PART II: How Machine Learning Works With Your Existing Data

Home / PART II: How Machine Learning Works With Your Existing Data

PART II: How Machine Learning Works With Your Existing Data

Having data and metrics is great! But there comes a point when so much data is collected that a user has no clue what to make of it. And data overload can be a serious problem. Here is a perfect example;

Recently while visiting a client site, we asked the VP of Operations how their current MES solution was working and if it was providing what they needed. He told us that while it was much better than the old days of putting everything on paper, sadly many of his plant managers and supervisors spend 6-7 hours a day interpreting data. When there was an issue, they could absolutely see the problem. Reports and dashboards were everywhere. Giant TVs hung with pride on the shop floor!  But why did the issue occur? Better yet, how could it be prevented next time? He explained that his team would have to spend countless hours diving into the data to figure out the root cause to any problem. This pulled them away from the shop floor where he wanted them spending more time, making sure quality product was getting produced.

 

He also explained that the shop floor operators had begun to ignore alarms and didn’t trust the data or found it lacking context. His request was simple, can a new solution run the plant? He wasn’t advocating the removal of all human interaction, but rather challenging that we had to take the next step. How could we leverage all this data and use it to provide his team a solution that makes recommendations based on concise actionable data? A dashboard that would only show the user what they need to do to complete their job, and not force them to become data analyst.

Thankfully for this client, the push right now is towards Machine Learning. ML is a subset of AI and refers to the ability of machines to learn from data without being explicitly programmed. It is a method by which machines improve their performance over time using data. Machine learning algorithms detect patterns and trends in data, allowing them to make predictions or decisions based on this information. Everyone is talking about it and many companies have started implementing programs centered around ML. The big question is, how many companies are ready?
 
To answer that, we need to understand that ML works based on pre-existing data. If you feed the system bad data, you probably are not going to like the results. Essentially the system is only going to be as accurate as the data it’s given. Therefore, the first question is: where are you getting the key manufacturing data from? Operations, performance, and quality data are foundational elements regardless of how they are collected. As our clients discuss their desire to implement ML solutions, it’s critical that their current systems have accurate real time data from the production process.
 
Before moving forward with ML initiatives, we must understand the current landscape and whether you are truly ready to take the next step. That’s why Actemium takes a consultative approach to everything we do. It’s fundamental to our clients’ success that we help them define not just the challenges they are facing and how to solve them, but also look at creating the foundation to embrace and correctly implement new technology in the future. Essentially, we must avoid recreating the earlier problem of implementing MES just to implement MES, only this time calling it ML. We have to be open to adopting many new systems in the future that we can’t even begin to comprehend yet.
 
I still believe strongly that MES is relevant today. In fact, it’s more relevant than ever. While the focus is rightfully on next gen AI and ML solutions, we can’t lose sight of day-to-day operations of a manufacturing plant and how to harness their data. MES fits well into the overall software architecture as a key component of any strategy. We still need easy- to- use solutions that bring value to the manufacturing process and the people using it. It’s critical that any digital manufacturing strategy defines how data will be collected from the shop floor equipment, systems, and humans. Skipping over this step will potentially negatively impact any other initiative. Without MES, it would be impossible to transition into more AI and ML solutions.
 
With so much focus on data initiatives, many companies are jumping in headfirst. It’s an exciting time for manufacturing but I hope we don’t repeat the same mistakes. These programs take time and require a clear strategy and roadmap. The good news is that Actemium is here to help. Our digital transformation team is made up of experts in automation and controls, MES, and Data Ops. We believe bringing an experienced cross functional team together provides our clients not only value but de-risks their initiatives. We partner with our clients and become one team driving towards the same goals.

Leave a Comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.