This project will be supervised by Professor Matteo Savino (Matteo Savino is an Associate Professor interested in Operations Management, Product Development, Supply Chain, Production Engineering, Logistics, working in the Department of Engineering , areas of Energy Engineering and Production Management.

Despite the progress made by research and politics, the workplace continues to be characterized by numerous injuries that often are deadly for the operator. The goal of having a safe workplace is often jeopardized by procedures and behaviors that don’t pay attention to prevention.

From the point of view of safety, the companies must focus their efforts on a prevention policy as a set of measures to be implemented in order to anticipate a potential risk.

Ergonomic risks cause losses of efficiency to the workers and, in the long period, social, health and economic harm for employers and the whole economy.

This aspect is an essential point to ensure the valorization and well-being of workers in factories, directly affecting company performance and productivity.

An ergonomically designed workstation also helps to reduce the risk of injury. Therefore, in order to increase the efficiency and quality of work in the factories, a preventive and objective evaluation of the ergonomic exposure level for each workstation of the production system is required.

The suggested project activity focuses on the research and experimentation of new tools and methods for posture analysis and its ergonomic evaluation through the use of Artificial Intelligence Techniques, with virtualization of the production process. In the field of Artificial Intelligence planning, the use of the Artificial Neural Networks will be explored through MATLAB software with which simulations will be carried out in order to find the optimal training network.

Specifically, the project consists of three macro-activities:

· Development of ergonomic matrices that describe posture / machine requirements researching for risk factors;

· Working with the developed matrices as input for the learning techniques developed by experienced systems

· Validation of the model by comparing the training data and the expected outputs.

The Neural Network Toolbox offers a wide variety of architectures and training functions for modeling complex nonlinear systems easily, using artificial networks. Apps available in the Neural Network Toolbox allow you to plan, train, visualize, and interactively simulate the network to subsequently generate the equivalent MATLAB code so you can automate the process.

Toolbox supports supervised, unsupervised and reinforced learning by using architectures such as perceptron, multi-layer perceptron, radial base, self-organizing maps, competitive networks, and so on. With the goal of accelerating complex training processes from large data volumes, you can deploy calculations on multicore, GPU, and cluster machines using the Parallel Computing Toolbox.

The aim of this project is to understand what can be the most suitable network configuration in order to make a correct ergonomic evaluation. You will also need to understand which countermeasures could be used to reduce the ergonomic risk level in order to have a safer working area.

The project will be supervised by Professor Matteo Mario Savino and his PhD student Marialuisa Menanno.