Monitoring manufacturing processes live and getting better with every component. Reference twins are used to detect defects right when they happen. And for constant improvement, reference twins learn from workers‘ feedback and constantly improve their threshold values.
Lower sunk costs:
Early defect detection and live notification of deviations and rejects
Optimized component requirements:
Adjustment of tolerance windows with each new component.
More process stability:
Automatic detection of defect patterns and parameter drift.
The main aim of the Online Monitoring application is to monitor component quality during production. This means, manufacturing defects should not be detected only at the end of the production chain, but at an early stage, when defects might still be repaired. This saves sunk costs for further processing of defective components.
Using reference twins to detect quality deviations online
The Online Monitoring application is based on a reference model that is either predefined by means of component requirements or automatically created by the nebumind software on the base of components that have already been manufactured. During the manufacturing process, the online monitoring application detects deviations from the reference model and informs the user or even stops production. The user can evaluate the deviations and, if necessary, adjust the process or pass on suggestions for changes to the engineering department.
AI algorithm constantly learning from feedback and improving detection
Furthermore, the reference model is not considered irrevocable in the online monitoring application, but the tolerance windows can be continuously updated across the entire component: Each evaluation of a user on deviations of a newly manufactured component can be fed back automatically into the reference model. Over time, a reference twin with high-resolution tolerance windows is created. Particularly in quality-critical industries, where tolerance windows are set extremely narrowly to guarantee the highest quality, unnecessary scrap can be avoided that lies outside the initially defined tolerances but is actually of sufficiently high quality.
The user manually defines a reference twin based on component requirements or rough tolerance windows developed during initial tests.
The user automatically creates a reference twin based on already manufactured components and their evaluation (OK/ NOK).
During manufacturing, the digital component twin is built up and compared with the reference model. Deviations are marked in the digital twin and noted in an evaluation list. The worker can immediately receive a message by mail or on his cell phone.
The user goes through the list of all marked deviations and evaluates them with OK or NOK. Based on the user’s evaluations, the reference twin is automatically updated.
Self-created scripts and algorithms can be loaded to use them for live analysis of data and trigger alerts if necessary.
Based on the user’s evaluations, the reference twin is automatically updated.