Towards Energy Efficient Control for Commercial Heavy-Duty Mobile Cranes: Modeling Hydraulic Pressures using Machine Learning

TL;DR

This paper develops machine learning models to accurately predict hydraulic pressures in heavy-duty cranes, aiming to enhance energy efficiency and control systems in construction machinery.

Contribution

It introduces a generalized machine learning approach for modeling hydraulic pressures in complex cranes, incorporating detailed procedures for deriving key input variables from real-world data.

Findings

Models accurately predict actuator and pump pressures across various conditions

Demonstrates improved modeling accuracy using real crane data

Supports energy-efficient control strategies for heavy-duty machinery

Abstract

A sizable part of the fleet of heavy-duty machinery in the construction equipment industry uses the conventional valve-controlled load-sensing hydraulics. Rigorous climate actions towards reducing CO$_{2}$ emissions has sparked the development of solutions to lower the energy consumption and increase the productivity of the machines. One promising solution to having a better balance between energy and performance is to build accurate models (digital twins) of the real systems using data together with recent advances in machine learning/model-based optimization to improve the control systems. With a particular focus on real-world machines with multiple flow-controlled actuators and shared variable-displacement pumps, this paper presents a generalized machine learning approach to modeling the working pressure of the actuators and the overall pump pressures. The procedures for deriving reaction forces and flow rates as important input variables to the surrogate models are described in detail. Using data from a real loader crane testbed, we demonstrate training and validation of individual models, and showcase the accuracy of pressure predictions in five different experiments under various utilizations and pressure levels.