MATLAB/Simulink Framework For Modeling Complex Coolant .

2y ago
1.72 MB
8 Pages
Last View : 2d ago
Last Download : 2y ago
Upload by : Randy Pettway

NREL/CP-5400-66324. Posted with permission.Presented at the SAE 2016 World Congress &Exhibition, 12-14 April 2016, Detroit, Michigan.MATLAB/Simulink Framework for Modeling ComplexCoolant Flow Configurations of Advanced AutomotiveThermal Management Systems2016-01-0230Published 04/05/2016Gene Titov, Jason Lustbader, and Daniel LeightonNational Renewable Energy LaboratoryTibor KissThermal Sciences ConsultingCITATION: Titov, G., Lustbader, J., Leighton, D., and Kiss, T., "MATLAB/Simulink Framework for Modeling Complex Coolant FlowConfigurations of Advanced Automotive Thermal Management Systems," SAE Technical Paper 2016-01-0230, 2016, doi:10.4271/201601-0230.Copyright 2016 SAE InternationalAbstractThe National Renewable Energy Laboratory’s (NREL’s) CoolSimMATLAB/Simulink modeling framework was expanded by includinga newly developed coolant loop solution method aimed at reducingthe simulation effort for complex thermal management systems. Thenew approach does not require the user to identify specific coolantloops and their flow. The user only needs to connect the fluid networkelements in a manner consistent with the desired schematic. Usingthe new solution method, a model of NREL's advanced combinedcoolant loop system for electric vehicles was created that reflected thetest system architecture. This system was built using componentsprovided by MAHLE Inc. and included both air conditioning andheat pump modes. Validation with test bench data and verificationwith the previous solution method were performed for 10 operatingpoints spanning a range of ambient temperatures between -2 C and43 C. The largest root mean square difference between data andsimulation results for pressure, temperature, energy and mass flowrate was less than 7%.IntroductionWhen operating, the air conditioning (A/C) system is the largestauxiliary energy consumer in a conventional vehicle. A/C loadsaccount for more than 5% of the fuel used annually by light-dutyvehicles in the United States [1]. Climate control loads can have aneven larger impact on hybrid electric vehicle (EV), plug-in hybridEV, and all-electric vehicle performance. Hybrid EVs show a 22%lower fuel economy with the A/C on [2]. For all-electric vehicles, theeffect of the climate control system usage is even more severe. Due toa shortage of waste heat, heating of the passenger cabin in EVs has torely on battery energy. Cooling the cabin can also take a significantportion of energy available in the battery, significantly reducingvehicle efficiency and range. Mitsubishi reports that the range of thei-MiEV can be reduced by as much as 68% with heating and 46%with cooling of the cabin on Japan's 10-15 cycle [3]. The AdvancedPowertrain Research Facility at Argonne National Laboratory hasreported 59.3% and 53.7% reductions in range due to maximumheating and maximum cooling, respectively, for the Ford Focus EVoperating on the Urban Dynamometer Driving Schedule cycle [4]. Inaddition to these climate control impacts, electric-drive vehicles mayhave additional cooling requirements for the electric traction drivesystem components, including batteries, power electronics, andelectric machines.To solve these challenges, alternative heating methods and moreefficient cooling systems are needed for EVs. These methods ofteninvolve running the A/C system in heat pump mode to reduce theheating power requirements of the cabin. In some advanced concepts,the traditional liquid coolant-based thermal management issupplemented with refrigerant-based cooling systems, which can makethe thermal management system as a whole significantly morecomplex. When developing a thermal management system for aninternal combustion engine vehicle, it has traditionally been sufficientto simulate the A/C system and the liquid coolant-based cooling systemseparately. For advanced vehicles, especially for hybrid and all-electricvehicles, the benefits of interconnectedness of the thermal managementand A/C systems outweigh the associated complexity. This, in turn,results in a requirement for more integrated simulation approaches.The more complex thermal management systems of advanced vehiclestypically allow for various modes of operation that can be selectedbased on driving and ambient conditions. Investigating a number ofsystem alternatives and determining the best ranges for variousoperating modes with experimental methods can be very timeconsuming. A good system simulation tool can greatly reduce the timeand expense of developing these complex systems. Such tools shouldalso be able to efficiently co-simulate with vehicle simulation programsand should be applicable for evaluating various control algorithms. The

