Predictive maintenance could save the industrial world billions annually, but this emerging field poses technological challenges that must be overcome first.
Most industrial equipment was not designed to be monitored. Although the world’s factories account for 16 percent of global GDP and 14 percent of employment, they are filled with production lines seemingly designed to complicate maintenance and defy modern technology that can detect and prevent breakdowns.
The value of solving this maintenance mess is tremendous. The McKinsey Global Institute estimates that predictive maintenance, powered by the “Internet of Things” (IoT), could have a $240 to $627 billion dollar annual impact on factories by 2025. Indeed, they could reduce annual costs by 10 to 40 percent, depending on the actions that innovators and factories take over the next 10 years.
Here’s the challenge: how do we collect, aggregate and analyze data from production lines that were never intended to provide any data? At minimum, we have to instrument machines with sensors, give those sensors a way to communicate and use software to make sense of the data while it’s still actionable. My goal here is to discuss each of these sub-challenges and provide some direction that will help the industrial world gain the full economic impact of predictive maintenance.
The Need for Ruggedized Sensors
Factories are some of the most hostile environments in the world. Heat, dust, grease, pressure, corrosive chemicals and powerful machinery can overwhelm all but the most ruggedized sensors. Of course, the more ruggedized a sensor is, the higher the price tag. Moreover, a large manufacturing operation could require tens of thousands of sensors to monitor for all potential breakdowns.
Prices on ruggedized sensors can range from hundreds to thousands of dollars, but one way to put the cost in perspective is to look at the price of a breakdown. In the auto industry, for instance, Nielsen Research found that downtime costs an average of $22,000 per minute (Disclosure: the study was commissioned by ATS). One extended breakdown could easily cost more than all the sensors needed to prevent it, but the upfront cost of sensors is a barrier that no manufacturer can circumvent.
There not yet a collection communication standards for digital devices in this space adding a great deal of confusion and complexity to a predictive maintenance strategy. Things in the “Internet of Things” have to communicate.
The major tech companies are pushing different protocols, and as Network World illustrated in detail, it’s a messy situation. Google, for instance, released an IoT standard called Thread. Samsung Electronics, Silicon Labs, Yale Security and a few other companies support it. Qualcomm, Cisco, Microsoft and LG support the AllJoyn protocol. Atmel, Dell, Broadcom and Samsung (batting for multiple teams) support the Open Interconnect Consortium (OIC) standard. Meanwhile, Intel, Cisco, AT&T, GE and IBM formed the Industrial Internet Consortium.
Analysts have an endless stream of opinions on why each consortium is promising or not. Meanwhile, some factories are hesitant implement Wi-Fi in the first place because an industrial-grade wireless internet can cost hundreds of thousands of dollars. Until there is an agreed upon protocol, industrial IoT will be limited; manufacturers will be hesitant throw all their eggs in one consortium.
Some factories rely on industrial equipment designed before the computer era. Others rely on newer machinery that does produce monitoring data, but not necessarily data that will play friendly with predictive maintenance software.
Some machines allow a field worker plug-in with a computer and collect whatever maintenance data the manufacturer deemed important. However, there is no simple way to convert this data into a format that monitoring software will understand, nor there is an efficient way transmit this data to a central database where it can be analyzed. The data might not even be useful. In most cases, it provides information about that one machine without any context on the performance of the production line.
Thus, even if a machine does produce data, factories will need to retrofit the equipment with modern IoT sensors to get actionable data. While this might disappoint manufacturers that expected to benefit from self-monitoring equipment, it may be the only practical option.
The Maintenance Challenge in Perspective
Manufacturers could wait to see how industrial IoT evolves. Maybe better, cheaper sensors will debut? Maybe the tech companies will put egos aside and agree to one protocol so all their devices and software can talk? The consequence of waiting will be further breakdowns, hefty maintenance costs and expensive replacements when equipment is pushed beyond repair.
There are financial barriers to predictive maintenance and tradeoffs to consider. However, if predictive maintenance can start to produce the trend towards the predicted 40 percent savings by 2025, the risks are worth the reward.
Chris LeBeau is Global IT Director at Advanced Technology Services (ATS).