Field Hardened Machine Learning
Machine learning is the process by which a set of computer algorithms "learn" about a subject at hand. In the industrial setting this would be learning about how a manufactured product is best made, or how a manufacturing process is best run, the causes of problems, predicting the need for maintenance and so forth. This includes understanding and handling all the idiosyncrasies of each process and product, as two processes side by side built from the same "blue prints" at the same time are each different and behaviorally diverge over time.
In order to help our industrial customers improve their performance, reduce costs, increase yields and so forth, we have learned the ins-and-outs of creating robust, practical and performant machine learning technologies that work. We have created the tools that properly handle data, both on the desktop in bulk, but also streaming in real-time. We've created server technology to put these solutions on-line and close the loop with process control systems, and not merely sending a setpoint but the nuances of validity, validating, heart-beats, redundancy, retry, and all the other considerations a serious and proven solution must provide. We've created the means to drag-n-drop configure functionality to create the diverse solutions necessary to serve our diverse industrial customer base. This is not trivial. We've delivered and supported these solutions and fine tuned them to handle all the oddities and exceptions one encounters in real life; missing data, invalid data, catching up when data re-appears and more. We've even created autonomously recalibrating machine learning techologies that monitor their own performance and make adjustments automatically, without human intervention.
Example Characteristics of Our Industrial Strength Machine Learning Technology
- Heartbeat between our Intellect Server and One or More Distributed Control Systems (for availability monitoring and associated actions)
- Process State Detection (determining what state the process is in based on valves, registers, conditions such as temperatures and pressures)
- Sensor Availability Detection (is the sensor giving readings and taking appropriate actions if not)
- Sensor Validation (are the sensors' readings reasonable for the operating state of the process)
- Autonomous and Human-Driven Multi-Dimensional Non-Linear Predictive Regression
- Autonomous Rebuilding of Models Should Sensors Fail (can keep going even if sensors fail)
- Autonomous Rebuilding of Models Based on Performance Feedback
- Autonomous Recalibration (corrects for drift and non-stationary processes)
- Closed-Loop Cascade Control or Operator Guidance on Recommended Setpoints
- Autonomous Performance Reporting
When a customer says they are going to put our systems "on the control network", we know what that means and how to configure the system appropriately. To us, a "DCS", "SCADA" system or "MES" aren't just acronyms, but a way of life and familiar to us. We've even created a divsion specifically for this; IntelliDynamics.
We are well beyond most machine learning solutions providers, having a decade's head start, that when you use our technologies, you are using field hardened, experienced and capable machine learning solutions.
BioComp Systems is run by engineers with many decades' experience in the industrial setting. We have experience in oil and gas, discrete parts assembly, continuous and batch chemical manufacturing, coating films and papers, paper making, and the list goes on.
Call Us at 1-281-760-4007
If you think this technology might be a good fit for something you are working on, a challenge you may have, then please do pick up the phone and call us at 1-281-760-4007. A technical expert will be happy to discuss your opportunity.