Development and evaluation of modeling approaches for transients in centrifugal chillers
Abstract
Dynamic modeling of vapor compression equipment is gaining increasing attention in recent years. The availability of cheap and powerful computers is simplifying the heretofore formidable challenge of numerically solving the non-linear differential equations that arise in first-principle based system models. Recent advances in the HVAC industry such as fault detection and diagnostics (FDD) require, for the design and validation of algorithms, extensive chiller data with and without faults. Faulty performance data from chillers is extremely difficult to get from commercial installations and therefore models are necessary. This requirement necessitates the use of first-principles since the fault and its severity need to be varied in a physically meaningful manner to produce useful data. First-principles models involve the solution of transient mass, energy and momentum balances which are a system of non-linear, coupled, partial differential equations. Limited study has been done in the past on identifying efficient methods of formulating and solving these systems for centrifugal liquid chillers. Existing system models are either not for centrifugal compressors or do not contain enough detail in the heat exchangers to allow fault simulation. Also, little information is provided regarding the execution speed and accuracy of different modeling approaches. This thesis includes a comparison of the finite-volume and the moving-boundary formulations for modeling dynamics of shell-and-tube heat-exchangers. The finite volume models performance is investigated using direct solution methods with explicit integration and a sequential solution method with implicit integration. The moving boundary model is studied using a direct solution method and a state-space formulation. The goal is to identify the computationally most efficient approach for future model developers. Validation results with and without faults are presented, using data obtained from a 90-ton centrifugal liquid chiller test-stand available at Herrick Labs, Purdue University. It is found that the direct solution methods are preferable for both the formulations. The sequential solution with implicit integration is less robust in convergence than the direct solution when the finite volume formulation is used. The state-space solution when using the moving boundary formulation is restricted to small transients due to the linearization approximation. Between the formulations, the finite volume tolerates a wider range of inlet conditions than the latter and is preferable when start-up is needed. The moving boundary approach executes twice as fast as the finite volume approach for comparable accuracy.
Degree
Ph.D.
Advisors
Braun, Purdue University.
Subject Area
Mechanical engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.