skip to main content

Explore Research Products in the NSF-PAR

Title: Physics-Informed Machine Learning for Degradation Diagnostics of Lithium-Ion Batteries
State of health (SOH) estimation of lithium-ion batteries has typically been focused on estimating present cell capacity relative to initial cell capacity. While many successes have been achieved in this area, it is generally more advantageous to not only estimate cell capacity, but also the underlying degradation modes which cause capacity fade because these modes give further insight into maximizing cell usage. There have been some successes in estimating cell degradation modes, however, these methods either require long-term degradation data, are demonstrated solely on artificially constructed cells, or exhibit high error in estimating late-life degradation. To address these shortfalls and alleviate the need for long-term cycling data, we propose a method for estimating the capacity of a battery cell and diagnosing its primary degradation mechanisms using limited early-life degradation data. The proposed method uses simulation data from a physics-based half-cell model and early-life degradation data from 16 cells cycled under two temperatures and C rates to train a machine learning model. Results obtained from a four-fold cross validation study indicate that the proposed physics-informed machine learning method trained with only 60 early life data (five data from each of the 12 training cells) and 30 high-degradation simulated data can decrease estimation error by up to a total of 9.77 root mean square error % when compared to models which were trained only on the early-life experimental data.  more » « less
Award ID(s):
2015710 1611333
NSF-PAR ID:
10325841
Author(s) / Creator(s):
; ; ; ; ;
Date Published:
Journal Name:
47th Design Automation Conference (DAC)
Volume:
3A
Page Range / eLocation ID:
V03AT03A041
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. ; ; ( , Mechanical systems and signal processing)
    Wang, Dong (Ed.)
    Lithium-ion batteries have been extensively used to power portable electronics, electric vehicles, and unmanned aerial vehicles over the past decade. Aging decreases the capacity of Lithium-ion batteries. Therefore, accurate remaining useful life (RUL) prediction is critical to the reliability, safety, and efficiency of the Lithium-ion battery-powered systems. However, battery aging is a complex electrochemical process affected by internal aging mechanisms and operating conditions (e.g., cycle time, environmental temperature, and loading condition). In this paper, a physics-informed machine learning method is proposed to model the degradation trend and predict the RUL of Lithium-ion batteries while accounting for battery health and operating conditions. The proposed physics-informed long short-term memory (PI-LSTM) model combines a physics-based calendar and cycle aging (CCA) model with an LSTM layer. The CCA model measures the aging effect of Lithium-ion batteries by combining five operating stress factor models. The PI-LSTM uses an LSTM layer to learn the relationship between the degradation trend determined by the CCA model and the online monitoring data of different cycles (i.e., voltage, current, and cell temperature). After the degradation pattern of a battery is estimated by the PI-LSTM model, another LSTM model is then used to predict the future degradation and remaining useful life (RUL) of the battery by learning the degradation trend estimated by the PI-LSTM model. Monitoring data of eleven Lithium-ion batteries under different operating conditions was used to demonstrate the proposed method. Experimental results have shown that the proposed method can accurately model the degradation behavior as well as predict the RUL of Lithium-ion batteries under different operating conditions. 
    more » « less
  2. ; ; ; ; ( , Journal of The Electrochemical Society)

    Differential voltage analysis (DVA) is a conventional approach for estimating capacity degradation in batteries. During charging, a graphite electrode goes through several phase transitions observed as plateaus in the voltage response. The transitions between these plateaus emerge as observable peaks in the differential voltage. The DVA method utilizes these peaks for estimating cell degradation. Unfortunately, at higher C-rates (above C/2) the peaks flatten and become unobservable. In this work, we show that, unlike the differential voltage, the peaks in the 2nd derivative of the expansion with respect to capacity remain observable up to 1C and thus make possible diagnostic algorithms at these charging rates. To understand why that is the case, we have developed an electrochemical and expansion model suitable for model-based estimation. In particular, we demonstrate that the single particle modeling methodology is not able to capture the peak smoothing effect, therefore a multi-particle approach for the graphite electrode is needed. Additionally, model parameters are identified using experimental data from a graphite/NMC pouch cell. The proposed model produces an excellent fit for the observed electric and mechanical swelling response of the cells and could enable physics-based data-driven degradation studies at practical charging rates.