MATLAB/Simulink simulation environment, popular in the automotiveindustry, is well suited for development of such models and meets therequirements of dynamic modeling of complex systems.BackgroundTo meet the needs of advanced vehicle thermal system simulations,the U.S. Department of Energy’s National Renewable EnergyLaboratory (NREL) is building an integrated single- and two-phasethermal system modeling framework, CoolSim, in Simulink. Thisintegrated approach allows for rapid system analysis and design in aflexible and open modeling environment. Simulink is a commonengineering platform that allows for co-simulation with vehiclemodeling software Autonomie [5]. NREL previously developed anA/C system simulation modeling framework in MATLAB/ Simulinkand validated its results against test bench data. To match the widerange of A/C modeling needs, NREL developed models with threedifferent levels of detail: the Fully-Detailed, Quasi-Transient, andMapped-Component models.The three models involve different levels of trade-offs between speedand accuracy to meet a wide range of modeling needs. The FullyDetailed model captures the system transient behavior accurately butruns at 0.1 of real-time speed [6]. The Quasi-Transient and MappedComponent models are progressively more simplified while trying tomaintain accuracy and run at real-time speed and faster than 10 timesreal-time speed, respectively [7]. The goal of these newer modelversions was to provide faster simulation tools for less detailed,vehicle-focused drive-cycle-based evaluations of A/C systems. Forsteady-state conditions, the Quasi-Transient model providesessentially the same accuracy as the Fully-Detailed model. TheMapped-Component model does lose some accuracy in steady-stateconditions. For the SC03 drive cycle, the averaged results of powerand heat exchange rates obtained with the Quasi-Transient model arewithin 3% of the results of the Fully-Detailed model. The MappedComponent model results are within 15% of the results of theFully-Detailed model. Short transients, such as those occurring duringcompressor cycling, produce the most deviation from the FullyDetailed model for both simplified models. Conversion from theQuasi-Transient A/C system model approach to the other two modelsis relatively simple within the CoolSim framework. This allows a newsystem model to be developed with the Quasi-Transient version beforethe results are refined using the slower Fully-Detailed version oraccelerated using the faster Mapped-Component model version.As outlined in the Background section, there is a need for coupledthermal system simulations due to interconnectedness of therefrigerant and liquid coolant circuits used in advanced thermalmanagement systems, especially the ones developed for EVs. Toaddress this need, NREL’s refrigerant circuit simulation model wasextended with a liquid-coolant circuit simulation capability. Theoriginally implemented coolant Fluid Network solution method [8]was selected for its speed and algorithmic simplicity. This approachworks well for relatively simple systems that do not involve featuressuch as changes in fluid flow direction, a large number of operatingmodes, etc. The complexity and flexibility of next generationintegrated systems, however, put a higher burden on the user forsetting up the models using the originally developed Fluid Networkapproach. Certain system configurations with changing flowdirections based on mode and operating conditions were also found tobe challenging to simulate. To improve modeling of these morecomplex thermal systems, a more general method that considers thecoolant as a compressible medium with an artificially small bulkmodulus was developed. This approach is similar to that used fortwo-phase flow of refrigerant in the Quasi-Transient method.While this approach comes at a higher computational cost, theflexibility and ease of model development make it a preferablealternative for complex fluid networks. Furthermore, it wasdetermined that the bulk of computational effort is typically spent onthe refrigerant circuit, making additional computational expensesrelatively small. An additional benefit for developers comes from thefact that the solution methods for both single- and two-phase flowsbecome similar. As a result, the effort spent on developing models inthe CoolSim framework for specific systems is reduced.A model built with this updated version of CoolSim was developed forNREL's combined fluid loop (CFL) thermal management test bench inboth active cooling and heating modes. Comparisons of simulatedresults with measured data validate the new coolant loop solutionapproach. Additional verification was obtained by comparisons withresults produced by the Fluid Network solution method.A New Approach to Coolant Loop ModelingThe new single phase solution method integrates the “QuasiTransient” modeling method for the refrigerant circuit with a similarapproach for coolant loops. The details of the two-phase refrigerantloop solution method are discussed in [7]. This paper focuses ondetails of the single-phase coolant loop modeling. The original FluidNetwork approach for solving coolant loops in CoolSim relies on atheory similar to Kirchoff's law for electric circuits. This approach isefficient and will continue be of use for simpler systems; however,the approach proved to be complicated for quick development ofmore sophisticated models with many modes of operation. Figure 1shows the model interface in Simulink for the original solutionapproach and Figure 2 shows the new approach interface. Althoughthe new interface looks more involved, the time needed to build andtest the new model was notably less due to elimination of a coolantloop specification step of the original approach. An importantadvantage of the new approach is its ability to simulate complicatedmulti-mode fluid networks with changing flow directions instraightforward manner.Similar to the Quasi-Transient refrigerant circuit approach, coolantloops in the new method are represented by zero-dimensional (0-D)volume simulation blocks connected with simulation blocks forone-dimensional (1-D) pipes, valves, or orifices. In general, anysystem component that can provide a flow rate due to a pressuredifferential can be connected to the 0-D volume blocks. For thecoolant loop network topology, these 0-D volume blocks are referredto as junctions, as opposed to actual fluid reservoirs such asaccumulators, headers, compressor suction/exit volumes, etc., in caseof refrigerant circuits. These coolant junction simulation blocks canstill be used to model large volumes of coolant if needed.

