Industrial Setpoint Control With AI Safety Boundaries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing industrial process control systems rely on hardcoded rulesets or mathematical models, which do not learn or improve over time, and AI-based control systems are nondeterministic, posing safety risks in mission-critical systems.
Innovation Solution
A multi-layered safety architecture that uses an AI agent to determine setpoint values within defined safety boundaries, combined with a setpoint source selector and a limit module to ensure deterministic and explainable control actions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If AI-based control systems are used to predict control parameters, then system performance and adaptability are improved, but safety and reliability deteriorate due to nondeterministic and unpredictable actions
Solution Approach 1:
The control system is segmented into multiple independent modules: AI-based predictive control module, traditional deterministic control module (e.g., PID), and a switching mechanism. Each module operates independently with defined responsibilities, allowing the system to leverage AI's adaptability while maintaining deterministic safety through traditional control modules that can be activated when AI predictions fall outside acceptable ranges.
Solution Approach 2:
A switching mechanism acts as an intermediary between the AI-based predictive control module and the traditional deterministic control module. This intermediary monitors AI predictions, compares them against safety boundaries and performance thresholds, and selectively activates either the AI module or traditional control module based on current system state, thereby ensuring safety while preserving AI's performance benefits.
2Speed
If local AI-based control systems are deployed, then response time and autonomy are improved, but maintenance complexity and operational difficulty worsen due to requiring AI engineers to travel and develop onsite
Solution Approach 1:
The control system employs dynamic switching between local AI-based control and remote traditional control modes. The system can adapt its operational mode based on network availability, performance requirements, and maintenance needs, allowing flexible deployment where AI processing occurs locally for speed while maintenance and updates can be performed remotely through the switching mechanism.
Solution Approach 2:
The switching mechanism serves multiple functions: it acts as a safety boundary enforcer, a performance monitor, a network failure detector, and a maintenance coordination interface. This multi-functional component enables the system to handle both real-time control decisions and long-term maintenance operations through a single integrated architecture.
3Ease of operation
If remote AI-based control systems are used, then ease of maintenance and operation are improved, but reliability worsens due to network issues such as outages and latency that leave the industrial system without control instructions
Solution Approach 1:
The system implements beforehand cushioning by maintaining local AI models and traditional deterministic control algorithms as backup control capabilities at the industrial site. When network connectivity is available, the system operates in remote mode for ease of maintenance. When network outages or latency occur, the pre-positioned local models and traditional controllers provide immediate fallback control, ensuring continuous operation without interruption.
4Reliability
If hardcoded rulesets or mathematical models are used for control, then safety and determinism are improved, but adaptability and performance improvement over time worsen due to inability to learn or improve
Solution Approach 1:
The system implements feedback mechanisms where AI-based predictive control continuously learns from historical operational data and performance outcomes, while traditional deterministic control provides stable baseline performance. The switching mechanism monitors which control approach yields better performance for specific operational conditions, creating a feedback loop that gradually improves overall system performance while maintaining deterministic safety through the traditional control fallback.
Data Source
AI summary
In variants, a method for industrial process control can include: determining AI setpoint values using an AI agent; determining local setpoint values using a local control system; selecting a setpoint source from a set of candidate setpoint sources including the AI agent and the local control system; optionally determining a set of transition setpoint values based on setpoints provided by the setpoint source; and limiting the setpoint values; wherein the industrial system is operated based on the limited setpoint values.