Data-Driven Plasticity Modeling via Acoustic Profiling

Khalid El-Awady

cs.LGMar 26, 2026

Data-Driven Plasticity Modeling via Acoustic Profiling

Authors: Khalid El-Awady

Abstract

This paper presents a data-driven framework for modeling plastic deformation in crystalline metals through acoustic emission (AE) analysis. Building on experimental data from compressive loading of nickel micropillars, the study introduces a wavelet-based method using Morlet transforms to detect AE events across distinct frequency bands, enabling identification of both large and previously overlooked small-scale events. The detected events are validated against stress-drop dynamics, demonstrating strong physical consistency and revealing a relationship between AE energy release and strain evolution, including the onset of increased strain rate following major events. Leveraging labeled datasets of events and non-events, the work applies machine learning techniques, showing that engineered time and frequency domain features significantly outperform raw signal classifiers, and identifies key discriminative features such as RMS amplitude, zero crossing rate, and spectral centroid. Finally, clustering analysis uncovers four distinct AE event archetypes corresponding to different deformation mechanisms, highlighting the potential for transitioning from retrospective analysis to predictive modeling of material behavior using acoustic signals.

Figures & tables

Figure 1: Experimental setup for the original data acquisition (left) and acquired acoustic emission plot and its magnitude spectrum (right).

Figure 2: Magnitude spectrum of the signal after apply a band-pass filter in the frequency range 2KHz to 60 KHz. The spectrum exhibits some prominent peaks after filtering in the vicinities of 8KHz, 16KHz, 25 KHz, and 44 KHz.

Figure 3: Visual representation of identified events in each of the frequency bands. On the left the gray central plot is the filtered waverform. The blue dots on the blue line indicate times where the onset of an AE event in the 8KHz band is detected. The orange, green, and red dots similarly represent the same for the 16KHz, 25KHz, and 44KHz events respectively. On the right these are shown in histogram form as well.

Figure 4: Average magnitude spectra for identified events in each of the frequency bands.

Figure 5: Correlation of the AE events with drops in the stress curve. The AE events line up well with the stress drops.

Figure 6: Strain and cumulative AE energy released over the experiment duration. We see the onset of increasing strain rate following a large AE event.

Figure 7: Confusion matrices on the test set for naive KNN (k=4) and SVM (with PCA) on the time domain signals.

Figure 8: Scatter plot of events and non-events in the 2D feature spaces of RMS vs ZCR an RMS vs spectral centroid. The plots show good clustering and separation of events and non-events along these features.

Figure 9: Identified clusters using a variety of time and frequency features. In the top figures we identify 4 clusters via a Gaussian Mixture Model and visualize them in 2 dimensions. In the lower left we show sample waveforms in each cluster. In the lower right we overlay the histogram of cluster occurrences over time.

Cluster	Count	RMS	Peak abs	Energy	Kurtosis	ZCR	Spec. Centroid	Spec. Entropy	Median Duration	Median $d\sigma/dt$
						(cross/sample)	(KHz)		( $\mu$ s)
1	105	0.112	0.340	4.995	4.166	0.032	51.5	2.75	207	-0.134
0	80	0.0750	0.171	1.585	2.654	0.025	106.1	3.05	116	-0.0277
3	73	0.0639	0.138	0.915	2.395	0.034	86.8	2.70	110	0.00816
2	8	0.0569	0.109	0.693	2.004	0.022	61.1	2.40	105	0.0235

Table 1: Cluster summary (sorted by number of events). Values are medians within each cluster. ZCR is reported in crossings/sample, spectral centroid in KHz, and duration in

\mu

Figure 10: Identified clusters using a variety of time and frequency features. In the top figures we identify 4 clusters via a Gaussian Mixture Model and visualize them in 2 dimensions. In the lower left we show sample waveforms in each cluster. In the lower right we overlay the his

Figure 11: Identified clusters using a variety of time and frequency features. In the top figures we identify 4 clusters via a Gaussian Mixture Model and visualize them in 2 dimensions. In the lower left we show sample waveforms in each cluster. In the lower right we overlay the his

Feature	Before Transition	After Transition
Event rate	Moderate	Pronounced increase
AE energy rate	Small	Large increase
Spectral centroid	More high-frequency activity	Lower frequencies dominate initial transition
Spectral entropy	Some spectral concentration of energy	More broadband activity

Table 2: Qualitative comparison of acoustic emission (AE) characteristics before and after the onset of accelerated strain.

