logo SBA

ETD

Digital archive of theses discussed at the University of Pisa

 

Thesis etd-07122023-173018


Thesis type
Tesi di dottorato di ricerca
Author
DI RIENZO, FRANCESCO
email address
f.dirienzo1@studenti.unipi.it, francescodirienzo93@gmail.com
URN
etd-07122023-173018
Thesis title
Technologies for monitoring safety and well-being of workers in industrial environments
Academic discipline
ING-INF/05
Course of study
SMART INDUSTRY
Supervisors
tutor Prof. Vallati, Carlo
tutor Tognetti, Alessandro
tutor Genovesi, Simone
Keywords
  • indoor localization
  • industry 4.0
  • monitoring safety
  • object detection
  • wearable device
  • workplace safety
Graduation session start date
20/07/2023
Availability
Full
Summary
My thesis aims to identify innovative approaches to improve workplace safety, prevent accidents and enhance workers' well-being in the context of Industry 4.0. My research focuses on identifying Industry 4.0 technologies that can be exploited to create innovative systems to reduce risks for workers and prevent accidents in the industrial environment. To achieve this, I will exploit some of the technologies that are fuelling the Industry 4.0 revolution: Near Field Communication (NFC), Object Detection, Bluetooth beacons and non-invasive wearable devices.
The first chapter is an introduction where I analyse the problem of security in the workplace. From the studies and statistics gathered, it can be seen that worker safety is mainly given by the use of safety devices that reduce the possibility of fatal accidents in the industrial environment. Hence the need to implement systems that improve worker safety.
The second chapter presents NFC, a short-range wireless communication technology that can be used to exchange data between devices. For example, NFC can be used to monitor the use of Personal Protective Equipment (PPE) by workers in the context of Industry 4.0. In particular, NFC tags can be attached to PPE, such as hard hats and safety glasses, to track the use and location of devices in the work area by placing several readers at workstations. In this way, supervisors can quickly identify whether a worker is wearing the appropriate PPE, thus ensuring that the equipment is used correctly and continuously. My contribution is an analysis of the limits of this technology for monitoring in industrial and medical contexts, including experiments to assess the potential of NFC in these fields through experimental setups. The experiments demonstrated the feasibility of using NFC for occupational safety, and a prototype smart glove was developed that can monitor the external temperature of the glove and alert the worker when the hand moves near a hot surface. When the glove senses a temperature above a threshold, it sends a warning signal to the worker in real-time.
In the third chapter, I examined Bluetooth beacons intending to use them for indoor localisation. These small battery-powered devices use Bluetooth Low Energy to transmit signals that can be received by other devices, such as smartphones and tablets. These signals allow us to measure the relative distance between the Bluetooth Beacons and the receiving device. These measurements can be used to track the position of a Bluetooth Beacon or the receiving device in space. This approach can be used in various industrial applications, such as tracking equipment, vehicles and personnel. In addition, Bluetooth beacons can be used for inventory management, tracking the location of specific items within a warehouse or manufacturing facility, enabling efficient inventory management and reducing wasted time searching for specific products. My thesis focuses on a use case: monitoring the location of a product by tracking which workstations it passes in the assembly line using Bluetooth beacons. To this end, a prototype sensor table was developed and tested to evaluate the performance of the technology and demonstrate the feasibility of its use for precise tracking in closed environments. Different models of Regressors, Convolution Neural Network and Recurrent Neural Network were tested. After a data collection phase, the models were trained and conclusions were drawn. The model that achieved the best accuracy was the K Nearest Neighbors, obtaining a result of 99% accuracy.
In the fourth chapter, I looked at Object Detection in the industrial environment for PPE detection to monitor whether a worker is wearing the appropriate protective equipment, such as a hard hat, safety glasses or ear protectors at all times. One of the main advantages of object detection for PPE detection is its ability to operate in real-time: image analysis can be integrated into cameras and other sensors placed in industrial environments. As soon as a worker enters the field of view of a camera, the technology can detect whether he or she is wearing the appropriate PPE and alert him or her if the PPE is not being used or is being worn incorrectly. This real-time functionality can help prevent accidents and injuries. This chapter reports the development of a real-time PPE monitoring system running on an embedded system. The system consists of three nodes: an image and depth camera (Intel RealSense D435), a control board (Raspberry Pi 4) and an image inference board (Intel Neural Compute Stick 2). To train the selected neural networks, three image datasets were collected in industrial environments. The best-performing network was the Yolov4-Tiny with a speed of 6.8 Frames per Second and a mAP@50 of 86.3.
In chapter five, I considered the latest technology under scrutiny: wearable devices. Wearable devices are widely used in industrial environments to monitor the safety and well-being of workers. These sensors can track a wide range of data, including heart rate, body temperature and movement. They provide real-time information to help managers and supervisors make decisions about worker safety and productivity. Specifically, we analyse and develop an innovative sock with an integrated inertial measurement unit (IMU) and some pressure sensors for gait analysis. Gait analysis is important for predicting a possible worker fall and can send an alarm when the worker falls. The information gathered can be very useful in highlighting how repetitive work can generate specific occupational diseases for the worker.
The last chapter outlines the conclusions of the studies carried out. All the prototypes developed are part of an ecosystem of solutions and can be used together to improve the safety and well-being of workers. These systems are complementary in that each one considers a different aspect, from accident prevention to compliance: the smart glove prevents hand accidents by preventing burns through temperature monitoring, BLE tracking monitors hazardous products, PPE detection enforces occupational safety regulations, and wearable devices such as socks monitor workers' movement to prevent long-term illnesses.
File