Fast Data-Driven Modeling of Hydraulic Clutch Control Pressure with Latch-State Classification and Gaussian Process Regression

Jul 16, 2026

Upload by Levi

Data-driven modeling of hydraulic clutch control pressure is presented for a circuit with a variable-force solenoid, accumulator, pressure regulator valve, and latch valve. The system shows nonlinearities driven by hysteresis, latch transitions, and actuator dynamics. A baseline regression using commanded current captures the overall response but misses hysteresis and latch behavior. The input is extended with current-derivative features, then latch regimes are separated using classifiers and modeled via Gaussian Process regression partitions. Nonlinear SVC and gradient boosting achieve the best latch classification, and the selected SVC-based local regression matches measured pressure and hysteresis better than a physics-based Amesim model on unseen ramp-rate data.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages8

Reading Time20 min

Introduction

Background

Modeling Process

Hydraulic Subsystem and Test Data

Results and Conclusions

FAST DATA-DRIVEN MODELING OF HYDRAULIC CLUTCH CONTROL PRESSURE WITH LATCH-STATE CLASSIFICATION AND GAUSSIAN PROCESS REGRESSION∗ Yash Bagla Graduate Controls Engineer Drive System Design Farmington Hills, Michigan, 48335 [yashbagla321@gmail.com](yashbagla321@gmail.com) Jason Schneider Senior Controls Engineer Drive System Design Farmington Hills, Michigan, 48335 [Jason.Schneider@drivesystemdesign.com](Jason.Schneider@drivesystemdesign.com) arXiv :2607 . 10477v1 [ ee ss . SY] 11 Jul 2026 May 2020 ABSTRACT This paper presents a data-driven method for modeling the pressure response of a hydraulic clutch control circuit. The system consists of a variable-force solenoid, accumulator, pressure regulator valve, and latch valve, and exhibits nonlinear behavior caused by hysteresis, latch transitions, and actuator dynamics. A baseline model using commanded current variables captured the general pressure response but failed to represent hysteresis and latch behavior accurately. The input vector was therefore extended with current derivative information, and several classifiers were tested to separate latch-related operating regimes before fitting Gaussian Process regression models to the resulting partitions. Nonlinear SVC and gradient boosting produced the highest latch-classification accuracy, and nonlinear SVC was selected for the final local-regression pipeline. The proposed approach was evaluated on unseen ramp-rate data and compared against a physics-based Amesim model. The machine-learning model reproduced the measured pressure response and hysteresis behavior more accurately than the physics-based simulation for the tested operating conditions. These results suggest that machine-learning plant models can complement physics-based hydraulic models during hardware development and controller calibration when representative test-stand data are available. Keywords Hydraulic Control · Clutch Pressure Modeling · Machine Learning · Gaussian Process Regression · Support Vector Classification 1 Introduction Modern hydraulic control systems are difficult to model accurately because their measured behavior depends on coupled mechanical, hydraulic, and actuator dynamics. In clutch control circuits, effects such as valve friction, hysteresis, latch transitions, dither settings, and pressure-regulator dynamics can cause the same nominal current command to produce different pressure responses under different operating histories. These nonlinearities make calibration and plant-model development expensive when the model must be tuned manually. Physics-based hydraulic simulation remains valuable during early design because it can be built before hardware is available. However, high-fidelity correlation can require difficult tuning of lumped parameters such as flow friction and jet-force coefficients, and some local nonlinear effects are hard to capture without extensive test data [1, 2, 3] . Once hardware data are available, a learned plant model can provide a complementary path: it can use measured input-output behavior to reproduce the system response quickly enough for calibration, control development, and sensitivity studies. ∗This work was accepted to the 14th CTI Symposium and Exhibition,“Automotive Drivetrains, Intelligent, Electrified,” scheduled for May 13-14, 2020 in Novi, Michigan, USA. This study focuses on a hydraulic clutch control circuit within a valve body for an automatic transmission. The measured output is clutch pressure, and the primary input is the current applied to a variable-force solenoid (VFS), including dither amplitude and dither frequency. The objective is to learn a pressure model that captures the dominant nonlinear behaviors observed in test-stand data, especially hysteresis and latch behavior. The contribution of this paper is a staged modeling process for this hydraulic subsystem. First, a baseline regression model is trained using commanded current variables. Second, current deri

FAQs

Why is modeling hydraulic clutch control pressure difficult?

Because valve friction, hysteresis, latch transitions, dither settings, and pressure-regulator dynamics cause the same nominal current command to yield different pressure responses depending on operating history.

How does the method improve over a baseline current-command regression?

It extends the input vector with current-derivative information to encode direction-dependent hysteresis behavior, enabling more accurate pressure prediction.

How are latch-related behaviors handled in the proposed pipeline?

Latch operating regimes are identified using classification methods, and separate Gaussian Process regression models are trained for each resulting data partition before evaluation on unseen ramp-rate data.

Jul 16, 2026

Upload by Levi