ErgoTech is a leader in modelling and process optimization using machine learning and deep learning. There have been some fantastic advances in this field over the last three years, so if you tried machine learning and neural networks before, and think it won't work in your industry or application, it may be time to look again.

Machine learning and artificial intelligence are extremely powerful tools for process optimization but often very simple to use. This example of calculating optimal parameters based on historical data represents almost a case-book usage of a neural-network based learning system. The final TensorFlow model amounts to about 35 lines of code, much of the effort involves simply collecting and parsing the data. A traditional approach to this problem costs multiple millions of dollars.

More challenging is the idea of a "Golden" product and then classifying products as good or bad in real time. Building a model only from the process parameters with no additional measurement data is considered "unsupervised" learning. We've successfully tackled this problem for a particularly difficult data set in the semiconductor industry.

The problems in manufacturing are relatively simple compared to many problems being solved by modern AI systems. The time to make calculations between parts is generally long - seconds or even longer - and so doesn't call for massive hardware requirements. Many problems can be solved using these new tools and it's easier and so less costly than traditional approaches.

Image Credits: APC Conference Proceedings, Austin, TX Oct 2017 - From ErgoTech Presentations