The customer was working on the launch of a “box 4.0” that would be used to collect data from the robots on the customer’s production lines. The customer wanted to better understand the use of its robots, and feed its future predictive maintenance service.
In addition, for each new robot model, the customer subjected one or more robots to endurance tests, in order to gather information on possible model failures. It was on the basis of this data that the customer wished to evaluate DiagFit. The data supplied corresponded to weekly (position, force, etc.) and hourly (temperature) recordings on various 4-axis and 6-axis robot models. The data had a cyclic behavior.
The customer wanted to:
- Produce a predictive model for one type of robot: 4-axis (initially)
- Focus more specifically on predictive capability for two types of failure (harness failure and gearbox failure)
- To be advised on the minimum data to be transmitted and stored, and to see if the model generated could be used with another type of robot.
A model was built using DiagFit
The model generated showed potential for detecting changes of state, could be tested on real anomalies (there was sufficient data) and proved 100% effective.