Loading...

Lead-Free Solder Joints Fracture Load Prediction by Considering the Effect of Strain Rate on Joint Behavior Using Artificial Neural Networks

Soroush, Hossein | 2024

0 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 57283 (08)
  4. University: Sharif University of Technology
  5. Department: Mechanical Engineering
  6. Advisor(s): Nourani, Amir; Farrahi, Gholam Hossein
  7. Abstract:
  8. Given the extensive advancement of the industry and the widespread use of electronic devices, the safe operation of these tools is crucial. Electronic devices such as mobile phones and laptops are prone to impacts or falls, making mechanical loading evaluation on them particularly important. This research utilized DCB solder joints for mechanical loading and fracture testing on solder joints in the first mode of crack propagation. By initially examining environmental factors and geometrical constraints affecting the force and energy of fracture, influential factors were identified through single-variable analysis of variance (ANOVA) tests. Based on this, variables such as the adherend thickness, solder thickness, adherend width, maintenance temperature, and moisture content were found to have a significant impact on the solder joint fracture force. On the other hand, it was determined that the width and thickness of the joint do not play a significant role in determining the fracture energy of the samples. In the next step, using various machine learning tools including the k-nearest neighbor algorithm, regression, random forest, neural network, and maximum gradient boosting (XGBoost), the failure load and energy of solder joints were estimated at a strain rate of 0.03 per second. According to the results obtained, the neural network model provides the highest prediction accuracy and the lowest error among the algorithms used. Accordingly, the fracture load and energy were estimated using the designed neural network with accuracies of 86% and 81%, respectively. In the following step, by extending the strain rate interval, four factors including strain rate, solder length, adherend thickness, and solder thickness were selected as effective factors on fracture load using the Plackett-Burman test, and data collection was performed by selecting appropriate levels for each variable. Analyzing the obtained results revealed that in quasi-static strain rate, the average fracture force increases further with an increase in solder length. Additionally, an increase in adherend thickness also leads to an increase in the fracture force in all strain rate conditions. Another notable point in this regard is that the increase in strain rate from quasi-static to medium has a greater impact on the increase in fracture load compared to the scenario where the strain rate increases from medium to high. Furthermore, a comprehensive model for predicting fracture force has been presented using various machine learning algorithms. Based on this, the Random forest model with an accuracy of 83.2% demonstrated the best performance in estimating the fracture load of DCB solder joints. Finally, using the NSGA-II algorithm, optimization of the solder joint was carried out to maximize the fracture force and minimize the manufacturing cost of the specimen
  9. Keywords:
  10. Fracture Energy ; Strain Rate ; Plackett-Burman Method ; Machine Learning ; Artificial Neural Network ; Non-Dominate Sorting Genetic Algorithm (NSGAII) Method ; Solder Joints ; Double Cantilerer Beam ; Double Cantilever Beam (DCB)Solder Joints

 Digital Object List

 Bookmark

...see more