Manufacturing analytics and machine learning benefits
Analytics and machine learning (ML) are the norm in manufacturing and can help users get better, more actionable data. Two examples are highlighted.
Analytics and machine learning (ML) are hot topics in manufacturing today. There is a lot of buzz around the possibilities to leverage the same artificial intelligence (AI) technology the credit card companies use to monitor and alert customers to “unusual” activity on their cards, and GPS apps use to guide users on a journey to avoid traffic jams and tolls. It’s the manufacturing version of Alexa or Siri sharing insights and suggesting the correct parameters to run against to control plant operations.
Some manufacturers have harnessed the power of AI and ML through software languages like R and Python, leveraging common platforms to visualize the results. Consider these examples of 2021 AI realities leverage current tools to bring business value to manufacturing.
Process control loops are a common component of manufacturing operations and one that can easily be performing sub-optimally for a variety of reasons. One manufacturer implemented an AI application as an analysis tool to dig into the various control loops inside a facility.
The output was a report card showing how each loop was performing. Key data samples from the historian were gathered for each loop (manual/auto status, how well it came up to setpoint and control oscillation once it got there, was its response or performance degrading over time, did the loop function differently in various situations) and then analyzed through R programming.
From there, the information and results of the analytics generated a proportional-integral-derivative (PID) report card with associated suggestions to review and walk through to determine paths to take to improve performance.
The results led to recommendations and justifications to re-tune loops, collect more data, ensure loops are in automatic mode, make process or instrumentation changes and add multimodal functions to the loop to improve control throughout the spectrum of setpoint demands. These changes reduce energy consumption, improve product consistency and reduce cycle times.
Many manufacturers struggle with weekly production planning and how to minimize changeover times in between product runs. In another AI example, one manufacturer had a fairly simplistic approach. A change from one stock-keeping unit (SKU) to another could be broken into one of four categories depending on whether or not they had to do a washout or gear/tooling change. The best scenario was if they had to do neither, followed by having to do a washout only, then having to do a gear/tooling change only and finally having to do both. So, in their scheduling, they would look to have the most “no wash nor change” and the least number of “both wash and change”.
However, inside the groups that require a “wash” only or a “gear change” only, they didn’t have any distinction as to which should be first vs. last from an efficiency perspective. As with most systems, there was a wealth of historically tracked data, most notably the changeover duration for each change per SKU.
Even though there were hundreds of SKUs, there was dozens of samples for any combination of products. The data, which was provided by implementing Python code, gave the user had a simple a simple interface and the user enters different SKUs they need to produce and the optimal order of execution is displayed.
The AI and Industry 4.0 path these manufacturers took was not one that required a huge lift in the organization. They committed to a structured approach to analyze and improve their operational challenges by leveraging a new set of resources and tools.
AI opportunity is here. Manufacturers need to take advantage of it.