Technology Soft Sensing
Cybernetica Soft Sensing applications combine mechanistic models with estimator algorithms to provide online estimation (soft sensing) of process variables that cannot be directly measured.
Advantages of Soft Sensor implementation
- Calculate process variables that are not possible to measure
- Continuous prediction of variables that are infrequently measured
- Rapid feedback to controllers that depend on slow process measurements
- Condition-based maintenance (CBM) based on estimated equipment state
- More precise and successful process control
- Early warning to operators of:
- Plant signal failures
- Process deviations
- Need for equipment maintenance
- Improved safety
Typical process variables measured by Soft Sensors
- Reaction rates for chemical reactions
- Product quality parameters
- Concentrations of raw materials, reaction products and other chemical species
- Conversion of chemical reactions
- Temperature and other thermodynamic properties
- Catalyst activity
- Heat transfer coefficients
We deliver soft sensors based on either existing models or models developed in cooperation with the customer. We will work together with you to develop, tune and implement process models that provide the most valuable information.
What is a Soft Sensor?
A soft sensor is a computer program that calculates a new sensor-like signal based on process information as well as other measurements and signals. Soft sensors are typically used to extract additional low-cost information from existing sensors or to measure variables that cannot be measured directly (e.g. due to hostile process environment or missing sensor technology).
Soft Sensing Cybernetica CENIT applications
Our soft sensing applications are based on the Cybernetica CENIT software suite, and the underlying models are implemented in a Cybernetica Model and Application Component. The CENIT implementation allows for seamless integration with both advisory and closed-loop controller algorithms.
Cybernetica Soft Sensors typically utilise advanced estimation algorithms, such as Kalman Filtering or Moving Horizon Estimation (MHE). The resulting estimator is then able to use live plant measurements to correct and adapt a physical process model. This strategy generates a data-driven and enhanced accuracy description of the current process state and is particularly valuable for processes with one or more unknown parameters.