Xiaojun Li

Su Chen, Xianwei Liu, Lei Fu et al.

Ground-motion model (GMM) is the basis of many earthquake engineering studies. In this study, a novel physics-informed symbolic learner (PISL) method based on the Nest Generation Attenuation-West2 database is proposed to automatically discover mathematical equation operators as symbols. The sequential threshold ridge regression algorithm is utilized to distill a concise and interpretable explicit characterization of complex systems of ground motions. In addition to the basic variables retrieved from previous GMMs, the current PISL incorporates two a priori physical conditions, namely, distance and amplitude saturation. GMMs developed using the PISL, an empirical regression method (ERM), and an artificial neural network (ANN) are compared in terms of residuals and extrapolation based on obtained data of peak ground acceleration and velocity. The results show that the inter- and intra-event standard deviations of the three methods are similar. The functional form of the PISL is more concise than that of the ERM and ANN. The extrapolation capability of the PISL is more accurate than that of the ANN. The PISL-GMM used in this study provide a new paradigm of regression that considers both physical and data-driven machine learning and can be used to identify the implied physical relationships and prediction equations of ground motion variables in different regions.

Xiaojun Li, David Jauernig, Mengzhu Gao et al.

A Battery Cloud or cloud battery management system leverages the cloud computational power and data storage to improve battery safety, performance, and economy. This work will present the Battery Cloud that collects measured battery data from electric vehicles and energy storage systems. Advanced algorithms are applied to improve battery performance. Using remote vehicle data, we train and validate an artificial neural network to estimate pack SOC during vehicle charging. The strategy is then tested on vehicles. Furthermore, high accuracy and onboard battery state of health estimation methods for electric vehicles are developed based on the differential voltage (DVA) and incremental capacity analysis (ICA). Using cycling data from battery cells at various temperatures, we extract the charging cycles and calculate the DVA and ICA curves, from which multiple features are extracted, analyzed, and eventually used to estimate the state of health. For battery safety, a data-driven thermal anomaly detection method is developed. The method can detect unforeseen anomalies such as thermal runaways at the very early stage. With the further development of the internet of things, more and more battery data will be available. Potential applications of battery cloud also include areas such as battery manufacture, recycling, and electric vehicle battery swap.

Su Chen, Yi Ding, Hiroe Miyake et al.

In scientific machine learning, the task of identifying partial differential equations accurately from sparse and noisy data poses a significant challenge. Current sparse regression methods may identify inaccurate equations on sparse and noisy datasets and are not suitable for varying coefficients. To address this issue, we propose a hybrid framework that combines two alternating direction optimization phases: discovery and embedding. The discovery phase employs current well-developed sparse regression techniques to preliminarily identify governing equations from observations. The embedding phase implements a recurrent convolutional neural network (RCNN), enabling efficient processes for time-space iterations involved in discretized forms of wave equation. The RCNN model further optimizes the imperfect sparse regression results to obtain more accurate functional terms and coefficients. Through alternating update of discovery-embedding phases, essential physical equations can be robustly identified from noisy and low-resolution measurements. To assess the performance of proposed framework, numerical experiments are conducted on various scenarios involving wave equation in elastic/viscoelastic and homogeneous/inhomogeneous media. The results demonstrate that the proposed method exhibits excellent robustness and accuracy, even when faced with high levels of noise and limited data availability in both spatial and temporal domains.

Yifan Yang, Qiao Jin, Robert Leaman et al.

The remarkable capabilities of Large Language Models (LLMs) make them increasingly compelling for adoption in real-world healthcare applications. However, the risks associated with using LLMs in medical applications have not been systematically characterized. We propose using five key principles for safe and trustworthy medical AI: Truthfulness, Resilience, Fairness, Robustness, and Privacy, along with ten specific aspects. Under this comprehensive framework, we introduce a novel MedGuard benchmark with 1,000 expert-verified questions. Our evaluation of 11 commonly used LLMs shows that the current language models, regardless of their safety alignment mechanisms, generally perform poorly on most of our benchmarks, particularly when compared to the high performance of human physicians. Despite recent reports indicate that advanced LLMs like ChatGPT can match or even exceed human performance in various medical tasks, this study underscores a significant safety gap, highlighting the crucial need for human oversight and the implementation of AI safety guardrails.

Xiaojun Li, Jianwei Li, Ali Abdollahi et al.

For electric vehicles (EV) and energy storage (ES) batteries, thermal runaway is a critical issue as it can lead to uncontrollable fires or even explosions. Thermal anomaly detection can identify problematic battery packs that may eventually undergo thermal runaway. However, there are common challenges like data unavailability, environment and configuration variations, and battery aging. We propose a data-driven method to detect battery thermal anomaly based on comparing shape-similarity between thermal measurements. Based on their shapes, the measurements are continuously being grouped into different clusters. Anomaly is detected by monitoring deviations within the clusters. Unlike model-based or other data-driven methods, the proposed method is robust to data loss and requires minimal reference data for different pack configurations. As the initial experimental results show, the method not only can be more accurate than the onboard BMS and but also can detect unforeseen anomalies at the early stage.

Yuan Ma, Kewen Liu, Hongxia Xiong et al.

The presence of residual and dense neural networks which greatly promotes the development of image Super-Resolution(SR) have witnessed a lot of impressive results. Depending on our observation, although more layers and connections could always improve performance, the increase of model parameters is not conducive to launch application of SR algorithms. Furthermore, algorithms supervised by L1/L2 loss can achieve considerable performance on traditional metrics such as PSNR and SSIM, yet resulting in blurry and over-smoothed outputs without sufficient high-frequency details, namely low perceptual index(PI). Regarding the issues, this paper develops a perception-oriented single image SR algorithm via dual relativistic average generative adversarial networks. In the generator part, a novel residual channel attention block is proposed to recalibrate significance of specific channels, further increasing feature expression capabilities. Parameters of convolutional layers within each block are shared to expand receptive fields while maintain the amount of tunable parameters unchanged. The feature maps are subsampled using sub-pixel convolution to obtain reconstructed high-resolution images. The discriminator part consists of two relativistic average discriminators that work in pixel domain and feature domain, respectively, fully exploiting the prior that half of data in a mini-batch are fake. Different weighted combinations of perceptual loss and adversarial loss are utilized to supervise the generator to equilibrate perceptual quality and objective results. Experimental results and ablation studies show that our proposed algorithm can rival state-of-the-art SR algorithms, both perceptually(PI-minimization) and objectively(PSNR-maximization) with fewer parameters.

Kewen Liu, Yuan Ma, Hongxia Xiong et al.

During training phase, more connections (e.g. channel concatenation in last layer of DenseNet) means more occupied GPU memory and lower GPU utilization, requiring more training time. The increase of training time is also not conducive to launch application of SR algorithms. This's why we abandoned DenseNet as basic network. Futhermore, we abandoned this paper due to its limitation only applied on medical images. Please view our lastest work applied on general images at arXiv:1911.03464.