Figure 1. Original approach. NREL combined loop model, Simulink top-level viewFigure 2. New approach. Combined loop model, new approach SimulinkThe strategy of this new Quasi-Transient single-phase numericalmethod is to approach a steady-state solution that corresponds to theboundary conditions prevailing at each of the time steps. In this way,the model approximates a solution that would be obtained with ahypothetical quasi-steady-state model that at every time step computessteady-state conditions for the entire system. In 0-D junctions, acompromise is made between the accuracy of the implementedconservation equations and computational speed. This is achieved byintroducing and adjusting an artificial bulk modulus, which ensuresthat all volumes in the model have similar low numerical stiffness sothat the high stiffness of liquid is avoided. This allows for significantlylarger computational time steps while maintaining numerical stabilityand accuracy, resulting in faster simulations.The 1-D pipe block assumes a constant coolant mass flow rate alongits length. The flow rate then becomes a simulation state variable. Ateach time step, the coolant pressure differential across each line iscompared to pressure differences between the junctions (0-D volumeblocks) attached to them. A numerical method is applied tocontinuously adjust the coolant mass flow rate in each of the lines. Thegoal of this method is to match the pressure drop of the line to thepressure drop between the junctions that the line connects. The coolantmass flow rate, therefore, responds with a delay but it approaches thesolution that would develop under the steady-state conditions.The downside of this approach is that the total coolant mass in thesystem is fluctuating slightly and the energy balance is not strictlyenforced. The implications include lost accuracy for modeling of fasttransients that occur on the order of seconds, such as pump cycling.For steady-state conditions, however, the conservation of mass andenergy for each junction and each of the 1-D pipes in the model isensured. A typical thermal management network is a slowly drifting“quasi-steady” system, especially in cases with constant-RPMelectric pumps. In such cases, a true conservation of mass and energywill be closely approximated by this method at all times.Junction ModelingFor junctions, a mathematical concept of “artificial mass” of coolantis introduced and the conservation equations are written for thisartificial mass. This allows for adjustment of system “stiffness.” Massand enthalpy flows into and out of a junction are obtained fromadjacent blocks. The heat transfer rate across the solid boundary of ajunction is obtained separately. The time derivative for the artificial

