- Augmented Analytics
- Business Analytics
- Big Data Analytics
- Data Governance & Data Fabric Architecture
- Performance Management & Improvement
- Data-Driven Strategy & Transformation
- About us
- Insights Room
- Webroom & Events
The three major success factors in manufacturing processes? Performance, availability and quality. Learn how to tackle all of them to increase overall equipment effectiveness (OEE), create transparency and ultimately reduce cost by using process mining and data that is already available in this blog post by Gerrit Kohrs.
Apr 21, 2020
When we take a bird’s eye view on manufacturing, there are three main questions that decide about success or failure:
- Does your throughput meet your plans?
- Can you deliver the desired quality?
- Is your equipment available when you need it?
The concept of OEE
When we talk about these questions, we are talking about one of the most popular KPIs in manufacturing: OEE or Overall Equipment Effectiveness. Let’s take a minute to wrap our minds around the underlying concept. OEE is a combined score that is made up of three components:
- Availability: The net run time of a given piece of equipment compared to your planned production time. Unplanned stops like technical failures or material shortages and planned stops like tool change-over or daily cleaning are both taken into account. Example: 20 hours scheduled production time per day, 45 minutes of planned maintenance and cleaning and an unplanned stop due to electrical problems of 15 minutes result in an availability ration of 19h / 20h = 95% availability (one hour lost)
- Performance: This one is all about speed. We determine the theoretical fastest throughput time for one piece of finished product (= ideal cycle time) and multiply by the achieved output during a given period of time, with gives us the theoretical run time. Then we divide this by the net run time during the same period and receive a percentage value for performance. Example: With an ideal cycle time of 2 minutes, 500 pieces manufactured and a net run time of 18 hours, the result would be (2 x 500) / 1080 = 92,6% (1,33 hours lost)
- Quality: Practically self-explaining. We divide the number of good products after inspection by the total number of pieces produced and there is our third percentage value. Example: 485 pieces have passed final inspection, 500 pieces have been produced, so our quality score is at (485pcs / 500pcs) = 97% (15 pieces lost)
Now that we have calculated these three scores (or losses, if you look at it the other way around), we combine all three into one total score by multiplying:
OEE = 95% availability x 92,6% performance x 97% quality = 85,33 %
The perfect score of 100% is pretty much impossible to achieve. What is a good value for your operation, depends a lot on your type of industry and various environment parameters. Therefore, benchmarking with similar companies is probably a good idea. With the 85,33% in our example, we’d already be “world class” as stated by Seiichi Nakajima, who introduced TPM in the 1980’s . So, for most companies, this would already be a very ambitious target.
How does process mining help?
The plain OEE scores above will neither tell you how well you use your potential, nor can you deduct how or where you could improve. In order to do that, you need to understand which parameters have an impact on these KPIs. This is where process mining comes in.
With process mining technology, you can analyze the digital footprint of your production. Process mining tools take logfiles from various source systems like your ERP, shop-floor management, machinery and equipment controls etc. and translate them into a visual representation of your process. Assuming these systems are connected or at least share some common data like order numbers, you can even visualize surrounding processes like sales order management or intra-company logistics, which might help explain dips in the availability-part of your OEE.
The visualization and analysis of the processes in your production delivers content to the bare numbers and thus helps not only with their interpretation, but also directly to process steps or sequences that concern probable problems.
What does that look like?
In our experience, when process owners think of their processes, they quite often have the picture in their head as they designed it. For example, a simple process for manufacturing a plastic part of some sort might look like this:
Reality of course is rather more complex and visualizing this complexity is one of the main objectives of process mining.
All the sudden we see a completely different picture with all those extra steps, loops, a sometimes seemingly strange order of activities as well as some unexpected manual work. At first glance, we see those time-consuming steps (many tools show them in red) and all those situations that drive machine operators crazy, like a cancelled order when they already had everything prepared for it. We also see a problem with intra-company logistics right away because we always need to wait for material.
In short, this graph – along with numerous other features of common process mining tools – will already give you answers why your scores for performance, availability or quality look the way they do. Most of the tools in their ready-made templates offer numerous detailed analyzes that provide you with further insights in the second step. Common findings include, but are not limited to:
Bottlenecks in certain steps of the production process:
- Insufficient capacity
- Technical problems
- Mistakes by operator
- Waiting for material
- Frequent tool changes, machine calibration for different parts
- Unplanned maintenance
- Changed / cancelled production orders
- Changes after setting up machine
- Loops going back to a previous step
What you see above is a rather simple example, of course, but imagine we would include the logs from machine controls to show error codes, sensor data, maintenance activities and many more. This would make potential trouble visible even before one of the three scores go down and give you time to react. And the best part: You would know exactly where to start improving.
Where to begin?
Our experience shows that a short pilot project or proof of concept (PoC) is a good start. In order to see first results, there are basically four simple steps, after you have identified the first process you want to examine:
- Extract event log data from the source system(s)
- Enrich this data with additional information from ERP or existing BI-applications
- Import data to a process mining tool
- Customize a standard template app to fit your needs
Still sounds like a complicated endeavor? Guess what, it’s our daily business and we’re here to support you every step of the way.
Obviously, manufacturing processes are not the only ones worth analyzing. Pretty much every digital process in your organization will have more or less potential for optimization and the message is the same: Getting started is simple, the data is already there, and insights are almost instantaneous.
Process mining technology offers full process transparency and data-supported decision-making aids for your process optimization. Think about how you could bring its advantages to use within your organization or how you can fully exploit the potential of your existing process mining tool. Feel free to contact us.