# Isik University - INDE3233 - Ergonomics (Summer 2024)

###### tags: `INDE3233`

## Assignment Paper

**Name-Surname :** Emre Berk Günel
**Student Number:** 19INDE1012

### Abstract:
**The risk of vibration, a well-known source of diseases associated to the workplace, is covered in this article. In order to gather vibration data and divide the resultant signals into time frames, a wearable device has been designed. It is suggested to use a machine learning classifier to identify worker activities and evaluate the dangers associated with vibration exposure.**

**APA Citation:**
:::info
Aiello, G., Certai Islam Abusohyon, A., Longo, F., & Padovano, A. (2021). Machine learning approach towards real-time assessment of hand-arm vibration risk. IFAC-PapersOnLine, 54(1), 131-136. https://doi.org/10.1016/j.ifacol.2021.08.140
:::

### **Introduction**
The goal of the project was to monitor workers' conditions in real time using Human Activity Recognition (HAR) approaches in order to improve occupational health and safety in smart production settings.The goal was to improve ergonomics and worker safety in the context of Industry 4.0, with an emphasis on safer and healthier work environments in light of the aging workforce.

### **Methods**
As part of the research, a real-time monitoring system for determining the danger that employees face from hand-arm vibrations while doing their duties was developed and tested.The information included vibration signals captured by smart wearable devices equipped with triaxial accelerometers from instruments commonly used in production.Vibration signals from workers wearing a wearable prototype device that was used to record data during simulated tool operations in a lab environment were recorded. The wearable device's local processing performed a 40-second window segmentation and risk evaluation on the vibration inputs. Based on vibration characteristics, the data was divided into heavy-duty (HD) and low-duty (LD) activities using a k-nearest neighbor (k-NN) classifier.

### **Results**
With the help of the k-NN classifier, the system successfully and accurately classified HD and LD activities. The system's decentralized architecture made it possible to monitor several workers simultaneously and scalable.The results are consistent with previous research on the efficiency of HAR approaches and machine learning applications for risk assessment and activity identification in industrial environments.

### **Discussion/Conclusion**
The study proved that real-time monitoring solutions for enhancing workplace health and safety in intelligent production settings are both technically feasible and effective.The outcomes validated the concept of utilizing wearable devices and machine learning algorithms by demonstrating the system's ability to effectively categorize and monitor various job activities and related dangers.The study has important ramifications for improving ergonomics and workplace safety through real-time data collection and analysis, supporting the wider use of Industry 4.0 technologies.The controlled laboratory setting in which the trials were carried out might not have accurately reflected real-world circumstances. The right choice of parameters for the k-NN method, which was established empirically in this work, determines the classification accuracy.Testing and verifying the system in real-world industrial situations should be the main emphasis of future research. Increasing the k-NN classifier's parameters may improve the system's performance and accuracy.