Prototyping embedded systems for vibrodrivers monitoring

Prototyping embedded systems for vibrodrivers monitoring. The following thesis concerns the development of an embedded system for the monitoring of vibratory hammers during different working conditions, in order to ensure the proper maintenance of the machine and avoid catastrophic failures. The work has been conducted in collaboration with PTC Fayat Group, which is an international company specialised in the design and manufacturing of pile driving and soil improvement equipment. The first part of the thesis regards the general description of vibrodrivers, by outlining its working principle, and the definition of the twin virtual model, with particular attention to the contact models used to describe the interaction between the machine and the ground. The virtual model consists in a lumped parameter system, which is characterised by non linear equations. Non linearities are due to the excitation forces and the contact models. The Dahl’s friction model is used to represent the contact between the lateral surface of the profile to be driven and the soil. The contact force between the tip of the profile and the ground is modelled with an exponential law, which takes into account the interpenetration between the two surfaces. The exponential law is function of an internal variable, which is essential to describe the dynamics of the ground. The second part of the work illustrates the experimental tests, which have been carried out to obtain data about the dynamic behaviour of the machine. The sine sweep tests have been performed on a test bench with a frequency which varies from zero up to the nominal working frequency. The time domain signals have been processed and analysed in the frequency domain, in order to identify the main parameters of the system, which are indispensable for a proper definition of the virtual model. The third part concerns the implementation of the extended Kalman filter algorithm, which can predict in real time the behaviour of the system and identify the force exchanged between the machine and the ground. The Kalman algorithm combines experimental measurements and estimates coming from the virtual model, in order to find the best prediction of the system’s states. The purpose is to provide to the extended Kalman filter only the signals recorded by the accelerometers, avoiding the integration of further sensors on the machine. The acceleration signals used in this part are simulated through the virtual model, then some noise is added to make the signals more similar to real ones. The procedure allows to test the effectiveness of the algorithm, by comparing the estimates with the simulated states and signals. In the fourth part, the extended Kalman filter algorithm is applied to the real system to perform the monitoring and identification of the interaction force with the environment. The acceleration signals used in this part are the real signals recorded by the accelerometers during the experimental tests on the test bench. The identified force is the force exchanged between the machine and the test bench.