The application of advanced analytics into OEE
Recently I have been reflecting on my life and career and was amazed to discover that I just hit 30 years of working with industrial and manufacturing applications. While it’s a bit tough to admit my age, I’m quite happy with my career and have enjoyed manufacturing and industrial applications. I love to see how stuff is made and it has been a cornerstone of many economies. So, whatever I can do to help it, I will.
I’ve been to hundreds of plants and walked many production lines. There are three things that are quite universal that these plants care about and they are:
Safety is always number one. It’s never a good story when an operator gets his or her arm stamped into a car door. The company is on the hook for a lot of money and medical support and of course it’s not too good for the operator either! Safety has become a pervasive topic in manufacturing – as it should be – but I’ve been more focused on the latter two items on the list.
Manufacturing plants make the product but are not responsible for selling it. This means they focus heavily on costs. They do this by focusing on items over which they have direct control, which include quality and productivity. As manufacturing has matured over decades, many of their processes have become automated and they’re on a continual journey to digitize their operations through software technologies such as Supervisory Control and Data Acquisition (SCADA) and Manufacturing Execution System (MES). The data coming from these systems has helped to drive the popularity of one of the best-known and most frequently used key performance indicators (KPI) in manufacturing: Overall
Equipment Effectiveness (OEE).
OEE gained popularity because it directly measures manufacturing asset against quality and productivity measures. OEE itself is a combination of three underlying KPIs which are:
Availability: Is the asset available to produce product when it is scheduled to produce? This is mainly a measure of unplanned downtime.
Performance: When the asset is available and producing, is it producing the amount of product we expect? (for example, 10 widgets a minute)
Quality: Of the 10 widgets produced per minute, how many are good? (For example, 1 out 10 are scrap)
Thus, this one KPI captures quality and productivity at the asset level and gives manufacturers a way to measure their equipment. It’s no wonder that this KPI has become so popular in manufacturing. For a numerical example, say your equipment is available to produce 90% of the time, it is producing 90% of what it is capable of, and 1 in 10 widgets are bad (90% good). OEE is a product of all of these, and it would be (90% X 90% X 90%) 72.9%.
How can advanced analytics help?
Manufacturers have been focused on improving OEE for the equipment in their plants for decades. This is nothing new. It started with automating processes, improving performance rates, and increasing repeatability for quality. The advent of manufacturing software systems have allowed them to track OEE effectively and understand where to focus to gain improvements. For instance, if the equipment is down half the time, its OEE can never be above 50% even if it is producing 100% quality at standard rate.
Manufacturers have been looking at the process and equipment data, generating pareto charts, and implementing preventive maintenance programs to increase the availability of their assets.
The question becomes, then: With all of this investment, why does equipment still go down unexpectedly?
There are things we still don’t know about our equipment and processes. Enter in advanced analytics. We have tons of equipment data available with all the sensors we now have on our assets. There is additional information to be extracted from this data. We have gleaned all the obvious information we can see and chart but what else is in our data that we can’t see? Advanced analytics is capable of being adapted to asset performance to gain further insight. We can discover patterns in those sensor data streams that give us a precursor to an unplanned downtime not visible to the naked eye. Downtime is only one measure: We can also determine causes of degraded operations. This hits directly on availability and performance rates. There is another underlying component of OEE, the quality rate. In addition to equipment data, these manufacturing systems have been able to capture and store all the process data that went into making these products. Quality systems have been in place for a while to manage products to their specifications and help manufacturers identify scrap or product to be reworked. Again, advanced analytics can be used for production quality. Here we can let the math discover unknown correlations and relationships with processes, operations, and operators. These discoveries can be put into action and provide predictive quality alerts so that the cost of quality can be mitigated or avoided. Predictive analytics enable deeper process understanding and alarming before scrap is produced.
The second benefit to a deeper process understanding is the ability to optimize production processes. Predictive quality analytic models define the relationships between the inputs and that target (throughput, yield, etc.). That understanding allows prescriptive analytics to be generated and recommend actions back to production engineers and operators. They can set up the production process with the recommended process setpoints to maximize quality while reduction costs.
While I’ve spent decades implementing controls, SCADA systems and MES systems to improve on OEE and help manufacturing maintain competitive advantage, much of the low-hanging fruit has been picked and the next level of productivity and quality comes from the use of advanced and predictive analytics. Let the math find patterns or unknown correlations that lead to special cause variation in your processes. This deeper process understanding will allow you to be proactive and achieve the next level of performance from your operations.
David Frede, Product Manager, IoT