In this paper, a MATLAB-Simulink based general co-simulation approach is presented which supports multi-resolution simulation of distributed models in an integrated architecture. This approach was applied to simulating aircraft thermal performance in our Vehicle Systems Model Integration (VSMI) framework. A representative advanced aircraft thermal management system consisting of an engine, engine fuel thermal management system, aircraft fuel thermal management system and a power and thermal management system was used to evaluate the advantages and tradeoffs in using a co-simulation approach to system integration modeling. For a system constituting of multiple interacting sub-systems, an integrated model architecture can rapidly, and cost effectively address technology insertions and system evaluations. Utilizing standalone sub-system models with table-based boundary conditions often fails to effectively capture dynamic subsystem interactions that occurs in an integrated system. Additionally, any control adjustments, model changes or technology insertions that are applied to any one of the connecting subsystems requires iterative updates to the boundary conditions. When evaluating a large set of trade studies, the number of boundary condition models and time to generate these models becomes intractable and affects capturing the results accurately. A single interconnected model of all the subsystems may be impractical and using additional external packages may be prohibitive in terms of cost or compatibility. This general approach requires no additional MATLAB toolboxes. Two different data interchange mechanisms are presented. A dynamic vehicle system integrated model was developed to enable customizability and flexibility. The developed co-simulation approach was combined with this flexible architecture to enable system evaluation. Example applications using the vehicle system model integrated architecture with the co-simulation approach are discussed.

The combination of increasing performance demands, increasing system complexity, and the need for reduced program development schedule and budget costs in the aerospace industry is driving engineers to increasingly rely upon modeling, simulation, and analysis (MS&A) in the platform development cycle. One approach to ensuring that such integrated system simulations remain computationally tractable is co-simulation utilizing technology found in commercially available packages, such as PC Krause and Associates, Inc.'s (PCKA's) Distributed Heterogeneous Simulation (DHS) / FastSim software. In such co-simulation environments, dynamic models are executed in independent model spaces, with coupling between subsystems achieved by exchanging a minimal set of required data typically found at subsystem boundaries. In such environments, an important challenge that must be overcome is the estimation of communicated signals whose true values may only be updated at some reduced rate compared to the model in which they are used. This estimation has traditionally been achieved with simple sample/hold techniques, wherein the communicated signals are sampled at the discrete communication intervals, and interpolation (most often utilizing polynomial functions) is used to reconstruct the signal in between samples. While conceptually simple, this approach often gives rise to step discontinuities in the communicated signals due to the difference between the interpolating function used and the signal itself. In certain models, such step discontinuities are undesirable, as they may give rise to spurious high-frequency dynamics, resulting in slower simulation speeds and potentially inaccurate solutions.In this paper, the problem of signal estimation and reconstruction is recast in a generalized framework based on notions of prediction and error correction. Abstracting the problem in this form allows for describing different signal estimation schemes in a unified manner, as application of different prediction and error correction formulas, including the sample/hold techniques commonly used in co-simulation tools. Based on this framework, a new prediction and error correction scheme is derived which allows for maintaining various levels of continuity in the estimated signal and in its derivatives, ensuring that smoothness is retained and step discontinuities associated with traditional sample/hold approaches are eliminated. This paper presents a detailed comparison of the proposed technique against traditional approaches, whereby it is shown that the new method yields more desirable performance under certain conditions. Ultimately, the generalized framework and the proposed approach put forth in this paper provide engineers engaged in the MS&A of aerospace platforms additional configuration flexibility to achieve meaningful results.

This paper describes a coupled-circuit physical-variable modeling of multiphase induction motors. The presented modeling interface makes it straightforward to implement an induction machine with arbitrary number of phases and/or phase groups on the stator and the rotor. The 3-, 6-, and 9-phase motors are simulated and compared. It is shown that machines with higher number of phases have less severe torque pulsation and the stator current increase following a loss of one phase. For the 9-phase machine, several studies involving loss of multiple phases are also presented, wherein the relative location of the faulted phases is shown to have a significant impact on redistribution of currents and resulting electromagnetic torque. The proposed models can be used to represent induction motors and generators for transient studies involving multiple faults, system-level reconfiguration, and survivability.

