Just back in the office after Hannover Messe with a huge pile of material to think through, and some very interesting questions around the architecture model for IIOT and Industry 4.0. Once I have read through all the different point of views, I will condense my view into a follow up blog.
At the conference, in a coffee break, I had a conversation with a colleague about Google Now, and its promise to provide the right information at just the right time – without having to search for it. With all the predictive capabilities and big sensor data at hand we should be able to offer the same convenience to the factory workers. So for today, I would like to focus on the “human side” of IOT.
In my last blog on IOT I promised to take a closer look at “insight to action”. I would rephrase it to “providing the right information to enable smart decisions”.
This goes beyond “data driven decisions” in several aspects:
a) Information, or insight, is more than just data. It’s like the difference between a sensor reading, and the early warning that an asset is prone to fail in the next days.
b) “Insight to action” includes the element of alerting, making the user aware of the need to take action.
At SAP, “run simple” is our overarching tagline & promise, where “insight to action” is a key ingredient. This is also a design principle in our next generation business suite SAP S/4 HANA. But most importantly, this will make the life a lot easier for the person taking the decision.
The question is: Can we reach the same ease of use & smartness also on the factory floor?
Let’s approach this step by step:
Step 1: Plot sensor data on the screen, often including the allowed ranges (e.g. temperature, vibration or thickness) and some basic alerting if out of the allowed range. This is easily done – with SAP MII, or with the machine health monitor in SAP Predictive Maintenance, or with many other tool on the market.
Step 2: Advanced analytics and data mining on the sensor’ historic data. Purpose is to enable predictive maintenance, predictive quality. This may include to derive improved production parameters to avoid the issue. You may not need really “big” data, but you will need predictive tools, like SAP Predictive Analysis, or SAP Predictive Maintenance, and some data science skills.
Step 3: Run the model against active sensor data to alert on upcoming issues – and display the alerts to the operator and draw the attention to where intervention is needed.
The previous 3 steps are all “insight” – but at a different maturity level. Compare step 3 to classical reporting and printed excels pinned to machines or discussed in weekly operations meetings. This is a major cultural change.
Now let’ s prepare for
Step 4: Take ACTION. There are slight differences – who takes the action and how.
Alternative 1: The operator takes action, like creating a maintenance notification, adjusting some production parameters e.g. run speed.
If you prefer a more structured approach, short interval control (SIC) is worth to consider. SIC is a type of Kaizen process that uses real-time production data to guide immediate decision making on the shop floor. Within the time frame of a shift, you look back and review previous problems and losses. With prediction in place, you may even “look into the future” and consider upcoming quality or maintenance issues. Then you prioritize next actions and implement those.
Practically, SIC is implemented as a series of short reviews (5 – 10 minutes) several times during a shift – to enable continuous improvements on the shop floor level. (Find more information on SIC here: http://www.leanproduction.com/short-interval-control.html )
Alternative 2: The system suggests potential solutions. Each solution is evaluated in real-time and the user can preview the potential viability and impact before accepting a solution. A great example for this is the MRP cockpit in SAP S/4 HANA – described nicely in the SCN blog Real-Time Value for Manufacturing powered by SAP HANA.
The figure above gives you an impression of this interaction model – note especially the four tiles below the shortage graph: this are proposed viable remedies to the material shortage problem.
The level of simplification for the operator is dramatically. We still meet customers in our industry where workers need to walk through the plant to check whether the machines operate properly.
If you want to read more about this, and how it is realized in SAP S/4 HANA – search for “Smart Business Cockpits”.
Alternative 3: Automate.
In easy use cases full automation is possible e.g. to always trigger a maintenance notification or send an alarm message.
In more complex use cases, like when. prediction recommends to adjust the production parameters, this is rarely done automatically.
In Mill Products, we find customer that are still before step 1, and others already running proof of concepts and pilots for step 4.
Customer look at:
- reasons for unplanned downtime on a paper machine (due to paper breaks)
- dyeing variations or weaving on-loom quality control in textile manufacturing (see e.g. here or here)
- thermal imaging to analyse surface and subsurface defects in metal sheets, also thickness fluctuations causing cracks in later manufacturing steps
- predictive maintenance for grinding mill gear boxes
and many more.
Most mill customers I talked to were looking at alternative 1, although a few expressed intention to move into full automation were feasible.
I personally believe that our manufacturing processes are not deterministic enough to fully automate everything. Thus I would expect a mix of automatic rules and manual decision – which is good news for our factory people. They may need to learn new tools, but we will need their expertise also in the future.
Let us know about your plans. If you need help, to identify use cases for your industry and your business, please talk to us.