01 / APPLICATION
02 / INNOVATION
03 / CONTROL EXPERTISE
Applied chemical engineering software for refinery units, petrochemical plants, gas processing and NGL Fractionation.
Machine-Learning for effective automation and stability of plant operations for direct real-time control of plant valves and actuators.
The PID (Proportional-Integral-Derivative) method of control is the fundamental building block of industrial automation. Eigen technology is a step-change in simplicity and ease of integrating machine learning with direct PID control.
Three Nested PID Control Loops
04 / AI's ABILITY TO PREDICT COMPLEX SYSTEMS
The animation demonstrates a simulation of three nested PID loops that quickly become unstable with very minute disturbances; the outcome is very hard to predict with simple PID tuning. That is known as the Butterfly Effect: even the smallest perturbations to a complex system can propagate into a dramatically divergent future.
Under the effect of process disturbances, an MPC controller alone is usually ineffective to regain control without manual intervention by the control room operator.
Eigen Control technology aids the MPC Controller to maintain control by making use of novel machine learning models to predict the chaotic evolution of PID systems at energy or industrial plants.
Machine learning has been demonstrated to have an amazing ability to predict evolutions of chaotic systems.
"Chaos theory suggests that even in a deterministic system, if the equations describing its behavior are nonlinear, a tiny change in the initial conditions can lead to a cataclysmic and unpredictable result."
Our story was born out of the Houston tech ecosystem with a game-changing product born out of the intersection of several domains of expertise: chemical engineering, electrical engineering, applied mathematics, and machine-learning.
Our vision is to avail our game-changing process control technology to process control engineers across the globe.
The equivalent CO2 emissions of 7.4 million cars are eliminated if our technology were to be implemented at every distillation process plant in the U.S.
Significant increases in plant capacity, increases in product purity/quality, and reduction of energy and fuel use are the three pillars of value proposition to plant operators.
The essence of our technology is real-time control of a plant in a fully transparent and auditable way.
We achieve this by breaking open the base layer and linearizing the process or mathematically speaking, finding the eigenfunctions of every non-linearity embedded in the process, in a piecewise method, from the MPC layer to the valve positioner. The effect is a reduction of degrees of freedom in an existing control system.
In some sections of the piecewise architecture, we employ deep learning to find the eigenvalues and eigenvectors.
Our methodology of piecewise eigenfunctions dramatically increases the performance of an existing MPC controller or a base layer system without one.