Power systems that include regulated power-electronic sources and/or loads are susceptible to potentially destabilizing interactions between these components. A variety of techniques and methodologies have been developed to characterize the small- and large- displacement stability of such systems. Perhaps the most common approach is to establish the input/output impedance-versus-frequency characteristics of all sources and loads, whereby Nyquist- and/or Bode-inspired criteria may be used to characterize interconnected system stability. Essential to this methodology is a means of accurately and efficiently determining the input and/or output impedance-versus-frequency characteristics of the power electronic components that comprise the overall system. These frequency-domain characteristics can be established by (1) direct measurement, (2) exercising detailed simulations, or, more commonly, (3) using state-space average-value models. The primary disadvantage of using direct measurements is that the hardware must be available a-priori which makes it difficult and/or expensive to change or tailor the impedance characteristics if instabilities occur. Calculation of the impedance characteristics from detailed simulations is generally time consuming, especially if the low-frequency characteristics are needed, and little insight is gained as to how the impedance characteristics are affected by the various design parameters. Average-value models overcome the previous disadvantages; however, they introduce a new one. In particular, the derivation of an average-value models is typically time consuming, especially if the circuit topology is complex and/or the power converter exhibits multiple load-dependent modes of operation. In this paper, an automated approach of establishing average-value models of power electronic converters of arbitrary complexity is set forth. The user-supplied inputs consist of a standard Spice-like circuit description (branch parameters and network graph) whereupon the input/output impedance-versus frequency characteristics are automatically and rapidly established. In addition to eliminating the need for the analytical derivation of average-value models, this technique readily permits the inclusion of secondary effects such as conduction losses, switching losses, and magnetic nonlinearities, to name a few. This technique has been successfully applied to characterize the output impedance of a one-quadrant dc/dc buck converter and a three-phase generator/rectifier source.

To support research and analysis requirements in the development of future power systems, a flexible and efficient means of predicting the dynamic performance of large-scale multi-disciplinary systems prior to hardware trials is crucial. With the development of Distributed Heterogeneous Simulation (DHS), the technology now exists to enable this type of investigation. Previously, DHS was shown to allow the interconnection of component simulations running on a single-or distributed-computer network and developed using any combination of a variety of commercial-off-the-shelf (COTS) software packages for the Microsoft Windows operating system. However, for large-scale systems, all subsystem models may not be developed in software packages operating under Windows thereby requiring a translation of such models in order to incorporate them within a system simulation. In this paper, the DHS technique is expanded to support the UNIX operating system, thus, allowing subsystem models developed and executed on either UNIX- or Windows-based computers to be interconnected to form a dynamic system simulation. For the purpose of demonstration, a more-electric fighter (MEF) power system, such as that found on the Joint Strike Fighter (JSF), has been selected as a study system. This system is comprised of ten component models each developed using MATLAB/Simulink (TM), EASY5 (TM), or ACSL (TM). Utilizing the system simulation, studies have been performed to illustrate the dynamic interactions between the subsystems when simulated on a heterogeneous computer network containing both Windows- and Unix-based machines.

In this paper, a parametric average-value modeling approach is applied to a high-frequency six-phase aircraft generation subsystem. This approach utilizes a detailed switch-level model of the system to numerically establish the averaged dynamic relationships between the ac inputs of the rectifier and the dc-link outputs. A comparison between the average-value and detailed models is presented, wherein, the average-value model is shown to accurately portray both the large-signal time-domain transients and the small-signal frequency-domain characteristics. Since the discontinuous switching events are not present in the average-value model, significant gains can be realized in the computational performance. For the study system, the developed average-value simulation executed more than two orders of magnitude faster than the detailed simulation.

Future Air Force intelligence, surveillance, and reconnaissance (ISR) platforms, such as high-altitude Uninhabited Aerial Vehicles (UAV), may drastically change the requirements of aircraft power systems. For example, there are potential interactions between large pulsed-power payloads and the turbine engine that could compromise the operation of the power system within certain flight envelopes. Until now, the development of large-scale, multi-disciplinary (propulsion, electrical, mechanical, hydraulic, thermal, etc.) simulations to investigate such interactions has been prohibitive due to the size of the system and the computational power required. Moreover, the subsystem simulations that are developed separately often are written in different commercial-off-the-shelf simulation programs. In this paper, a new technique useful for the numerical simulation of large-scale systems to overcome these obstacles, known as Distributed Heterogeneous Simulation (DH$), is utilized to form a dynamic system-level simulation of a high-altitude, long-endurance UAV-type of power system. This system includes detailed dynamic models of a turbine engine, high- and low-spool generators, and payloads. Although not necessary, all of the component models for this system were developed within the same simulation environment, specifically with MATLAB/Simulink. This enabled a single-computer integrated system model and a distributed computer system simulation to be formed thereby allowing for a direct comparison of simulation accuracy and computational performance for the two simulation approaches. From this comparison, it was determined that by distributing the system simulation across three computers, a 21-fold increase in simulation speed could be realized while producing nearly identical results.

A novel multirate method of simulating power-electronic-based systems containing a wide range of time scales is presented. In this method, any suitable integration algorithm, with fixed or variable time-step, can be applied to the fast and/or slow subsystems. The subsystems exchange coupling variables at a communication interval that can be fixed or varied dynamically depending upon the state of the system variables. The proposed multirate method is applied to two example power systems that include power-electronic subsystems. Increases in simulation speed of 183-281 % over established single-rate integration algorithms are demonstrated.

