The new book written by the Lean Six Sigma expert Michael L. George and others is the first one I have read that goes deep into improvement areas in manufacturing by application of deep learning technologies.
The book touches several important aspects and needs in production environments and people with a certain level of knowledge in manufacturing feel comfortable with deployed thoughts and assumptions.
The authors emphasize in Chapter 3: “…Artificial Intelligence coupled with Lean Six Sigma is creating the Fourth Manufacturing Revolution. The result is increasing flexibility, autonomous self-adaptability to demand, and lower costs in the face of shorter product lifetimes, lower or static prices, and market demand fluctuations. AI is effectively a new “factor of production” in addition to labor and capital, as discussed in the author’s note…”
It is a little bit funny that the authors state in the same chapter under the headline “Third Manufacturing Revolution”: “…Lean Six Sigma does not provide a cost-effective answer to the twenty-first-century problem of low-volume, non-repetitive Job Shop Manufacturing, in which part numbers may not be repeated weekly, but only monthly or quarterly…”
This means that the philosophy of Lean Six Sigma has to change in the age of AI. Fortunately, there is already a different proven philosophy available that deals with low-volume, small-batches, none repetitive tasks, Pull systems, Replenishment, etc. in Job Shop Manufacturing. This is the TOC – Theory of Constraints developed by the late Eli Goldratt. Of course, neither Lean Six Sigma nor TOC took advantage of neural networks in the past. Therefore, the book is useful for both (meantime) very similar concepts.
Michael George makes the following suggestions in the books:
- Focus on Setup-Time reduction. Toyota’s Four Step Rapid Setup is not sufficient.
- The Setup-Time reduction should go hand in hand with standardization activities to reduce detail complexity as well.
- Implement a Data Mining setup for all kind of company data in a cloud to identify areas of unneeded but today hidden waste.
- Make use of data for Job Shop Scheduling. The major suggestion of the book is to run the scheduling tasks with the help of deep learning algorithms (neural networks). These algorithms can run much faster compared to MILP (Mixed Integer Linear Programming) or simple Linear Programming algorithms.
- Furthermore, the authors propose to reduce engineering time by making reuse of standardized components (point #2).
- Already discussed in the AI community is the proposal to apply neural networks for visual inspections in quality.
Michael L. George and his co-authors go deep into the technical details of the Neural Networks and they share quite a lot of important information without losing the focus on the manufacturing environment needs.
Therefore, I can highly recommend the book for people who are in manufacturing or in charge of production that want to prepare their plant for the future.