HANDS-ON APPROACH ON DEVELOPING A DEEP LEARNING ALGORITHM FOR STATE CLASSIFICATION OF A HYDRAULIC ACCUMULATOR

Abstract

Hydro-pneumatic accumulators are essential components in fluid power systems, serving various purposes like dampening pulsations, stabilizing flow, and ensuring safety.Monitoring their energy state is challenging due to gas leakage, requiring knowledge of gas pressure, temperature, and volume.Real-time measurements of gas temperature and volume are difficult due to transient changes during operation.This paper introduces a novel approach to classify gas mass in bladder pressure accumulators using Deep-Learning, particularly long short-term memory (LSTM) networks.The study aims to classify load situations with minimal sensor data, providing insights into the workflow for classifying multivariate time-series data with deep neural networks.Therefore, a bladder accumulator is equipped with sensors, its dynamic behaviour is recorded and used to train and validate the LSTM network's ability to classify the gas amount inside the accumulator.This method offers a promising way to determine the pre-charge pressure, a crucial parameter for assessing the accumulator's energy state.

Authors 3

1. Faras Brumand‐Poor Aachen Institute for Fluid Power Drives and Systems ifas

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen University , Institute for Fluid Power Drives and Systems (ifas

   RWTH Aachen University, Institute for Fluid Power Drives and Systems (ifas

2. Peter Kloft
   Affiliation as printed

   Hydac Technology GmbH , Germany

   Hydac Technology GmbH, Germany

3. Katharina Schmitz Aachen Institute for Fluid Power Drives and Systems ifas

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen University , Institute for Fluid Power Drives and Systems (ifas

   RWTH Aachen University, Institute for Fluid Power Drives and Systems (ifas