Figure 12: Frequency of clusters before and after the strain regime change.

Figure 13: AE events over time as they occur.

Materials Science

Materials science studies how composition, structure, processing, and defects determine material properties and how to design useful materials.

See all 301 papers

Predicting Stress-strain Behaviors of Additively Manufactured Materials via Loss-based and Activation-based Physics-informed Machine Learning
Mar 15, 2026Chenglong Duan, Dazhong WuMaterials Property Prediction Physics-Informed ML
Department of Mechanical and Aerospace Engineering, College of Engineering and Computer Science, University of Central Florida, Orlando, FL 32816, USA
Strain in Sound: Soft Corrugated Tube for Local Strain Sensing with Acoustic Resonance
Apr 21, 2026Michael Chun, Ananya Nukala, Tae Myung Huh
All authors are with Dept. of Electrical and Computer Engineering, University of California Santa Cruz, Santa Cruz, CA, USA
CP-Agent: A Harness-Engineered Agent for Crystal Plasticity Simulation Workflows
Sep 25, 2026Samuel Onimpa Alfred, Abhishek Kumar, Veera SundararaghavanScientific Workflow Automation Tool-Using Agents
Department of Aerospace Engineering, University of Michigan, Ann Arbor, 48109, MI, USA

Mar 15, 2026J/K move · Enter open · S save

Predicting Stress-strain Behaviors of Additively Manufactured Materials via Loss-based and Activation-based Physics-informed Machine Learning

Chenglong Duan, Dazhong Wu

Department of Mechanical and Aerospace Engineering, College of Engineering and Computer Science, University of Central Florida, Orlando, FL 32816, USA

Materials Property Prediction Physics-Informed ML Additive Manufacturing

Open paper AskPDF arXiv

Predicting the stress-strain behaviors of additively manufactured materials is crucial for part qualification in additive manufacturing (AM). Conventional physics-based constitutive models often oversimplify material properties, while data-driven machine learning (ML) models often lack physical consistency and interpretability. To address these issues, we propose a physics-informed machine learning (PIML) framework to improve the predictive performance and physical consistency for predicting the stress-strain curves of additively manufactured polymers and metals. A polynomial regression model is used to predict the yield point from AM process parameters, then stress-strain curves are segmented into elastic and plastic regions. Two long short-term memory (LSTM) models are trained to predict two regions separately. For the elastic region, Hooke's law is embedded into the LSTM model for both polymer and metal. For the plastic region, Voce hardening law and Hollomon's law are embedded into the LSTM model for polymer and metal, respectively. The loss-based and activation-based PIML architectures are developed by embedding the physical laws into the loss and activation functions, respectively. The performance of the two PIML architectures are compared with two LSTM-based ML models, three additional ML models, and a physics-based constitutive model. These models are built on experimental data collected from two additively manufactured polymers (i.e., Nylon and carbon fiber-acrylonitrile butadiene styrene) and two additively manufactured metals (i.e., AlSi10Mg and Ti6Al4V). Experimental results demonstrate that two PIML architectures consistently outperform the other models. The segmental predictive model with activation-based PIML architecture achieves the lowest MAPE of 10.46+/-0.81% and the highest R^2 of 0.82+/-0.05 arocss four datasets.

Abstract

Materials Science

Explore similar work

Predicting Stress-strain Behaviors of Additively Manufactured Materials via Loss-based and Activation-based Physics-informed Machine Learning

Strain in Sound: Soft Corrugated Tube for Local Strain Sensing with Acoustic Resonance

CP-Agent: A Harness-Engineered Agent for Crystal Plasticity Simulation Workflows