Situational Overview
Autonomous Advisory Systems (AAS) is a relatively new terminology for the process industries. To understand the terminology and thus to create the background understanding and familiarity for applications, Dominus advises certain key definitions and clarifications.
AAS vs. Closed Loop Controls & RAG
Autonomous Advisory System (AAS): An isolated, human in the loop intelligence platform that acts purely as an advisor. It calculates unmeasured process values and optimized setpoint suggestions but remains mechanically incapable of injecting changes into the control loops directly. It is an AI model grounded by physics: it cannot hallucinate, it learns plant behavior, and it corrects human mistakes and errors.
Retrieval Augmented Generation (RAG): A design method that prevents AI hallucinations by anchoring a localized 8B Small Language Model (SLM) strictly to deterministic first principles physics equations and local unstructured text data repositories (P&IDs, Standard Operating Procedures, maintenance shift records).
How the Dominus AAS Works
It Reads the Plant: Plant data and documents cross a one way data diode into the workstation in the industrial DMZ.
It Estimates and Forecasts: Physics is the basis. A time series model forecasts, physics leads when the two disagree, and physics answers every what if.
It Answers From Documents and Logic: When IRIS answers, the plant’s own logic decides and the language model only translates. When the documents and data do not hold the answer, she says so. She never invents a setpoint or procedure.
It Keeps Score: Every answer is stored and later checked against what the plant recorded, and the models adapt to each site.
Technical Interconnect Architecture & Operator Interfaces
OPC Data Interface: The proxy software streams real time 1 second interval tag telemetry packets directly from the host OPC Server into the hardware diode’s Transmit/Receive (TX/Rx) node inside the secure instrument cabinet room.
Edge RAG Workstation: A hardened on premises 4U rackmount server featuring 24 Core CPU blocks, a 48GB VRAM NVIDIA Ada GPU, and partitioned RAID 1 storage configurations to pin model weights and time series data caches.
Operator & Engineer Experience: System displays function on two 27 inch rugged touchscreens side by side with the main consoles. Typical query interactions evaluate tower levels or compressor curves, returning plain language, 3 part predictive advisory steps.
On Premises AI Advisory & Retrieval Augmented Generation (RAG) Platform
Defines the True Computational Architecture: Incorporating “Retrieval Augmented Generation (RAG)” clarifies to automation engineers exactly how the system works. It communicates that the system isn’t blindly guessing process adjustments based on generic text weights. Instead, it is dynamically binding real time 1 second streaming telemetry with static, highly technical facility files (P&IDs, heat & material balances, and flow curves) to ground its mathematical deductions in verified physical realities.
The Dominus Position?
The more AI specific AAS knowledge that process industries can engage and embrace, the more efficient and optimal the solution.
Dominus Industries is currently headquartered in Atlanta but will relocate to Houston in early 2027.