Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

In-Situ Model Validation for Continuous Processes Using In-Network Computing

Ike Kunze, Dominik Scheurenberg, Liam Tirpitz, Sandra Geisler, Klaus Wehrle

TL;DR

CIVIC addresses the need for continuous, in-situ validation of industrial process models used in control schemes like MPC by leveraging in-network computing on programmable data planes. It combines a data collection unit that extracts instantaneous and historical process signals with modular model validation that compares observed behavior against known dependencies, classifying states into normal, warning, or error. Implemented on an Intel Tofino switch and demonstrated on a lab-scale coupled-tank water treatment plant, CIVIC achieves high detection accuracy for faults such as clogged pipes and failing pumps, and supports fast, edge-driven responses or reconfigurations. The work highlights the practicality and benefits of edge-based, lightweight model validation for real-time industrial process health monitoring and potential model refinement.

Abstract

The advancing industrial digitalization enables evolved process control schemes that rely on accurate models learned through data-driven approaches. While they provide high control performance and are robust to smaller deviations, a larger change in process behavior can pose significant challenges, in the worst case even leading to a damaged process plant. Hence, it is important to frequently assess the fit between the model and the actual process behavior. As the number of controlled processes and associated data volumes increase, the need for lightweight and fast reacting assessment solutions also increases. In this paper, we propose CIVIC, an in-network computing-based solution for Continuous In-situ Validation of Industrial Control models. In short, CIVIC monitors relevant process variables and detects different process states through comparison with a priori knowledge about the desired process behavior. This detection can then be leveraged to, e.g., shut down the process or trigger a reconfiguration. We prototype CIVIC on an Intel Tofino-based switch and apply it to a lab-scale water treatment plant. Our results show that we can achieve a high detection accuracy, proving that such monitoring systems are feasible and sensible.

In-Situ Model Validation for Continuous Processes Using In-Network Computing

TL;DR

CIVIC addresses the need for continuous, in-situ validation of industrial process models used in control schemes like MPC by leveraging in-network computing on programmable data planes. It combines a data collection unit that extracts instantaneous and historical process signals with modular model validation that compares observed behavior against known dependencies, classifying states into normal, warning, or error. Implemented on an Intel Tofino switch and demonstrated on a lab-scale coupled-tank water treatment plant, CIVIC achieves high detection accuracy for faults such as clogged pipes and failing pumps, and supports fast, edge-driven responses or reconfigurations. The work highlights the practicality and benefits of edge-based, lightweight model validation for real-time industrial process health monitoring and potential model refinement.

Abstract

The advancing industrial digitalization enables evolved process control schemes that rely on accurate models learned through data-driven approaches. While they provide high control performance and are robust to smaller deviations, a larger change in process behavior can pose significant challenges, in the worst case even leading to a damaged process plant. Hence, it is important to frequently assess the fit between the model and the actual process behavior. As the number of controlled processes and associated data volumes increase, the need for lightweight and fast reacting assessment solutions also increases. In this paper, we propose CIVIC, an in-network computing-based solution for Continuous In-situ Validation of Industrial Control models. In short, CIVIC monitors relevant process variables and detects different process states through comparison with a priori knowledge about the desired process behavior. This detection can then be leveraged to, e.g., shut down the process or trigger a reconfiguration. We prototype CIVIC on an Intel Tofino-based switch and apply it to a lab-scale water treatment plant. Our results show that we can achieve a high detection accuracy, proving that such monitoring systems are feasible and sensible.
Paper Structure (15 sections, 3 figures, 2 tables)

This paper contains 15 sections, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Controlling Continuous Industrial Processes
  3. Continuous In-situ Validation of Industrial Control models (CIVIC)
  4. Data Collection
  5. Model Validation
  6. Monitoring a Water Treatment Plant
  7. Water Treatment Plant
  8. Data Collection
  9. Model Validation
  10. Evaluation
  11. Methodology
  12. Identifying Clogged Pipes
  13. Detecting a Failing Pump
  14. Discussion
  15. Conclusion

Figures (3)

  • Figure 1: The coupled tank system uses pump P$_{1}$, and valves V$_{1}$/V$_{2}$ to control the water levels L$_{1}$-L$_{3}$. The manual valves V$_{M1}$/V$_{M2}$ can emulate faults.
  • Figure 2: CIVIC consists of two components: its data collection unit collects instantaneous and long-term process information which the model validation unit uses to assess the fit of the process to the given model.
  • Figure 3: Our prototype tracks instantaneous control data (P$_{1}$,V$_{1}$,V$_{2}$) and long-term information on L$_{1}$-L$_{4}$. We compare the long-term slopes against given references when the instantaneous data matches certain thresholds.