Across decades of automating, operating, and optimising process units, our founding engineers saw the need for an early event-detection system to detect and advise console operators and unit-operations personnel about events requiring intervention. The foundation of this expert system required real-time process sensor data validation software to assure sensor and controller data was accurate as it supported the operations advisory system. The use of signal-characterisation rules and models enabled this Sensor & Controller Diagnostics module to detect faulty data in less than a minute.
Plant sites today have a data-quality gap: a layer between the historian and the AI pipeline that nobody owns. OT teams assure the availability of the data historian app while disregarding the quality of the data. Instrument & Control technicians assume instruments are fine until a work order is initiated. Process engineers assume the historian data is fine. Data scientists assume both. Data reconciliation forces model closure to normalise the data errors while enabling the models to continue performing in spite of faulty inputs. Tier 1 assets receive additional focus at a macro level while the sensors for the balance of plant are ignored.
VALID8A.AI exists to be that layer — engineered specifically for process sensor data, by people who have been in plant operations. It is the layer we wished we’d had for our many touchpoints and workflows requiring accurate process data. It’s built so the next generation of industrial AI runs on inputs an operator, a process engineer, and a manager can all stand behind.