Virtual Sensors for Human Safety Monitoring in Factory 4.0 Applications

The safety of operators in Factories is a key problem since the Industrial Revolution. The growth of the technology let the development of new safety monitoring systems, often integrated with the industrial machinery itself. However, with the Factory 4.0 machines became gradually more autonomous and so the role of the worker has shifted towards the maintenance and the malfunctions management of these ones. The main goal of the future society should be to understand the meaning of this new interaction and to exploit the technology not only for improving performances, but also for protect the safety and the security of the operator. This Thesis work, developed in collaboration with Modelway s.r.l, starts from these considerations and aims to design a virtual sensor for human state monitoring during work activities. In includes two work packages (WP07, WP09) inside the Disloman project in which two different virtual sensors have been implemented. For both of them an Artificial Intelligence algorithm developed by Modelway and called DVS(Direct Virtual Sensor) has been used. The first one is able to detect the fall of the operator starting from acceleration measurements. They are acquired at a sampling time of 50 ms by a wearable sensor placed in a bracelet in order to detect the wrist movement. The DVS/MDD (Man Down Detection) reads these variables and generates in output the estimation of the current state of the operator; after this a state machine has been implemented in order to elaborate the output of the virtual sensor, to provide the transitions between status and to return a feedback of the current one. In this project another DVS called DVS/R able to reset a wrong estimation of the DVS/MDD has been developed. The control unit chosen to manage these procedures is the Raspberry Pi 3 B+. The second virtual sensor is able to track the arm position of the operator that works near a mechanical press (at Honestamp Factory). In particular, it stops the action of the machinery each time the hand of the worker is inside the dangerous area. For this purpose two wearable sensors are used (placed in a bracelet and on top of the arm) which send measurements of accelerometer, gyroscope and magnetometer at a control unit (Raspberry Pi 3 B+). Also in this case it analyses these signals and performs the Artificial Intelligence algorithm for the Arm Tracking estimation. Its output is received by another state machine that provides the transition of the states.