Self Aware Factory in the Making

Manufacturing is a complex business. Product volume is exploding in industries like electronics, semiconductors, automobiles etc. On top of that, products are getting more and more complicated and design changes need to be incorporated every now and then to keep up with the market. All of this makes it difficult for line engineers to maintain quality effectively. In fact, companies are losing 15-30% of sales revenue because of poor quality. Saving this huge chunk could transform marginally successful companies into hugely profitable ones.

Understanding the root cause of every defect takes engineering data analysis. While an immense amount of data is generated from different manufacturing systems, process engineers spend 37-50% of their time just collecting, cleaning and organizing failure data (including visual data) coming from various sources before they are able to analyze it. This process is like firefighting and highly reactive in general. The engineers pull together all resources to fight one issue, while the root cause still persists in the line and defective products are being produced.  A lot of times, 2 new issues might crop up while the engineer is trying to solve one.

Most automation systems in a present-day factory are integrated. They have an MES or similar software to track the product, and machine vision systems to detect defects at multiple stages. In case of defect identification, the material is discarded into a reject bin and the process continues with good material. But the visual information is not used for further analysis. It acts like a simple go-or-no-go gauge with no information about why the failure happened.

Now imagine if there was a single platform connecting these three data sources:

  1. Product defect information in the form of image data received from machine vision systems
  2. Product tracking (include which machine it was run on) and yield information received from the MES
  3. Process Data (received from machine display or HMI)

If the data from all these three sources was automatically obtained, and then cleaned and correlated for defect analysis, it would save so much time for the engineers. They could use this automated platform for defect analysis directly.

What’s more, imagine if this platform had ML algorithms running in the background which could learn from factory data and engineers insights. Engineers could set the desired yield limits for their processes, and whenever the yield was below this level, the engineer records the root cause. This AI would learn the most likely root cause and pair that with data trends obtained from the line. In all future cases, this AI system could predict the likely root cause whenever similar data trends occurred. None of the engineers would need to do repetitive data analysis. Experience and learnings from past issues would be obtained by new engineers automatically, without getting lost when the experienced engineer leaves.

If a system like this was used at every processing point in the product life cycle, it would effectively track the product performance and create a feedback loop when failures occur. This platform would be a single point reference for both line operator and CEO to trace product and line performance, predict yields and throughput accurately.

At FireVisor, we have built a single point data analytics platform that connects to data sources in the manufacturing line, and automatically performs engineering failure analysis in real-time. Our predictive engine, backed with clean engineering datasets and process engineering insights, will be able to predict and prevent defects before they happen.

The manufacturing world is now at industry 3.0 stage, where if-else type automation in robotic arms and logic controllers is driving manufacturing. Our mission is to transform the industry and build the next stage of cognitive automation, where industrial systems are smarter, more productive, and have much lower error rates. We are building self-aware factories of the future which learn from itself.

Leave a Comment

Your email address will not be published. Required fields are marked *