Mondragon Innovation | Humanity at work – Detection of Anomalies in Components of Repetitive Part Production Machine Tools
Mondragon Innovation | Humanity at work
Sector: Industry
Business Case
The LG400 grinding machine performs the machining of large batches of repetitive parts in an automotive company. The machine is connected and sending data related to its use to the Danobat cloud without interruption. On November 2, 2020, the customer reported a problem with the machine’s part head.
Objectives
The goal is to create an algorithm capable of identifying anomalies or changes in the grinding process for repetitive parts, enabling the detection of issues in spindles. This is achieved by analysing machine variables captured at a frequency of 1 second.
Use case
The problem is approached in three phases: (i) Exploratory data analysis; (ii) generation of the anomaly detection algorithm; and (iii) implementation.
Infrastructure
Hybrid On premise Cloud
Technology
Machine learning and deep learning
Data
Temperature, intensity and current data from the machine in question.
Resources
The technical work was led by a PhD student from IDEKO and in terms of infrastructure, the data collection and initial exploration work was based on the features offered by Danobat’s cloud platform and the DANOBATBOX.
Difficulties and learning
In the context of high variability, such as in manufacturing with numerical control machines, a problem has been identified where the application of AI has the potential to yield positive results. In this context, the importance of having a repetitive part process is highlighted. The importance of pre-processing of raw data is also emphasised.
KPIs (business impact and metrics of the model)
Defect detection / labelled defects.
Funding
Private
Collaborators, Partners
IDEKO’s data analysis area was involved in the technological leadership of the use case.
