Deep learning-based prediction of time-resolved adhesive forces in viscoelastic Hertzian contacts

2026-07-21 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems · Depth: Expert, quick

Summary

A deep learning model has been developed to predict time-resolved adhesive forces in viscoelastic Hertzian contacts, addressing a significant challenge in soft robotics and manipulation tasks where traditional numerical simulations are computationally expensive. This scalar-conditioned, stateful, sequence-to-sequence model predicts full force evolution from displacement history across short- and long-range adhesion regimes. The training dataset covers four orders of magnitude in loading/unloading rates, varied dwell times, and Tabor parameters from 0.2 to 3.2. A fixed-measurement-step (FMS) representation was introduced to convert variable-length trajectories into fixed-length sequences while preserving physical-time information. Among tested architectures, an LSTM with concatenated conditioning performed best, achieving a held-out mean-squared error of 5.0e-4, a median pull-off-force error of ≈2.2%, and a median hysteresis error of ≈1.1%. This model provides complete force trajectories with a median inference time of 0.16 s, serving as a rapid surrogate for numerical evaluations.

Key takeaway

For Soft Robotics Engineers designing gripping and manipulation systems, this deep learning model offers a critical solution for real-time adhesive force prediction. Your reliance on computationally expensive numerical simulations can be replaced. This model predicts complete force trajectories in just 0.16 s, achieving high accuracy with ≈2.2% pull-off-force error. You should integrate this rapid surrogate model into control-oriented applications to accelerate design optimization and operational speed.

Key insights

A deep learning model accurately predicts time-resolved adhesive forces in viscoelastic contacts, enabling real-time soft robotics applications.

Principles

• Deep learning can surrogate complex physical simulations for real-time needs.
• Fixed-measurement-step (FMS) representation handles heterogeneous time scales.
• LSTM architectures excel in sequence-to-sequence force prediction.

Method

Train a scalar-conditioned, stateful, sequence-to-sequence deep learning model using FMS representation for variable-length trajectories, comparing LSTM, TCN, and time-distributed dense architectures with different conditioning mechanisms.

In practice

• Integrate into soft robotics for real-time gripping control.
• Accelerate design optimization loops for adhesive contacts.
• Replace computationally expensive numerical simulations.

Topics

• Deep Learning
• Soft Robotics
• Adhesive Forces
• Viscoelastic Hertzian Contacts
• Sequence-to-Sequence Models
• LSTM Networks

Best for: Research Scientist, AI Scientist, Machine Learning Engineer, Robotics Engineer

Related on AIssential

• Deep learning-based prediction of time-resolved adhesive forces in viscoelastic Hertzian contacts — Deep sequence models can rapidly predict full viscoelastic adhesive force trajectories, overcoming slow numerical simulations.
• There Will Be a Scientific Theory of Deep Learning — A scientific theory of deep learning, "learning mechanics," is emerging, focusing on dynamics, aggregate statistics, and quantitative predictions.
• Automating the Expert Eye: A System-Agnostic Deep Learning Framework for Rare Event Discovery in Imbalanced Force Spectroscopy — Deep learning automates rare event discovery in SMFS, overcoming extreme data imbalance with high accuracy and interpretability.
• Deep Learning Method for Stationary Distribution of Reflected Brownian Motion — The deep learning method accurately computes Laplace transforms for high-dimensional Reflected Brownian Motions, overcoming analytical intractability.
• Dense Force Estimation with an Event-based Optical Tactile Sensor — Event-based tactile sensors can reconstruct dense 3D force fields by mapping displacement data to forces via iFEM and neural networks.
• From Pixels to Newtons: Predicting In Vivo Joint Contact Forces from Monocular Video — A physics-free pipeline accurately predicts in vivo hip and knee contact forces from monocular video, matching traditional simulation.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.