     
    more » « less
  3. ; ; ; ; ; ( , 2021 IEEE International Conference on Prognostics and Health Management (ICPHM))
    Failure prognostics is the process of predicting the remaining useful life (RUL) of machine components, which is vital for the predictive maintenance of industrial machinery. This paper presents a new deep learning approach for failure prognostics of rolling element bearings based on a Long Short-Term Memory (LSTM) predictor trained simultaneously within a Generative Adversarial Network (GAN) architecture. The LSTM predictor takes the current and past observations of a well-defined health index as an input, uses those to forecast the future degradation trajectory, and then derives the RUL. Our proposed approach has three unique features: (1) Defining the bearing failure threshold by adopting an International Organization of Standardization (ISO) standard, making the approach industry-relevant; (2) Employing a GAN-based data augmentation technique to improve the accuracy and robustness of RUL prediction in cases where the deep learning model has access to only a small amount of training data; (3) Integrating the training process of the LSTM predictor within the GAN architecture. A joint training approach is utilized to ensure that the LSTM predictor model learns both the original and artificially generated data to capture the degradation trajectories. We utilize a publicly available accelerated run-to-failure dataset of rolling element bearings to assess the performance of the proposed approach. Results of a five-fold cross-validation study show that the integration of the LSTM predictor with GAN helps to decrease the average RUL prediction error by 29% over a simple LSTM model without GAN implementation. 
    more » « less
  4. ; ; ; ; ( , Optics Express)

    Images captured from a long distance suffer from dynamic image distortion due to turbulent flow of air cells with random temperatures, and thus refractive indices. This phenomenon, known as image dancing, is commonly characterized by its refractive-index structure constantCn2as a measure of the turbulence strength. For many applications such as atmospheric forecast model, long-range/astronomy imaging, and aviation safety, optical communication technology,Cn2estimation is critical for accurately sensing the turbulent environment. Previous methods forCn2estimation include estimation from meteorological data (temperature, relative humidity, wind shear, etc.) for single-point measurements, two-ended pathlength measurements from optical scintillometer for path-averagedCn2, and more recently estimatingCn2from passive video cameras for low cost and hardware complexity. In this paper, we present a comparative analysis of classical image gradient methods forCn2estimation and modern deep learning-based methods leveraging convolutional neural networks. To enable this, we collect a dataset of video capture along with reference scintillometer measurements for ground truth, and we release this unique dataset to the scientific community. We observe that deep learning methods can achieve higher accuracy when trained on similar data, but suffer from generalization errors to other, unseen imagery as compared to classical methods. To overcome this trade-off, we present a novel physics-based network architecture that combines learned convolutional layers with a differentiable image gradient method that maintains high accuracy while being generalizable across image datasets.

     
    more » « less
  5. ; ; ; ( , IEEE Transactions on Computers)
    In this work, we present a novel approach to real-time tracking of full-chip heatmaps for commercial off-the-shelf microprocessors based on machine-learning. The proposed post-silicon approach, named RealMaps, only uses the existing embedded temperature sensors and workload-independent utilization information, which are available in real-time. Moreover, RealMaps does not require any knowledge of the proprietary design details or manufacturing process-specific information of the chip. Consequently, the methods presented in this work can be implemented by either the original chip manufacturer or a third party alike, and is aimed at supplementing, rather than substituting, the temperature data sensed from the existing embedded sensors. The new approach starts with offline acquisition of accurate spatial and temporal heatmaps using an infrared thermal imaging setup while nominal working conditions are maintained on the chip. To build the dynamic thermal model, a temporal-aware long-short-term-memory (LSTM) neutral network is trained with system-level features such as chip frequency, instruction counts, and other high-level performance metrics as inputs. Instead of a pixel-wise heatmap estimation, we perform 2D spatial discrete cosine transformation (DCT) on the heatmaps so that they can be expressed with just a few dominant DCT coefficients. This allows for the model to be built to estimate just the dominant spatial features of the 2D heatmaps, rather than the entire heatmap images, making it significantly more efficient. Experimental results from two commercial chips show that RealMaps can estimate the full-chip heatmaps with 0.9C and 1.2C root-mean-square-error respectively and take only 0.4ms for each inference which suits well for real-time use. Compared to the state of the art pre-silicon approach, RealMaps shows similar accuracy, but with much less computational cost. 
    more » « less
  • Citation Formats
  • MLA
  • APA
  • Chicago
  • BibTeX
  • Save / Share this Record
  • Send to Email