An AI model for predicting device failures
Directing the company's preventive maintenance efforts to the machines which are likely to have downtime in the coming month. Done using a Deep Learning predcition model.
We are here for you!A developer and manufacturer of specialized technological products for the life science research and clinical diagnostics markets
Prediction of a critical failure for devices which are installed at customer sites, so to direct preventive maintenance.
Most business engagements are done using a Lease model. This means that if a machine is down at a customer site it has an immediate implication of no-revenue for the downtime. Thus, it is critical to predict the failures in advance to prevent them by focused preventive maintenance.
Method:
continuous collection of Logs from the devices at customer sites.
Data cleansing and smart manipulation on the raw data in a way that supports running Machine Learning models.
Running iterations of different statistical prediction models until reaching the appropriate accuracy.
Outcome:
A model that provides a list of machines that are likely to have a critical failure in the coming month. This allows doing preventive maintenance on these machines to prevent this failure from happening.
The model that was chosen for reaching the best prediction results is a Deep Learning one that exhibits an excellent ROC test (A test combining True Positive Rates and False Positive Rates of a model) as shown in the graph plot above.