Once headlined as the ‘killer app’ for IIoT, predictive maintenance has taken a while to find its feet, but progress has in fact been sure and steady, with some standout examples of successful niches.
By Martin Keenan, Technical Director, Avnet Abacus
In 2016 predictive maintenance was touted as one of the key applications for IIoT. Even in 2018, analysts such as Gartner made strong predictions of future success. Gartner predicts that by 2022, spend on IoT-enabled predictive maintenance would increase to $12.9 billion, up from $3.4 billion in 2018, with improved operational efficiencies through predictive asset maintenance leading to substantial savings of up to 40%.
However, the market has not exploded quite yet. Indeed, a new survey of more than 600 high-tech executives from Bain found that industrial customers were less bullish about predictive maintenance in 2018 than they had been in 2016. This was mainly due, according to the analyst firm, to challenges in effectively gaining insight from IIoT data once gleaned, and at the other end of the equation, difficulties in implementing systems in the first place.
According to Bain, another key barrier could be summarized as: “Device makers and other vendors of industrial and operational technology need to dramatically improve their software capabilities—not a historical strength for most of them.” In spite of this, the analyst firm still predicted rapid growth for IIoT, with the market doubling in size to more than $200 billion by 2021.
Preventive vs predictive
One of the biggest challenges to predictive maintenance adoption has been the fact that many industry sectors are still working their way through the implementation of preventive maintenance systems. Arguably the forerunner of predictive, preventive maintenance systems can range from quite simplistic, such as a ‘traffic light’ health system for individual machines or plant elements, to far more complex networks of sensors feeding data back to centralized dashboards. However, it generally relies on manufacturer lifetime predictions, human operators or direct sensor data to highlight potential problems, rather than use complex algorithms to predict maintenance schedules.
This means that the benefits of preventive maintenance are becoming well-entrenched, but the staged adoption has left many industrial players waiting for the machine learning and AI market to mature further, easing adoption pains, and lowering costs.
Food for thought
The current situation has created a range of opportunities, such as in the food industry. One example is the Mitsubishi Electric Smart Condition Monitoring (SCM) system that slots neatly into the niche between “traffic light” preventive systems and full-fat predictive IIoT. The system monitors the condition of individual assets but layers these to provide a holistic picture of overall plant health. Local preventive systems still provide visual ‘health check’ indicators, but real-time data are transferred over Ethernet to a PLC for in-depth monitoring and cloud-based analysis. A teach function ‘learns’ the normal operating state of the machine, then vital information such as bearing defect detection, imbalance, misalignment, temperature measurement, lack of lubricant, cavitation detection, phase failure recognition and resonance frequency detection can be viewed in a cloud dashboard.
Improving transport efficiency
There are certainly clear indications that predictive maintenance is still front of mind in many sectors, such as the transport industry. One example is trackside maintenance, a significant operating cost for rail firms that also requires qualified personnel to operate around the clock in potentially dangerous conditions. However, by deploying IIoT sensors and analytics technologies rail operators can move from wasteful inspection cycles (where perfectly serviceable equipment is checked and rechecked irrespective of condition) towards preventive, conditions-based and predictive maintenance.
For example, Nokia created a rail asset lifecycle optimization application that brings all three elements together, not only modelling maintenance schedules for each asset based on learned operating parameters and incorporating external data such as weather conditions, but also building in crucial risk-related data around the consequences of a component failure.
Keeping track of renewables
Predictive maintenance technology originally designed for the mining industry has found an application in the renewables industry, in an interesting pivot. An Australian startup, Ping Services, developed an acoustic sensor that was intended for mining and drilling applications, able to monitor the acoustic signature of a drill bit over its lifetime, and then harness machine learning to predict fault development ahead of time. While reducing astronomically expensive drilling stoppages is clearly an area of considerable interest, the company embarked on pilot programs with Australian and US-based wind farms to monitor turbines with similar goals in mind.
The solar-powered, satellite-connected sensors actively listen to the turbine blades’ acoustic signature to detect the development of pitting or cracks caused by lightning strikes or hail. As such issues begin to develop, they can be monitored and targeted for maintenance remotely, rather than requiring highly-trained teams to tour windfarms and conduct routine testing.
Predictive comes of age
Overall, while predictive maintenance may have taken some time to mature, there are signs that the market is beginning to open up, especially in niche use cases. More generalized ‘plug-and-play’ systems targeting wider industry sectors are also beginning to emerge, highlighting that R&D investment is beginning to translate into real-world demand. It seems that predictions of demise have in this business case at least, been exaggerated.
Martin Keenan is the Technical Director at Avnet Abacus, which assists and informs design engineers in the latest technological advances. With the IoT and Industry 4.0 changing manufacturing, Avnet Abacus helps designers find the best technological fit for their industrial applications, and accelerates the process all the way from idea to market.