Practical aspects of virtual metrology and predictive maintenance model development and optimization

This paper describes practical aspects of development and implementation of novel process control entities such as Virtual Metrology (VM) and Predictive Maintenance (PdM), which utilize multivariate statistical models and machine learning techniques for prediction of process quality parameters and equipment faults. The description is based on the experiences collected during model development for VM and PdM. An overview of the main development steps including main challenges, potential solutions and applicable algorithms is given. The implementation of the steps is described at the example of the prediction of the filament breakdown in an implanter ion source