Allianz Global Corporate & Specialty (AGCS) at the AGCS Expert Days, the industrial insurer's engineering conference in Munich.
'Predictive maintenance' is among the new forms of Industry 4.0 technology. According to the AGCS Trend Compass, predictive data analyses and automation are one of the most significant emerging technologies across all insurance sectors. Through the use of sensors, machines and facilities will be proactively maintained so as to keep downtimes to a minimum. The main difference to 'condition-based maintenance', in which maintenance is carried out on the basis of component condition, is the ability to predict the service life of machine parts by means of rule-based models, simulations and artificial neural networks. What's more, large quantities of data are recorded by sensors, which are then saved and analyzed. The predictive maintenance system is intended to identify and interpret anomalies in the machine data automatically. In an ideal world, this would allow faults to be predicted before they result in negative consequences or even outages.
As part of an in-depth analysis, the Allianz Center for Technology (AZT), run by AGCS, engaged itself with the issue of predictive maintenance – and thereby elicited the first tangible ramifications for opportunities and risks. Predictive maintenance systems can indeed contribute to changing technical risks and preventing loss events. "In specific terms, they could identify signs of vibration on a gradually growing crack on a shaft, for instance, and allow this to be rectified in a timely manner, thus averting a dangerous shaft breakage with substantial consequential losses," AZT engineer, Thomas Gellerman, offers as an example.
According to the Allianz Risk Barometer 2019, the risk and severity of business interruptions, the largest business risk for enterprises globally, can be reduced through predictive maintenance, if particular kinds of faults could be identified in time and a replacement is organized at an early stage, without impacting the availability of the facilities. "Above all, where manufacturers have not earmarked any scheduled reconditioning or reviews of their machines and facilities, and where the causes of outages and the modes of action are largely known, predictive maintenance can minimize disruptions and consequently save costs. Wind turbines are an excellent example of this," Gellermann explains.
The investigation by AZT also demonstrates, however, that some risks remain even with the deployment of predictive maintenance – and that new risks can in fact arise. Despite new forms of technology, potential spontaneous events, above all, cannot be eliminated if there are no measurable effects to be identified in advance. The burst of the low pressure turbine shaft at the Irsching power plant on New Year's Eve 1987 in Germany, which still today remains one of the greatest metal ruptures globally, would not have been averted with predictive maintenance, since the component which triggered the incident was not monitored. "Even maximal loss events cannot be excluded on the back of modern maintenance methods," Gellerman explains. Indeed, even if this method was extended by the use of the new maintenance form, review cycles, wear-and-tear and faults to unmonitored machine parts could not be identified in advance.
A further risk posed by predictive maintenance systems is the quality of the data produced. This data is likewise susceptible to outages and malfunctioning, and could be manipulated through cyber-attacks or acts of sabotage. Data could be misinterpreted or may initiate potentially damaging control commands.
As one of the leading global industrial insurers, AGCS wants to get involved with the introduction of predictive maintenance methods. AGCS is convinced that this new technology-based form of maintenance will sooner or later gain currency in businesses. Due to fault mechanisms and the complexity of machine facilities, future maintenance cannot be carried out on the basis of the predictive model alone. Instead maintenance measures, scheduled at set time intervals, will continue to be of necessity.
At the same time, however, AGCS wants to raise awareness about the new risks. "The industry and insurers must engage intensively with the opportunities and risks which maintenance technologies present – and cooperate closely with technical providers," says Thomas Meschede, Head of Allianz Risk Consulting in Central and Eastern Europe. What's more, it would be wrong to make a blank assessment of the risk under the guise of predictive maintenance; instead it necessitates an optimization objective and professional survey by an expert, each tailored to the specific use case. What is important is that critical and time-proven principles which serve to protect the facilities are not simply thrown overboard, such as differentiating predictive maintenance systems from the facilities' established protective functions.”
