Running machinery until it broke down used to be the norm but it was always a debatable course of action. It inevitably results in unscheduled downtime, lost production, poor quality, consequent essential rework and increased costs. Competitive businesses use preventive and predictive maintenance, sometimes arranged months in advance, to avoid unplanned stoppages, breakdowns and disruptions.
While predictive maintenance is better than the alternative, it often relies on manually-collected data, which ‘looks backwards’. Information is collected at the end of a shift, at the end of a day or at the end of the week and then collated. There is a vulnerability here: it is recognized as being important but if something urgent arises elsewhere in the plant, then the important item may be put to one side for the moment while the urgent matter is dealt with. The temptation is to overlook that particular report. The net result is that the ongoing condition of the machine is not being effectively monitored. It may well be OK, but it may not.
And even if everything is apparently ‘OK’, it is not properly optimized or performing as it truly could.
BCM Analytical Services, which analyzes and tests health and beauty products, used OEE (operational equipment effectiveness) techniques to reduce lost lab time by 40 percent and compress testing lead time from 15 weeks to under a month. The Moreton factory of Premier Foods, one of the UK’s leading food manufacturers, was able to reduce waste reduction by £200,000 a year; improve labor variance by over £800,000, and improve OEE by 41 percent within two months.
OEE, often called the gold standard of measurement, identifies the percentage of manufacturing time that is truly productive and provides essential insights on how to systematically improve processes. Performance measurement is based on scheduled hours and is compared with both expectations and specification.
It is unlikely that any process can run at 100 percent OEE. A more realistic, but still ambitious, target is around 85 percent.
But what if we could do even better?
To improve performance, we need data. Sensors gather information about operating performance and machinery condition, including wear. The collection of this data over a period of time enables operators and maintenance departments to forecast, accurately, when components will need to be replaced.
It can help understanding of why a bearing can wear out after six months, and why it does on one machine and not another. Why is an electrical component replaced when it still looks to be performing well?
To get the most accurate picture of how equipment is actually performing, data must be collected and analyzed in real time. Something like a short-term power surge may otherwise be missed and, along with it, the true reason why a component fails; long-time operators may expect failure ‘around that time’ from their experience and from the manual collection of data but they may not know exactly why it does.
While the accrued knowledge and experience of operatives and engineers is very valuable, it is not always reliable. Different people may have different experiences. Their knowledge should be collected but dispassionate collection of real data from the equipment itself is, ultimately, the most dependable basis on which to build.
The equipment to record data in real time is already there in the form of temperature sensors, vibration sensors, rev counters, programmable logic controllers and so on - all the monitoring equipment is already in place. Just as the Internet connected personal computers all over the world, the Internet of Things (IoT) connects machinery and equipment, brings them together and makes the whole greater than the sum of their parts. It enables real-time data collection, and analytics make it all useful.
Consider laser cutting and sorting
A company I worked with designed a 14-axis Cartesian (gantry) robot to undertake the bottleneck functions that had previously been performed by human personnel such as picking up steel plate, placing it with high precision on the cutting table and then removing the parts and stacking them on a pallet. The robot now does all that. The program on the robot calls to the warehouse for the raw material it needs for upcoming jobs and decides the best tool to use.
The robots are loaded up with sensors to measure all the relevant parameters and determine the best way to do any particular job.
What the company did was collect all this data, from all its machines across the world. It enabled it to understand how the robots were performing, how they accelerate, the torque being exerted, the speed of operation, and so on.
That data enabled performance to be optimized. Values outside of the normal range trigger alerts that something is wrong and enable intervention. More than that, the data enables better understanding of what the problem may be, quickly. This allows better planning of maintenance – and it’s very useful in the warranty period, particularly.
This process calculates MTBF (mean time between failures) much more accurately in the field and identifies more precisely where and why failures are occurring.
Getting a clearer overall picture means that spare parts warehouse locations across the world, along with their stocking levels and profiles, can be much better planned. That data is valuable to manufacturers and to their customers. It helps cut costs and improve performance, and becomes part of the total offer.
Keep the home fires (or boilers) burning
ELM Leblanc makes, installs, services and supports domestic boilers. They have implemented a remote monitoring solution developed by Orange and using Microsoft Azure, to collect and analyze a range of inputs, including temperature, electricity usage, energy outputs and water consumption. The system alerts technicians if there is a malfunction, which enables more efficient maintenance intervention and cost reduction, with fine-tuned cause diagnosis based on historical maintenance data. Possible causes are analyzed and corrective measures determined.
The response process includes feedback to help improve future alert diagnosis. The operator can also access the database that contains data on all ELM LeBlanc’s boilers, wherever they are installed. Every event added to the database will enable faster identification of patterns of failure or out-of-specification performance, thereby building a robust predictive maintenance solution. Ultimately, the machine learning algorithms will be able to calculate a reliable “remaining useful life” of the boilers.
Out of the darkness
The old way of doing things – reacting individually to customer complaints or failures and line hold-ups after a breakdown had already happened – was like crawling through a dark tunnel. No matter how fast or how efficient support services were, failure patterns could take months to be identified and, meanwhile, lines would be stopped, costs would be accumulating and dissatisfied customers – whether B2B or B2C - would be inclined to take their business elsewhere. Using IoT to manage predictive maintenance is more responsive, more effective, improves customer satisfaction and saves money.
To read more about how Orange Business is helping industrial enterprises leverage the power of IoT, please visit our manufacturing solutions website.
Before joining Orange Business, Pier Giuseppe Dal Farra served as Chief Operating Officer of ASTES4 SA, a Swiss-based tech company specializing in mechanical design and automation software. He was also cofounder of IPenable Inc., a company which pioneered the deployment of IoT in the Energy Vertical Sector. Today he is an IoT Industry Business Expert at Orange, focusing on all things IoT.