Charts are available to convert from one type of measurement to another as shown in Figure 19-13. Many of these charts also show approximate vibration limits. The charts demonstrate the independence of velocity meas-urements relative to frequency， except at very low and very high frequencies where the amplitude limits are constant throughout the operating speedrange. These limits are approximate.the type of machinery， casing， founda-tion， and bearings must be considered to determine final vibration limits.

.ibration Instrumentation Selection

The type of vibration instrumentation， its frequencyranges， its accuracy，and its locationwithin， or on the machine， must be carefully analyzed with respect to the diagnostics required to be achieved. These guidelines have been previously discussed.

The displacement noncontacting eddy current sensor is most effective for monitoring and measuring vibrations near rotational and subrotational speeds. While the displacement sensor is capable of measuring vibrationfrequencies of more than 2kHz， the amplitude of vibrational displacementlevels that occur at frequencies above 1 kHz are extremely small， and are usually lost or buried in the noise level of the readout system. The accelera-tion sensor is best suited for measurements at high frequencies， such as blade-passing and gear-meshing frequencies; however， the signals at one rotationalspeed are usually at low accelerationlevels， and may be lost in the noise level of the measurement system monitoring. .ow-pass filtering and additional amplification stagesmay， therefore， be necessary to bring out the rotational speed signals when measurements are made with accelerometers.
.elocitysensors， because of their limited operational frequency range(usually) from 10 Hz to 2kHz， are not recommended for application in a diagnostic system for high-speed machinery. .elocity sensors have moving elements and are subject to reliability problems at operational temperatures of more than 250 0F (121 0C). Gas turbine engine casing temperatures are usually in the 500 0F (200 0C) level or above; hence， sensor locations must be carefully examined for temperature levels. Accelerometers for these higher temperatures are more easily available than velocity sensors. At these ele-vated operational temperatures， high-frequency accelerometers (20 kHz and above) are available from only a few selected manufacturers.
Selection of Systems for .nalyses of .ibration .ata
The overall vibration level on a machine is satisfactory for an initial orrough check. However， when a machine has a seemingly acceptable overalllevel of vibration， there may be hidden under this level some small levels of vibrations at discrete frequencies that are known to be dangerous. An example of this is subsynchronous instabilities in a rotor system.
In the analysis of vibration data there is often the need to transform thedata from the time domain to the frequency domainor， in other words， to obtain a spectrum analysis of the vibration. The original and inexpensive system to obtain this analysis is the tuneable swept-filter analyzer. Becauseof inherent limitations of thissystem， this process， despite the use of auto-mated sweep， is time-consuming when analyzing low frequencies. When thespectra data needs to be digitized for computer inputing， there are further limitations in capability of tuneable filter-analysis systems.
.eal-time spectrum analyzers using ..time compression.. or the ..fast Four-ier transform.. (FFT) techniques are used extensively for performing vibra-tion spectrum analysis in computerized diagnostic systems. The FFTanalyzers use digital-signal processing， and hence are easier to integrate with the modern digital computer. FFT analyzers are often hybrids using micro-processors and FFT-dedicated circuitry.
The FFT can be implemented in a computer using the FFT algorithm for obtaining a pure mathematical computation. While this computation is anerror-free process， its implementation in a digital computer can introduceseveral errors. To avoid theseerrors， it is essential to provide signal con-ditioning upstream of the computer. Such signal conditioning minimizes theerrors， such as aliasing and signal leakage introduced in sampling and digitizing the time domain. Such signal conditioning systems will introduce considerable expense and complexity in effecting the mathematical FFT in a computer. The computerized FFT is also slower than a dedicated FFTanalyzer. It also has limitations in frequency resolution.Hence， the use of a dedicated FFT analyzer is considered to be the most reliable and cost-effective means for performing frequency spectrum analysis and plots in a computerized system for machinery diagnostics.
　
中國航空網 www.k6050.com
航空翻譯 www.aviation.cn
本文鏈接地址：燃氣渦輪工程手冊 Gas Turbine Engineering Handbook 3(51)