An allocation algorithm for optimally assigning the various subsystem simulations, within a distributed heterogeneous simulation, to a specific set of computational resources has been developed. This algorithm uses a cost function that approximates the simulation execution time for each of the subsystems based upon the model complexity and the performance parameters of the available computer resources. The cost function is then evaluated to determine the optimal allocation that ensures the overall simulation execution time is minimized. In this paper, the allocation algorithm is applied to a large-scale power-electronic-based aircraft electrical power system. This study system is comprised of ten component simulations that together are modeled by 85 state variables and include 74 switching devices. Both optimal and sub-optimal allocations are considered and the predicted simulation run times are verified experimentally.

A new paradigm for simulating large-scale dynamical systems is described and applied to an integrated naval power system. This approach enables the overall system simulation to be formed as an interconnection of interdependent dynamic simulations, each representing a specific electrical, mechanical, hydraulic, and/or thermal component/subsystem. Each simulation may be developed independently using possibly different commercial off-the-shelf simulation programs thereby allowing the most suitable language or tool to be used based on the design/analysis needs. The proposed simulation paradigm is applied to an advanced naval electric power system that includes a zonal dc electric distribution system and an integrated electric generation and propulsion system. The composite system simulation is implemented on a three-computer network of personal computers by interconnecting simulati ons of the constituent subsystems. It is shown that significant improvements in computational speed are achieved using this approach.

In this paper, a new technique useful for the numerical simulation of large-scale systems is presented. This approach enables the overall system simulation to be formed by the dynamic interconnection of the various interdependent simulations, each representing a specific component or subsystem such as control, electrical, mechanical, hydraulic, or thermal. Each simulation may be developed separately using possibly different commercial-off-the-shelf simulation programs thereby allowing the most suitable language or tool to be used based on the design/analysis needs. These subsystems communicate the required interface variables at specific time intervals. A discussion concerning the selection of appropriate communication intervals is presented herein. For the purpose of demonstration, this technique is applied to a detailed simulation of a representative aircraft power system, such as that found on the Joint Strike Fighter (JSF).This system is comprised of ten component models each developed using MATLAB/Simulink(TM), EASY5(TM), or ACSL(TM). When the ten component simulations were distributed across just four personal computers (PCs), a greater than 15-fold improvement in simulation speed (compared to the single-computer implementation) was achieved.

An algorithmic method of developing the state equations of complex power circuits and systems has recently been developed. In this approach, referred to as the Automated State Model Generator (ASMG), the system is described by the pertinent branch parameters and the circuit topology; however, unlike existing circuit-based approaches, the difference equations are not implemented at the branch level. Instead, the composite system state equations are established automatically and updated dynamically for each new topology of the switching network. Since the state equations are available, linearization, eigensystem analysis, and model-order reduction can be performed quickly and conveniently at the system level. In addition, it becomes possible to identify the operational modes of converters and inverters based on the cyclically repeated sequence of topologies. The ASMG includes a built-in switching logic for common power-electronic switching devices. The proposed simulation approach is particularly effective when modeling advanced power electronic systems that include special-purpose multi-phase electrical machines, transformers, and other components that are not available in the libraries of commonly used circuit simulators. An ASMG toolbox that is compatible with MATLAB/SIMULINK has been developed and used to implement a detailed simulation of a 6-phase generator/rectifier system.

An automated state model generator (ASMG) is a tool for modeling and analysis of lumped-parameter power-electronic-based systems. In this modeling approach, the minimal state-space representation of the overall system is generated automatically and updated dynamically based upon the topological state of the system. However, due to the changing topology, simulation of a switched circuit using the ASMG requires the concatenation of solutions to the initial value problems (IVPs) corresponding to the time intervals between commutations. In this paper, a transformation of state variables is derived such that the states are continuous throughout the simulation process. This feature eliminates the need to re-initialize the ODE solver. The continuous state algorithm is verified on a high-pulse-count power supply and sets the stage for state-space averaging and system-level analysis of switched circuits

The state variable approach is a numerically efficient and analytically meaningful method of predicting and characterizing the transient responses of power-electronic-based systems. An automated method of establishing state-space models of switched electrical networks is considered. In this modeling approach, the minimal state-space representation of the overall system is generated dynamically as each topology is encountered. Due to the changing topology, the simulation of a switched circuit requires concatenation of the solutions of initial value problems (IVPs) corresponding to time intervals between switching events. In this paper, a transformation of state variables is derived such that the computer-generated equations have the same dimension and the new states are continuous throughout the study. This feature eliminates the need for re-initializing the ODE solver and sets the stage for automated state-space averaging and system-level analysis of switched circuits. The algorithm is verified using a practical example of a synchronous machine-rectifier system