Temperature is calculated from the specific enthalpy, using thegeneric enthalpy-temperature relationship for a specific coolant.mass in a junction volume is the difference between the sum ofincoming and the sum of outgoing mass flow rates as is formulatedby the following equation:(1)where ma is the artificial mass, t is time, and ṁin,i and ṁout,j areincoming and outgoing mass flow rates, respectively. Conservation ofenergy is treated in a similar manner in a form of a control volumeequation. The size of the volume is constant, which implies that there isno work done by solid boundaries. The resulting time derivative of thetotal energy in a junction volume is the sum of incoming enthalpy flowrates minus the sum of outgoing enthalpy flow rates plus heat addition:Equations (1) and (2) become accurate for conservation of mass andenergy when applied to steady-state conditions. If the sum ofincoming mass flow rates is greater than the sum of outgoing massflow rates, the artificial mass will increase, and therefore the pressurein the volume will increase. Such a pressure rise will tend to reducethe incoming mass flow rates and will increase the outgoing massflow rates. As a result, the system will be driven to a steady-statesolution. A similar statement can be made for enthalpy, provided themass flow rates in and out have already reached a steady state;therefore, Eqs. (1) and (2) approach the rigorous mass and energyconservation equations in steady-state conditions, and they will tendto drive the system towards a correct steady-state solution from anytransient state.1-D Pipe modeling(2)where U is the internal energy, Ḣin and Ḣout are the enthalpy flowrates in and out of the volume, respectively, and is the heat transferrate into the volume through its boundaries.Naturally, ma and U are simulation state variables. By integration,Eqs. (1) and (2) produce the artificial mass and total energy in ajunction, making these values available before values of all the othervariables are computed as time is advanced by a step.The artificial coolant mass is introduced to allow changing howpressure and density are related through the coolant material propertyequations. This approach assumes a uniform coolant bulk modulusvalid for all conditions, making pressure a function of the artificialdensity only. The bulk modulus is also proportional to the size of thevolume. This ensures that all junction volumes in the model haveadjustable and identical “stiffness,” meaning similar coolant flowrates will result in similar pressure changes regardless of the size ofthe volume. The result is a higher allowed simulation time step andtherefore a much faster model execution.Accordingly, the pressure in a junction is:(3)where B is the bulk modulus measured in Pa, V is the size of thevolume, and ρref is a reference density. Note that while volume V isvarying from junction to junction in the system, B/V for each ofjunction remains the same. B/V and the volume are input parametersfrom which B is calculated. The lower the value of B/V, the “softer”the system will be. By dividing the total enthalpy by the artificialmass, the specific enthalpy in the volume can be obtained as:(4)For the 1-D pipe model, the governing equations are also developedwith the goal of approximating quasi-steady solutions. The approachassumes a constant coolant mass flow rate along the length of a pipeat any time. The flow rate is, however, allowed to vary in time. Afinite volume formulation is used to determine the lengthwisedistribution of flow parameters. With the coolant mass flow rate fixedalong the length of the pipe, the finite volume equations can beapplied with a marching scheme in the direction of the flow. For eachfinite volume (or segment) and at each time step, the flow variables atthe outlet boundary of the segment can be calculated from the flowvariables at the inlet to that segment and the wall temperature of thesegment. Assuming that the magnitude and direction of the coolantflow are known, it can be considered that the inlet boundaryconditions of the first segment are those prevailing in the junctionattached at the upstream side of the pipe block. Starting with thiscondition, the pressure at the exit boundary of the first segment, pout,is calculated using the Darcy-Weisbach equation (Eq. 5.8.7 in [9]).(5)where Pin, ρin, and Vin are the pressure, density, and velocity at theinlet boundary of the segment; L is the length of the segment; Dh isthe hydraulic diameter; Vin is the constant mass flow rate, ṁ, dividedby the inlet boundary density, ρin and by the pipe cross sectional area.The wall friction coefficient, f is obtained from the Hagen-Poisseuilleequation (Eq. 5.10.12 in [9]) for laminar flows and from a modifiedversion of the Colebrook equation (Eq. 5.10.13 in [9]) for turbulentflows. Next, the local heat transfer coefficient is calculated with theDittus-Boelter correlation [10] and the effectiveness-number-oftransfer-units (E-NTU) method [10] is applied to obtain the coolantexit temperature assuming that the pipe wall temperature is uniform.This approach ensures that the coolant exit temperature from thesegment does not overshoot the wall temperature:(6)

