Deep learning-based position detection for hydraulic cylinders using scattering parameters

Position detection of hydraulic cylinder pistons is crucial for numerous industrial automation applications. A typical traditional method to excite electromagnetic waves in the structure and analytically solve piston position based on scattering parameters measured by a sensor. The core this approach physical model that outlines relationship between targeted position. However, has shortcomings accuracy adaptability, especially extreme conditions. To address these limitations, we propose machine learning deep learning-based methods learn directly data-driven manner. As result, all models paper consistently outperform one large margin. We further deliberate choice domain knowledge provide in-depth analyses combining performance with real-world characteristics. Specifically, use Convolutional Neural Network (CNN) discover local interactions input among adjacent frequencies, apply Complex-Valued (CVNN) exploit complex-valued nature parameters, introduce novel technique named Frequency Encoding add weighted frequency information input. combination techniques results our best-performing model, CNN Encoding, which exhibits substantial improvement an error reduction 1/12 compared model.