A question that seems to get asked often is how many industrial applications can we name for machine learning? Well, that’s the wrong question, writes Joseph Zulick from MRO Electric.
Machine learning is not a device you can plug into a production line and make it operate better than it did before. Machine learning is a process that needs inputs from many devices to feed data to it so that data can be collected, evaluated and used to develop knowledge about how it produces the products and parts it does. That knowledge can then be used to determine how it can have a higher throughput of parts, operate at a lower cost, and run more reliably. In that way, machine learning transforms an industrial operation into systems that can get products to market faster at a lower cost so the company that owns it can remain competitive in its market and keep its customers happy by delivering the products they want. If you're going to put a label on that application of machine learning, it’s a higher profit margin that will create more innovative products to make the customers even happier.
Process-based machine learning – a complex systems engineering solution
From Wikipedia via the peer-reviewed Springer journal, Machine Learning:
“Machine learning (ML) is the scientific study of algorithms and statistical models that computer systems use to perform a specific task effectively without using explicit instructions, relying on patterns and inference instead. It is seen as a subset of artificial intelligence. Machine learning algorithms build a mathematical model based on sample data, known as ‘training data’ to make predictions or decisions without being explicitly programmed to perform the task.”
Let’s add a modifier to the idea of machine learning and call it ‘process-based’ machine learning. That allows us to get to the heart of the matter in identifying the industrial technology that had to be created or modified because of the desire to use machine learning computer algorithms to enable the era of smart manufacturing.
Machine learning uses training data to teach its computer algorithm on what to expect from the production machines it’s monitoring to obtain that training data, relying on pattern recognition and inference to develop the capability for the algorithm to make decisions and predictions without having to write code to be explicitly programmed to perform that task. The training data is collected, processed, and evaluated in a structured sequence of steps to prepare that data for use in the machine learning algorithm. That structured sequence of steps is a process, and the creation of that process introduces new technologies in the form of IoT devices to create the data, networks to store and process the data, and computers to process and clean the data for accuracy and relevancy.
Industrial applications and transformations
The list of new technology that can be attributed to machine learning is exhaustive and not possible to be covered in its entirety in this article. Thus, we will hit on the higher-level issues that are more readily identifiable.
The possibility of being able to predict disruptions to the production line in advance of that disruption taking place is invaluable to a manufacturer. It allows the manager to schedule the downtime at the most advantageous time and eliminate unscheduled stoppages. Unscheduled downtime hits the profit margin hard and also can result in the loss of the customer base. It also disrupts the supply chain causing the carrying of excess stock. The need to bring additional manpower to bear via your third-party field engineering support can cost a lot of money as well.
A PwC study, Digital Factories 2020: Shaping the Future of Manufacturing, predicts that the adoption of machine learning to enable predictive maintenance is expected to increase among manufacturers by 38% because of the ability to increase profit margin by eliminating unscheduled stoppages.
Part 2 tomorrow.