where Tin is temperature at the inlet boundary, Tw is the segment walltemperature, A is the heat transfer area (segment length times innerchannel perimeter), α is the heat transfer coefficient, and Cp is theconstant pressure specific heat. Then, the heat transfer rate from thecoolant to the wall can be calculated as follows:(7)Once the heat transfer rate is computed with Eq. (7), the specificenthalpy on the outlet boundary can also be calculated with:(8)where Ḣin is the enthalpy flow rate through the inlet boundary of thesegment. With hout and pout obtained, all the other coolant propertiescan be calculated at the outlet boundary of the first segment. Theprocedure can be repeated for

Coolant Flow Configurations of Advanced Automotive . Thermal Management Systems. 2016-01-0230 Published 04/05/2016. Gene Titov, Jason Lustbader, and Daniel Leighton. National Renewable Energy Laboratory. Tibor Kiss. Thermal Sciences Consulting . CITATION: Titov, G., Lustbader, J., Leighton, D., and Kiss, T., "MATLAB/Simulink Framework for Modeling Complex Coolant Flow Configurations of .

Related Documents:

ES360 Introduction to Controls Engineering MATLAB and SIMULINK Help Page 2 of 6 Starting SIMULINK SIMULINK can be started by: 1) Opening a SIMULINK model file (model files use the .mdl extension). 2) Starting MATLAB and clicking on the icon in the tool bar. The SIMULINK Library Browser SIMULINK

4) Simulink . Simulink is a program for simulating signals and dynamic systems. As an extension of Matlab, Simulink adds many features specific to the simulation of dynamic systems while retaining all of Matlab’s general purpose functionality. Simulink has two phases of use: mode

Test Driven Development powered by MATLAB and Simulink 45 Model-Based Design –Simulink and Stateflow Manage Requirements –Simulink Requirements Author and Execute Tests –Simulink Test Measure Test Completeness –Simulink Coverage Refactor and Verify Compliance –Simulink Check

6. Vision implementation in Matlab/Simulink 16 6.1 Matlab 16 6.1.1 Results 16 6.2 Simulink 17 7. Conclusion 19 8. References 20 Appendix A: Explanation of Matlab functions 21 Appendix B: Matlab code for color definition 22 Appendix C: Matlab code for pixel labeling 23 Appendix D: Matlab code for object recognition 24

Luigi Biagiotti Systems and Control Theory Introduction to Simulink-- 2 Simulink introduction Simulink (Simulation and Link) is an extension of MATLAB that offers modeling, simulation, and analysis of dynamical systems under a graphical user interface (GUI) environment. Simulink is based on block diagrams of Dynamic SystemsFile Size: 1MB

Introduction to Simulink Todd Atkins 4 Outline What is Simulink? Working with Simulink. How Simulink works. Continuous and discrete models Componentizing models. 5 Simulink Applications. 6 Simulink

Simulink and LEGO MINDSTORMS EV3 9 P a g e Project 1: Explore Simulink and LEGO MINDSTORMS EV3 P1.1 Get Started: Program EV3 Status Light with Simulink Motivation At the end of this project you will be able to program an EV3 brick from Simulink. Objective Create first model in Simulink Check hardware and software installation

MATLAB and Simulink Automatically generate C and HDL Verify hardware and software implementations against the system and algorithm models C MATLAB and Simulink Algorithm and System Design Real-Time Workshop Embedded Coder, Targets, Links V e r i f y Simulink HDL Coder Link for ModelSim Li