China
Africa
Explore chapters and articles related to this topic
∞ Control of Nonlinear Processes Using Multiple Linear Models
Published in Roderick Murray-Smith, Tor Arne Johansen, Multiple Model Approaches to Modelling and Control, 2020
A. Banerjee, Y. Arkun, R. Pearson, B. Ogunnaike
Therefore the underlying premise of this chapter is that a controller has to be designed for a nonlinear system that operates in several significantly different modes. This makes it necessary to have a nonlinear model that accurately matches the plant behaviour in all operating regimes. The first principles models are usually difficult to develop if they have to cover a wide range of conditions. Even if such a global model can be obtained it may not be appropriate for controller design. The alternative is to identify an empirical model from plant input-output data. However unmodelled dynamics which are negligible at one operating point may be dominant at another. Therefore it may not be easy to select a model structure that works well in all regimes. Furthermore in order to uncover all the necessary plant dynamics, the inputs required by the identification algorithm may not be practically implementable due to their large amplitude and/or large frequency. Therefore this chapter presents an alternative approach wherein multiple local linear models are identified at the different regions of operation, and controller design is carried out using these models.
Process Identification
Published in Raghunathan Rengaswamy, Babji Srinivasan, Nirav Pravinbhai Bhatt, Process Control Fundamentals, 2020
Raghunathan Rengaswamy, Babji Srinivasan, Nirav Pravinbhai Bhatt
In Chapter 2, we discussed models for control. These models are derived from first-principles such as mass, energy, or momentum balance equations. In contrast to the first-principles models, data-driven or empirical models are derived from process input-output data. Typically, we are interested in developing transfer function models that describe relationships between process outputs and inputs (both manipulated and disturbances). First-principles models are useful for developing a good understanding of the underlying processes. However, in many cases, it might be difficult to develop first-principles models because a comprehensive understanding of the process might not exist. Moreover, unknown parameters in the first-principles models, if developed, have to be estimated from experimental data. These experiments are costly and the estimation process might be quite complex. Further, first-principles models are generally simplified to render them computationally tractable for control system analysis. As a result, one might resort to directly developing data-driven or empirical models from process data. The subject of developing models from data is called process or system identification [51, 93].
Systems for Interpretation and Diagnosis
Published in Adrian A. Hopgood, Intelligent Systems for Engineers and Scientists, 2021
Practically all physical devices are made up of fundamental components such as tubes, wires, batteries, and valves. As each of these performs a fairly simple role, it also has a simple failure mode. For example, a wire may break and fail to conduct electricity, a tube can spring a leak, a battery can lose its charge, and a valve may be blocked. Given a model of how these components operate and interact to form a device, faults can be diagnosed by determining the effects of local malfunctions on the global view, that is, on the overall device. Reasoning through consideration of the behavior of the components is sometimes termed reasoning from second principles. First principles are the basic laws of physics that determine component behavior.
Development and validation of a batch fluidized bed dryer model for pharmaceutical particles
Published in Drying Technology, 2021
Francis Gagnon, Jocelyn Bouchard, André Desbiens, Éric Poulin
The pharmaceutical production, and other sectors like food and mining, frequently relies on batch fluidized bed dryer (FBD) for their water removal needs in particulate materials. While they are efficient dryers, their operation in industry are typically fixed procedures, which is presumably suboptimal. Process modeling is one of the most versatile and promising tools to optimize production efficiency. In opposition to empirical approaches, starting from the first principles can potentially increase the applicability range. However, the growing complexity of modern industrial processes puts a curb on pure first principle developments. Phenomenological modeling is a good compromise in such circumstances.
A review of fault detection and diagnostics methods for building systems
Published in Science and Technology for the Built Environment, 2018
Woohyun Kim, Srinivas Katipamula
The second category under rule-based methods uses rules derived from first principles, which are implemented in a tree structure (Brambley et al. 2011; Fernandez et al. 2009; Wang et al. 2012a). A first principles-based model is typically generated by using laws governing system behavior, such as mass and energy balance. A steady-state algorithm is used to filter out transient data, because the first-principles method is based on steady-state operating conditions.