Market Downturn is a Great Time for AI in Manufacturing

Inflation, rising material costs, reduced consumer spending…AI to the rescue. 

As the market tightens, efficiency gains and waste reduction are now top priorities.  So, how does artificial intelligence (AI) fit in?

Here at Amatrium, we are focused on delivering affordable AI solutions to manufacturers.  Our tools use machine learning, which is a subset of AI.  Machine learning is a method that identifies the factors that are driving scrap, and cutting scrap reduces cost and improves the production levels that are so critical with today’s labor challenges and high material costs.  

How does it work?  

We work with a manufacturer to help them understand that data that is required to make use of these tools.  In metals, this structured dataset typically is delivered as a spreadsheet on input variables, and output variables.  The inputs often comprise of the metal composition elements, each in its own column, and the process variables in columns.  The outputs are the metrics required by customers, typically mechanical properties, yet may include appearance, pressure tightness, X-ray grades and so on.  Scrap is typically associated to an output metric, and the challenge to process engineers is to understand which input variables are driving scrap.  With tens to hundreds of input variables, it is not humanly possible to determine the cause of scrap, which is why companies often have scrap levels between 5 and 30%.  In one case, a manufacturer had 90% scrap on a single day, with no clear explanation.    

Machine learning is well-suited in multi-variable analysis because it uses algorithms to detect patterns in data that drive an outcome.  There is no bias, no person saying ‘we have always solved it this way,’ or ‘we tried that before, it didn’t work.’  Manufacturing is far from stagnant.  In metals, coatings and lubrications evolve, tooling wears out, and alloys are tweaked for various reasons.  What worked 20 years ago may not be applicable today.  No one is suggesting that intuition is any less useful, yet the ML tools bring the ability to evaluate many variables and provide insight on the biggest factors for the situation at-hand. 

Besides solving today’s scrap issue, are there other applications for the machine learning tool? 

Absolutely.  Once a robust model is built to solve for the factors that are driving a particular output, the tool can be used in other directions.  Example, a customer wants to maximize the strength of 356 aluminum; the output yield strength can be fixed at the target level, and then the model can solve for the factors that were the input variables in the scrap calculation.  It can solve for a process condition or a chemical composition within the allowable ranges. 

Are you ready to investigate affordable AI solutions for your manufacturing operations?