End-effector tools wear prediction: a multimodel approach

Nowadays there is a need for many companies emerging in the context of Industry 4.0 to save the costs, to increase efficiency and improve the factory management. For this reason the research for predictive maintenance techniques and state of health (SoH) estimation of a production machine is one of the most relevant areas in the scientific field. This thesis is inside the research and development MOREPRO project owned by Brain Technologies. Morepro aims to bring to the field an innovative solution based on multi-level distributed intelligence logic to improve the management of production plants thanks to new capabilities. The objective concerns the development of a prototype which is capable of: • Monitoring the SoH of machine and its critical components through embedded sensors and hence applying machine learning and data mining techniques. • Monitoring the wear condition of the machine’s tool using a digital twin approach, combining real-time signals with estimated quantities in a virtual simulation environment. • Developing predictive models able to estimate the on-line SoH and the trend of degradation states of machine and system components over time. The contribution of this thesis work can be divided in three main sections: 1. Development of a basic model: the first phase is the physical model’s basic design of a CNC machine, in order to be able to develop a prediction algorithm. The approach was to combine the equations of state of a dc motor with the mechanical equations of a cnc machine and simulate the plant to collect the values of the analysed state variables. 2. Predictive multimodel: the core of the project where a prediction analysis on the state variable and wear’s parameter estimation are developed using a bank of Extended Kalman Filters and a logic of residual error management. 3. Model upgrade and predictive multimodel update: an in-depth modelling of the interaction between CNC end-effector and workpiece, through the study of how various parameter can impact the end-effector’s wear condition. An update of the EKF’s bank is made accordingly to the upgrade of the model. Finally various test were carried out to check the overall system correct behaviour.