Nonlinear Programming Applied to Calibrating Thermal and Fluid Models to Test Data (Semi-Therm 2002)

This paper describes readily available techniques for automating the search for worst-case (e.g., “hot case”, “cold case”) design scenarios using only modest computational resources. These methods not only streamline a repetitive yet crucial task, they usually produce better results.

The problems with prior approaches are summarized, then the improvements are demonstrated via a simplified example that is analyzed using various approaches. Finally, areas for further automation are outlined, including attacking the entire design problem at a higher-level.

In recent years, loop heat pipe (LHP) technology has transitioned from a developmental technology to one that is flight ready. The LHP is considered to be more robust than capillary pumped loops (CPL) because the LHP does not require any preconditioning of the system prior to application of the heat load, nor does its performance become unstable in the presence of two-phase fluid in the core of the evaporator. However, both devices have a lower limit on input power: below a certain power, the system may not start properly.

The loop heat pipe (LHP) is known to have a lower limit on input power. Below this limit the system may not start properly creating the potential for critical payload components to overheat. The LHP becomes especially susceptible to these low power start-up failures following diode operation, intentional shut-down of the device, or very cold conditions. These limits are affected by the presence of adverse tilt, mass on the evaporator, and noncondensible gas in the working fluid.

The NASA standard tool for thermohydraulic analysis, SINDA/FLUINT, includes thermodynamic and hydrodynamic solutions specifically targeted at the growing demand for design and analysis of liquid propulsion systems. Applications in this field have included:

This paper describes the application of the general purpose SINDA/FLUINT thermohydraulic analyzer to the modeling of vapor compression (VC) cycles such as those commonly used in automotive climate control and building HVAC systems. The software is able to simulate transient operation of vapor compression cycles, predicting pressures, coefficients of performance, and condenser/evaporator liquid positions in a closed two-phase system with a fixed fluid charge.

This paper describes a vehicle-level simulation model for climate control and powertrain cooling developed and currently utilized at GM. The tool was developed in response to GM's need to speed vehicle development for HVAC and powertrain cooling to meet world-class program execution timing (18 to 24 month vehicle development cycles). At the same time the simulation tool had to complement GM's strategy to move additional engineering responsibility to its HVAC suppliers. This simulation tool called "e-Thermal" was quickly developed and currently is in widespread (global) use across GM.

e-Thermal is a vehicle level thermal analysis tool developed by General Motors to simulate the transient performance of the entire vehicle HVAC and Powertrain cooling system. It is currently in widespread (global) use across GM. This paper discusses the details of the airconditioning module of e-Thermal. Most of the literature available on transient modeling of the air conditioning systems is based on finite difference approach that require large simulation times. This has been overcome by appropriately modeling the components using Sinda/Fluint.

Thermal analysis is typically performed using a point design approach, where a single model is analyzed one analysis case at a time. Changes to the system design are analyzed by updating the thermal radiation and conduction models by hand, which can become a bottleneck when attempting to adopt a concurrent engineering approach. This paper presents the parametric modeling features that have been added to Thermal DesktopTM to support concurrent engineering. The thermal model may now be characterized by a set of design variables that are easily modified to reflect system level design changes.