曝光台 注意防骗
网曝天猫店富美金盛家居专营店坑蒙拐骗欺诈消费者
Multlbus/IEEE 796 compatible expansion slots and a power supply. The chassis
contains the five boards shown In figure 5. The slmulator software executes on
an INTEL 86/30 single board computer (ref. 8) with an 8086 microprocessor, an
8087 floating point coprocessor, and 256 Kb of random access memory. Table Ill
lists the Jumper connections used on the 86/30 board. A Zendex ZX-2OOA single
board disk controller (ref. 9) Is Included to communicate with the dlsk drives.
A data translation DT 1742-32 DI 32 channel, differential input A/D converter
(ref. lO) accepts the analog contro] signals from the controller. Finally,
there are two data translation DT 1842-8-V 8 channel D/A boards (ref. lO) which
convert all of the slmulated outputs to analog voltages. Table IV lists the
pin connections from the D/A boards for each output variable.
The simulator software consists of 21 routines, II in FORTRAN and I0 in
8086 and 8087 assembly language (ref. Ill. In addltlon, the simulation uses
functions and utilities contained in four libraries. The routines share variables
through common blocks of memory. These common blocks are listed in
table V and their contents are described In table VI.
For proper stability and accuracy a good rule of thumb is that a numerical
(Euler) integration tlme of not more than one quarter the control Interval
should be used in the simulation. Use of this rule wlll reduce the Interactlon
between the simulation and the control by reducing any phase shift due to tlme
delays in the simulation. The ADIA control interval was 40 msec. Thus a slmulator
integration tlme of lO msec was the goal. As a full envelope slmulatlon,
the minimum achievable update time (integration tlme) for the slmulatlon was
approximately 40 msec or four times the desired Interval. To overcome thls
problem, a drastic reduction In the cycle time of the algorithm was requlred.
It was possible to determine the execution time for each major subroutine.
Most of the FORTRAN code had already been optimized (ref. 12) so the execution
time for each routine was essentially the minlmum possible. Several alternative
solutions to the execution time problem were considered. These included
uslng a faster processor, putting the simulation on multiple computers (parallel
processing), and/or modifying the structure of the software. To avoid having
to change the microcomputer hardware, the simulator software was modified.
The simulator was changed from a full-envelope model to an operating point
model. This was achieved by breaking the simulation into two loops: an
initialization loop and a real-time run loop. Now the base points and the
matrix elements are calculated in nonreal time (these are the longest routines)
and then, in the real-time loop, the system equations are evolved as a set of
linear equations to the new operatlng condition. This allowed the real-time
loop cycle time to be reduced to 12 msec. The result is a linear model valid
within a small reglon about a given operating point. This model gives excellent
steady-state results and good transient results for small perturbations,
such as small movements of the power lever angle (PLA). However, the model
will not perform accurately for large perturbations such as large PLA
movements.
Description of Modes
The simulation can operate in five different modes depending upon the
application. The modes are: initialization/run, PSL/hybrld, calibration,
open-loop/closed-loop, and actuators only. These modes are controlled by software
switches described in table VII.
Initialization/run. - In the Initialization mode, the simulator initializes
variables to their desired steady-state operating point values. Initialized
variables include the engine base points and the open-loop setpoints. In
the run mode, the simulation enters the real-time loop. Here the state equations
are evolved from the previous operating point to the desired operating
point using Euler's method for numerlcal Integration. These two modes are a
consequence of the fact that the simulation is not fast enough to accurately
model the whole flight envelope dynamically in real time.
Propulslon system lab (PSL)lhybrid. - The PSL mode scales the control slgnals
and alters the simulator outputs to correspond to those of the englne in
the PSL. Initlally the inputs and outputs of the simulator were scaled identically
to the inputs and outputs of the FlO0 Hybrld Simulation. These were all
±10 V, straight line representatlons of the engine inputs and outputs. However,
the actual engine input and output devlces consist of linear potentiometers,
resolvers, thermocouples, flowmeters, and electro-hydraulic actuators.
中国航空网 www.aero.cn
航空翻译 www.aviation.cn
本文链接地址：
A_real-time_simulator_of_a_turbofan_engine(3)
