Abstract
This paper presents a formal framework for the design and analysis of industrial plant controllers using Temporal Logic (TL) and $\omega$-graphs. By representing plant dynamics as infinite state sequences, the authors enable automated control rule synthesis, stability verification, and causal reasoning. The methodology is demonstrated through a qualitative model of a boiler system, showing how complex temporal requirements can be reduced to graph-search problems.
Introduction
Designing controllers for complex industrial plants, such as boilers or power stations, has traditionally relied on tools like ladder diagrams or specialized programming languages. While effective for implementation, these methods lack a formal logical foundation, making it difficult to verify behavior, perform causal analysis, or automatically generate control rules. This research from Mitsubishi Electric Corporation proposes a formal system that treats "time" as a first-class citizen to address these challenges.
Methodology: Temporal Logic and $\omega$-Graphs
The proposed framework bridges logic and control engineering using two primary tools:
- Temporal Logic (TL): An extension of propositional logic with modal operators to describe the timing of events, including $\Box$ (Henceforth), $\Diamond$ (Eventually), $\circ$ (Next), and $\mathcal{U}$ (Until).
- The $\omega$-graph: A graphical representation of a temporal formula where nodes represent states and edges represent transitions. This graph acts as a finite state automaton for infinite sequences of states ($\omega$-languages).
By converting system dynamics into an $\omega$-graph, logical deduction is transformed into a computationally efficient graph-search problem. This "inner model" approach is more practical for engineering than traditional theorem-proving methods.
Application: The Boiler Control System
The methodology is demonstrated using a formal model of a boiler system. This model consists of five temperature states ($E_1$ through $E_5$), a fire state ($H$), and a controller that manages the system via "ON/OFF" signals based on temperature feedback.
The authors frame the synthesis of control rules as a logical problem: $Dynamics \land Unknown\_Rules \supset Goal$. By solving for the unknown rules within the $\omega$-graph, the controller's logic is discovered automatically.
Key Insights and Results
The framework provides clear logical templates for common engineering goals:
- Eventuality: Ensuring the system reaches a desired goal state.
- Stability: Ensuring the system remains in a safe state once reached.
- Observability: Verifying that the controller can accurately perceive the plant's state.
Key results include the successful derivation of control rules for the boiler example. For instance, the system can automatically deduce the specific conditions for toggling the fire ($H$) to maintain a stable temperature range. Furthermore, the $\omega$-graph allows for qualitative simulation (forward traversal) and causal arguments (backward traversal) to diagnose system failures.
Conclusion and Future Work
While the $\omega$-graph approach offers a powerful pathway for automating industrial system verification, challenges remain regarding scalability. The "state explosion" problem may occur as the number of variables increases in real-world plants. Additionally, the current discrete state model may lose precision compared to continuous process models. Future work involves further development of software tools to manipulate these graphs in live industrial environments.
Summary updated