📄 208-211.html
字号:
<HTML>
<HEAD>
<META name=vsisbn content="0849398010">
<META name=vstitle content="Industrial Applications of Genetic Algorithms">
<META name=vsauthor content="Charles Karr; L. Michael Freeman">
<META name=vsimprint content="CRC Press">
<META name=vspublisher content="CRC Press LLC">
<META name=vspubdate content="12/01/98">
<META name=vscategory content="Web and Software Development: Artificial Intelligence: Other">
<TITLE>Industrial Applications of Genetic Algorithms:Space Shuttle Main Engine Condition Monitoring Using Genetic Algorithms and Radial Basis Function Neural Network</TITLE>
<!-- HEADER -->
<STYLE type="text/css">
<!--
A:hover {
color : Red;
}
-->
</STYLE>
<META NAME="ROBOTS" CONTENT="NOINDEX, NOFOLLOW">
<!--ISBN=0849398010//-->
<!--TITLE=Industrial Applications of Genetic Algorithms//-->
<!--AUTHOR=Charles Karr//-->
<!--AUTHOR=L. Michael Freeman//-->
<!--PUBLISHER=CRC Press LLC//-->
<!--IMPRINT=CRC Press//-->
<!--CHAPTER=10//-->
<!--PAGES=208-211//-->
<!--UNASSIGNED1//-->
<!--UNASSIGNED2//-->
<CENTER>
<TABLE BORDER>
<TR>
<TD><A HREF="206-208.html">Previous</A></TD>
<TD><A HREF="../ewtoc.html">Table of Contents</A></TD>
<TD><A HREF="211-214.html">Next</A></TD>
</TR>
</TABLE>
</CENTER>
<P><BR></P>
<P>Rocket plumes are emissive events subject to the same physics (with more complications, of course) as burning nickel over an open flame. The Optical Plume Anomaly Detection (or OPAD) program was initiated by researchers at MSFC as an effort to take advantage of the wealth of information contained in the exhaust plume of a rocket engine. The initial idea was to identify anomalous spectral events which were consistent with known mechanical failures and then use them as templates in the health monitoring of future engine tests (ground or in-flight). This could then be coupled with the anomalous events found in the vibrational and other sensor data to determine the overall state, or health, of the engine.
</P>
<P><A NAME="Fig3"></A><A HREF="javascript:displayWindow('images/10-03.jpg',300,531)"><IMG SRC="images/10-03t.jpg"></A>
<BR><A HREF="javascript:displayWindow('images/10-03.jpg',300,531)"><FONT COLOR="#000077"><B>Figure 10.3</B></FONT></A> Example emission spectra for nickel.</P>
<P>Using plume spectral acquisition instrumentation developed specifically for OPAD and mounted on the SSME Technology Test Bed (TTB) at MSFC, the process of building a cumulative database of the spectral templates began [9]. Researchers attempted to catalog the various spectral forms associated with changes in the SSME’s operation (i.e., changes in oxidizer/fuel ratio, engine startup transients, etc.) so that a baseline of “expected spectral signatures” would be established. The “template idea,” however, soon gave way to even more ambitious goals as a result of some initial findings in the TTB experimental program. The spectral data from one test in particular revealed a major occurrence of a metallic species which was indigenous to the SSME preburner faceplate. An even closer evaluation of the amount of metallic species present versus time showed an initial erosive event of the metal followed by numerous other anomalous emissions, all leading up to an engine-threatening erosion of the faceplate. This meant that anomalous events could be predicted.
</P>
<P>As a result of these findings, the focus of the researchers turned to not only anomaly detection but also metal quantification. In other words, health monitoring now involved the simultaneous tasks of anomaly detection and determination of the severity of the anomaly. This meant that the free atom densities of all the metals of interest within the engine would have to be predicted for every temporal scan taken by the instruments. The metal quantification process would essentially give metal concentration versus time. Spikes in this time trace would then be indicative of a metal erosion. Figure 10.4 summarizes the three major objectives of the health monitoring system.</P>
<P>Any spectrum obtained with the OPAD instrumentation is composed of three components: 1) a dominant OH component which arises from the burning of dissociated hydrogen radicals, 2) a background noise component caused by the scattering of background light, and 3) a metallic component, if indeed there is one, which would be indicative of a metal erosion. Thus, the quantification of a metal erosion and the subsequent identification of any anomalies requires a spectral “cleaning” procedure followed by an evaluation of the plume metallic state. In other words, methods for removing the OH and background components of the spectrum would need to be employed so that the underlying metallic component could be seen. Then, the metal quantities would have to be ascertained from the remaining metallic component.</P>
<P>For a given spectrum, ascertaining the metallic quantity could only be done through two methods: 1) by comparing the spectrum to past spectra obtained from plume seeding tests, or 2) using a theoretical model that emulates the emissive nature of the plume. The first option is plagued by an inability to precisely measure the erosion and survivability rates of the inserted species. Thus, there would exist plume spectra to compare to, but the metallic content associated with the spectra would be in error. Moreover, it would not be cost effective to run the SSME through all the possible metallic seeding combinations. For these reasons, option two was selected.</P>
<P><A NAME="Fig4"></A><A HREF="javascript:displayWindow('images/10-04.jpg',500,359)"><IMG SRC="images/10-04t.jpg"></A>
<BR><A HREF="javascript:displayWindow('images/10-04.jpg',500,359)"><FONT COLOR="#000077"><B>Figure 10.4</B></FONT></A> Overview of tripartite monitoring scheme.</P>
<P>Researchers at Vanderbilt University and AEDC developed a theoretical plume model known as SPECTRA [11,12]. The forward operation of the SPECTRA code involves the calculation of a theoretical plume spectrum from a pre-defined set of metal concentrations and flow parameters. The reverse operation of SPECTRA (obtaining the metallic components which made up the spectrum) cannot be written in a numerically convenient analytical form because of the almost insurmountable mathematics involved. This, therefore, mandates that the SPECTRA code be applied in an iterative manner until it converges with the spectrum obtained from the OPAD instrumentation. The set of SPECTRA input parameters which produced this convergence would then specify the current metallic state of the plume. Operating this iterative sequence is computationally exhaustive and precludes its use in real-time health monitoring systems. For this reason, neural network techniques have been investigated that accomplish this fundamental task in an expeditious manner. Described herein is a radial basis function neural network (RBFNN) architecture that models the “inverse” operation of the SPECTRA code and allows for real-time SSME anomaly identification and quantification via plume spectral assessment.
</P><P><BR></P>
<CENTER>
<TABLE BORDER>
<TR>
<TD><A HREF="206-208.html">Previous</A></TD>
<TD><A HREF="../ewtoc.html">Table of Contents</A></TD>
<TD><A HREF="211-214.html">Next</A></TD>
</TR>
</TABLE>
</CENTER>
<hr width="90%" size="1" noshade>
<div align="center">
<font face="Verdana,sans-serif" size="1">Copyright © <a href="/reference/crc00001.html">CRC Press LLC</a></font>
</div>
<!-- all of the reference materials (books) have the footer and subfoot reveresed -->
<!-- reference_subfoot = footer -->
<!-- reference_footer = subfoot -->
</BODY>
</HTML>
<!-- END FOOTER -->
⌨️ 快捷键说明
复制代码
Ctrl + C
搜索代码
Ctrl + F
全屏模式
F11
切换主题
Ctrl + Shift + D
显示快捷键
?
增大字号
Ctrl + =
减小字号
Ctrl + -