Soft Sensor

What is Soft Sensors

Soft-sensors: predictive models for sensor characteristic are called soft sensors Soft-sensors: software+sensor

Soft-sensor Categories

  • model-driven
    • First Principle Models (FPM)
    • extended Kalman Filter
  • data-driven
    • Principle Component Analysis + regression model, Partial Least Squares
    • Artificial Neural Networks
    • Neuro-Fuzzy Systems
    • Support Vector Machines

Soft-sensor Application

  • on-line prediction
    • prediction of process variables which can be determined either at low sampling rates
    • prediction of process variables which can be determined through off-line analysis only
    • (statistical or soft computing supervised learning approaches)
  • process monitoring
    • detection of the state of the process, usually by human
    • observation and interpretation of the process state (based on univariate statistics) and experience of the operator
  • process fault detection
    • detection of the state of the process

FPM

  • First Principle Models describe the physical and chemical background of the process.
  • These models are developed primarily for the planning and design of the processing plants, and therefore usually focus on the description of the ideal steady-states of the processes
  • based on established laws of physics
  • does not make assumptions such as empirical model and fitting parameters
  • using experimental data

Data-driven data-driven models are based on the data measured within the processing plants, and thus describe the real process conditions, they are, compared to the model-driven Soft Sensors, more reality related and describe the true conditions of the process in a better way. Nevertheless

The most commonly applied multivariate analysis tools are principal component analysis (PCA) for fault detection and projection of latent structures (PLS) for the prediction of key quality parameters at end of batch.

First-principle models may be the answer, using experimental data instead of statistical methods to estimate model parameters. They are not as quick and easy to build, but they have many advantages. In terms of simulation, first-principle models provide extrapolation in addition to the interpolation provided by data-driven models. But they also can be used for monitoring, control and optimization.

Soft-Sensor Modelling

Others

Dataset

Reference

Ebook