Application of deep neural network in fatigue lifetime estimation of solder joint in electronic devices under vibration loading

Salameh, Anas A.; Hosseinalibeiki, Hossein; Sajjadifar, Sami

doi:10.1007/s40194-022-01349-7

Cited by 9 publications

(5 citation statements)

References 46 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To deal with package failure in solder joints, correlation-driven neural network (CDNN) 34 and deep neural network 35 were proposed to predict useful lifetime of solder joints in electronic devices. In, Samavatian et al 36 enhanced the CDNN method by establishing a novel iterative machine learning-aid framework to improve the useful lifetime prediction results.…”

Section: Related Workmentioning

confidence: 99%

“…Compared with two model-based methods, the LSTM method achieves a higher accuracy, while a larger dataset is required to train the model. Attention mechanism was first applied to the IGBT RUL prognostics 4 .To deal with package failure in solder joints, correlation-driven neural network (CDNN) 34 and deep neural network 35 were proposed to predict useful lifetime of solder joints in electronic devices. In, Samavatian et al 36 enhanced the CDNN method by establishing a novel iterative machine learning-aid framework to improve the useful lifetime prediction results.A hybrid approach intends to fuse physical information into the model-building process of the data-driven approach.…”

mentioning

confidence: 99%

See 1 more Smart Citation

Remaining useful lifetime estimation for discrete power electronic devices using physics-informed neural network

Guo

Liu

et al. 2023

Sci Rep

View full text Add to dashboard Cite

Estimation of Remaining Useful Lifetime (RUL) of discrete power electronics is important to enable predictive maintenance and ensure system safety. Conventional data-driven approaches using neural networks have been applied to address this challenge. However, due to ignoring the physical properties of the target RUL function, neural networks can result in unreasonable RUL estimates such as going upwards and wrong endings. In the paper, we apply the fundamental principle of Physics-Informed Neural Network (PINN) to enhance Recurrent Neural Network (RNN) based RUL estimation methods. Through formulating proper constraints into the loss function of neural networks, we demonstrate in our experiments with the NASA IGBT dataset that PINN can make the neural networks trained more realistically and thus achieve performance improvements in estimation error and coefficient of determination. Compared to the baseline vanilla RNN, our physics-informed RNN can improve Mean Squared Error (MSE) of out-of-sample estimation on average by 24.7% in training and by 51.3% in testing; Compared to the baseline Long Short Term Memory (LSTM, a variant of RNN), our physics-informed LSTM can improve MSE of out-of-sample estimation on average by 15.3% in training and 13.9% in testing.

show abstract

Section: Related Workmentioning

confidence: 99%

mentioning

confidence: 99%

Remaining useful lifetime estimation for discrete power electronic devices using physics-informed neural network

Guo

Liu

et al. 2023

Sci Rep

View full text Add to dashboard Cite

show abstract

“…24,25 However, the proposed ML models have mainly focused on the physical parameters of solder joints for estimating the fatigue lifetime under different states. [26][27][28][29] To be specific, the models collected the input parameters, such as geometry features, thermal load specifications, and physical properties of solder interconnections, and established a ML-based algorithm to predict the fatigue lifetime as the target. Hence, until now there has been no published work characterizing the fatigue microstructure of solder joints through ML-based approaches.…”

Section: Introductionmentioning

confidence: 99%

A Micromechanical Data-Driven Machine-Learning Approach for Microstructural Characterization of Solder Balls in Electronic Packages Subjected to Thermomechanical Fatigue

Kurniawan

Sayed

Luna

et al. 2023

J. Electron. Mater.

View full text Add to dashboard Cite

A combination of nanoindentation mapping and machine-earning (ML) modeling has been used to characterize the micro-structural changes in SnPb solder balls exposed to thermal cycling. The model facilitated the microstructural evaluation of solder bumps through the prediction of microscale variations of Young's modulus in the joint zone. The outcomes revealed that the micromechanical data-driven ML model precisely classified the microstructural constituents, i.e., β-Sn and α-Pb, along with the grain boundary (GB) regions. However, some deviations were observed in GB recognition, when the elastic modulus gradient was not sharp enough. The predictive results also revealed that the increase in number of thermal cycles led to stiffening and grain coarsening of α-Pb, while the β-Sn matrix mainly remained stable. Moreover, it was found that the thermal cycling intensified structural heterogeneity in the solder and sharpened the elastic modulus variations at the GB regions. In summary, the outcomes of this study demonstrate the prediction possibility of microstructural features in SnPb solder balls with a predefined thermal cycle numbers, and unfolded the relationship between morphological characteristics and microscale mechanical properties.

show abstract

“…To examine random vibration phenomena acting on the solder service lifetime, vibrations in different directions and angles are applied to printed circuit boards using finite element analysis. By training the desired neural network, the useful lifetime of the solder joints has been calculated with 91% accuracy [28]. According to the literature review, no research has been done about failure load and energy prediction in solder joints, especially when they are under mechanical loading.…”

Section: Introductionmentioning

confidence: 99%

Application of data mining techniques for assessment of fracture load and energy in double cantilever beam solder joints

Soroush,

Nourani

2023

JDAF

View full text Add to dashboard Cite

Predicting the exact values of fracture load and energy has a significant effect in preventing solder joint failure. In this research, various double cantilever beam (DCB) solder joints with different geometric constraints, including adherend thickness, adherend width, loading arm length, and solder thickness were tested under mode I crack propagation and at a strain rate of 0.03 . According to ANOVA test results, all the mentioned geometric factors influence the fracture load, but adherend width and solder thickness don't change the failure energy remarkably, considering the type I error is equal to 0.05. The failure load and energy forecasting using the random forest algorithm showed that the prediction accuracy is 92% and 81% respectively. Linear regression and lasso regression were also utilized to identify the solder joint's fracture behavior. The coefficient of determination in both methods is acceptable for solder joint fracture force prediction, but for fracture energy, it has decreased to about 70%. Finally, multi-objective optimization was done with the help of the NSGA-II method, and the Pareto front diagram drawn according to the problem's constraints to find the optimal values for fracture force and energy.

show abstract

Application of deep neural network in fatigue lifetime estimation of solder joint in electronic devices under vibration loading

Cited by 9 publications

References 46 publications

Remaining useful lifetime estimation for discrete power electronic devices using physics-informed neural network

Remaining useful lifetime estimation for discrete power electronic devices using physics-informed neural network

A Micromechanical Data-Driven Machine-Learning Approach for Microstructural Characterization of Solder Balls in Electronic Packages Subjected to Thermomechanical Fatigue

Application of data mining techniques for assessment of fracture load and energy in double cantilever beam solder joints

Contact Info

Product

Resources

About