Physics-Enhanced Graph Neural Networks For Soft Sensing in Industrial Internet of Things

TL;DR

Graph neural networks are proposed, which are physics-enhanced GNNs, which integrate principles of physics into graph-based methodologies by augmenting additional nodes in the input graph derived from the underlying characteristics of the physical processes.

Abstract

The Industrial Internet of Things (IIoT) is reshaping manufacturing, industrial processes, and infrastructure management. By fostering new levels of automation, efficiency, and predictive maintenance, IIoT is transforming traditional industries into intelligent, seamlessly interconnected ecosystems. However, achieving highly reliable IIoT can be hindered by factors such as the cost of installing large numbers of sensors, limitations in retrofitting existing systems with sensors, or harsh environmental conditions that may make sensor installation impractical. Soft (virtual) sensing leverages mathematical models to estimate variables from physical sensor data, offering a solution to these challenges. Data-driven and physics-based modeling are the two main methodologies widely used for soft sensing. The choice between these strategies depends on the complexity of the underlying system, with the data-driven approach often being preferred when the physics-based inference models are intricate and present challenges for state estimation. However, conventional deep learning models are typically hindered by their inability to explicitly represent the complex interactions among various sensors. To address this limitation, we adopt Graph Neural Networks (GNNs), renowned for their ability to effectively capture the complex relationships between sensor measurements. In this research, we propose physics-enhanced GNNs, which integrate principles of physics into graph-based methodologies. This is achieved by augmenting additional nodes in the input graph derived from the underlying characteristics of the physical processes. Our evaluation of the proposed methodology on the case study of district heating networks reveals significant improvements over purely data-driven GNNs, even in the presence of noise and parameter inaccuracies.

Figures (10)

• Figure 1: Overview of the proposed Physics-Enhanced GNN architecture. In the absence of measurements in certain areas, a subset of available physical sensors is utilized to compute supplementary physical information based on component characteristics. After the preprocessing step, a graph using both physical sensor data and the calculated information is constructed. This graph is then fed into two graph layers with a skip connection, followed by an MLP for virtual sensor estimation.
• Figure 2: Graphs with augmented nodes, encompassing both physical mass flow rate sensors and physics-based information
• Figure 3: Pipe with its physical parameters.
• Figure 4: Heating demand and outside temperature patterns
• Figure 5: (a) Overview of the simulated district heating network, (b) Layout of the network simulation.