At the core of each equipment resides the PLC (Programmable logic controller). This is the computer that controls every aspect of the machine, including its sensors. Every sensor has a story to tell if you are paying attention. The problem is no human alone is capable of identifying data patterns that those sensors generate in the speed and precision required for any prediction to be useful. This is where a machine learning model can be trained. There are endless questions that can be answered by a machine learning model, take a look at some of them:

• How can the combination of engine speed and raw material temperature lead to a loss of performance of the equipment?

• Is there any relation between the environment conditions and product failures?

• What parameters are relevant to be monitored to predict equipment breakdown?

• Does humidity levels and end of line voltage correlate with producing better or worse products?

OEE (Overall Equipment Effectiveness) has been the industry standard to evaluate manufacturing performance at equipment level since the 1960s. We have come up with a new and modern way to predict efficiency at the operation level. Take a look below at some of the advantages of Operation Effectiveness:

• Using advanced statistical learning techniques, it is possible to predict parameters, such as planned process, rework and unavailable times;

• Those methods take into account historical data (same as OEE) but they can also use several other operation data points in order to better predict;

• Tailored to each manufacturing unit, after a machine learning model is created, predicting new values is incredibly easy and fast;

• Efficiencies are calculated at the operation level, meaning it provides a much better value for quotation and planning processes.

Supplier Management is a complicated subject that every manufacturing unit has to deal with. There are several decisions that are taken using information provided by suppliers as input, but, how trustfull are those data points? You may want to hire more operators knowing that material will be available as per promised dates from the suppliers; but what if those purchase orders are not delivered on time? Those are some of the topics machine learning and data science can help ease supplier management:

• What is the predicted delay for an open purchase order?

• What is the predicted quantity of fault pieces?

• What suppliers provide the best cost x benefit relation for a given raw material type?

• What is the chance for an open purchase order to be canceled by the supplier?

• Is it better to purchase from a supplier closer to you that can deliver their products with a shorter lead time or purchase from a supplier further away that provides higher data accuracy?

In reality, machine learning models can be applied to most areas, if not all, from a manufacturing unit. Every recorded transaction can be utilized to train a machine learning model. Below are some other questions that statistical learning methods can be leveraged:

• Manufacturing cost prediction. What if you could better predict the cost to manufacture a finished product? This will lead to a much better quotation pricing calculation;

• Delay threats. Knowing any potential threat of manufacturing delay can help focusing efforts on managing suppliers or operations;

• What shift is predicted to perform better for each type of product? Each operator is different from one another, so assigning the right products to the right team could be the difference between being on time or not;

• Where should you produce your finished product? What provides the best return on investment, being close to suppliers or to your market?