JPH0156244B2 - - Google Patents

Description

[Detailed description of the invention]

The present invention relates to a method for controlling the operation of a rotating machine, and more particularly to a method for controlling the operation of a rotating machine by detecting an abnormality in the rotating machine, which performs speed control while detecting vibration signals and diagnosing the state of the rotating machine. The present invention relates to the operation control of large rotating machines such as steam turbines and generators, and particularly to a control method for operating the machine while monitoring the vibration state at the time of startup. The problem of vibrations in rotating bodies plays a particularly important role in the operation and maintenance of thermal power plants, and investigating the cause and taking countermeasures requires accurate understanding of the situation and thorough technical consideration. In recent years, with the increase in the capacity of steam turbines, the distance between bearing centers, the weight of the rotor, the number of casings, etc. have increased, so the so-called vibration problem tends to become more complicated. Furthermore, as power system control shifts toward sharing the base load with nuclear power plants, thermal power plants are increasingly being operated for intermediate load adjustment. In the past, it was used for intermediate load adjustment, but recently for daytime load adjustment, even medium-capacity thermal power plants sometimes have to be shut down in the middle of the night when demand is low. During such plant start-up and stop operations, abnormal conditions are more likely to occur than during steady operation. In particular, abnormal vibrations due to thermal imbalance are likely to occur when the turbine is started. When operating a steam turbine or a generator, the operator's greatest attention is at the time of startup, and what to do if abnormal vibrations occur at this time is an important issue. Failure to take appropriate measures or take appropriate timing can lead to serious accidents in high-speed rotating bodies such as steam turbines. Furthermore, even during normal operation, if abnormalities are not detected quickly and appropriate measures are not taken in a timely manner, a serious accident could develop. This is a very similar problem at startup. The present invention determines whether these rotating bodies are in a normal or abnormal state by monitoring the state of a vibration signal from a vibration detector attached to a bearing, and performs operational control. The present invention was made against this background, and aims to prevent the occurrence of serious accidents by analyzing vibration detection signals and grasping the causes thereof while performing safe driving control. When the steam turbine is started, the turbine speed control device operates the main steam valve or the bypass valve in accordance with the target rotational speed of the turbine and the target rotational speed change rate to gradually increase the rotational speed. However, conventionally, when vibration occurs during speed increase, the following three control methods have been used depending on the amplitude of the vibration signal. The first method is to stop the speed increase control by the speed control device and switch to operation based on the operator's judgment, that is, manual control operation.The second method is to maintain the rotation speed at the time of vibration and increase the speed. The third method, which does not perform speed control, is to stop the turbine. The first method involves switching to manual control, leaving everything to the operator's judgment. Therefore, a skilled operator in that field is required, and there are differences in operation control due to individual differences, so this is by no means preferable. In the second method, there is a problem with the natural frequency when operating a rotating body, and it is not necessarily a good idea to hold the vibration in the rotating machine when vibration occurs. It is known that vibration increases in a critical speed range determined by the natural frequency of a rotating body consisting of a turbine and a generator directly connected to the turbine. Therefore, it is often very dangerous to uniformly maintain the rotational speed as in the second method. Next is the third method, which is safe for the rotating body, but may also cause unnecessary trips. Vibration factors include thermal imbalance, mechanical imbalance, lubricating oil condition,
Or there are many items such as rotation speed, and it is normal.
It is difficult to predict abnormalities. Therefore, in some cases, the vibration may converge to a safe vibration range by holding the rotation speed at a certain level as in the second method, so frequent use of the third method may result in unnecessary tripping. I end up having to do this. On the other hand, if the criteria for judgment are relaxed in an attempt to avoid this, there is a drawback that appropriate measures will be delayed in the event of an abnormality. However, what you need to be careful about here is to determine whether it is safe to avoid tripping and maintain that rotation speed. Therefore, it is not effective to maintain the rotation speed in the critical speed range, but only after determining what range the rotation speed is currently in and diagnosing what symptoms are appearing as vibration factors. Furthermore, among the vibration factors mentioned above, countermeasures against mechanical imbalance are taken in advance in an attempt to suppress the occurrence of vibration as much as possible. For example, JW Lund, J. Tonnesen: “Analysis
and Experiments on Multi−Plane Balancing
of a Flexible Rotor”, ASME Paper No.71
−Vibr−74 (ASME Vibration Conference,
Toronto, Canada, September 8-10, 1971). This paper describes a method for balancing by detecting vibrations at several locations on a rotating body, detecting the rotational speed, and calculating the corrected weight using the method of least squares. And for vibration measurement, shematic diagram of
It is shown in Fig. 4 as an instrumentation, and its explanation is given in the “Instrumentation” section (p. 3-4).
It is stated in Also, A. Clapis et al “Early Diagnosig of
Dynamic Unbalances and of Misalignment in
Large Turbogenerators”, Energy Nucleare,
Vol.23/n.5/maggio 1976, p271-277 shows an example of measuring the dynamic unbalance and axis misalignment of a large turbine generator and applying this as an early diagnosis method for failures. . In this paper, Part 2
two measurement methods, signal processing for monitoring and early diagnosis,
It describes the relationship between the amplitude and its phase due to unbalance of a rotating body with respect to rotational speed. In particular, Fig. 2 shows the installation state of the proximity switch, and Fig. 3 shows a block diagram of the signal processing. It is shown that the signal from the proximity switch is converted through a predetermined bandpass filter (BP). Here, we are talking about converting vibration phase signals, vibration effective value signals, etc. into DC components and recording them. In this way, in most cases conventionally the vibration signal is smoothed and treated as a DC component. However, these signal processing methods have a drawback in that, even when diagnosing an abnormality, for example, a large number of expected factors must be comprehensively diagnosed, making it difficult to perform detailed diagnosis. In addition, as an example of something actually applied to turbine start-up operation, “ACTUS” by FHBarratt et al.
An Automatically Controlled Turbine Run−
up System”AEI Engineering, September/
October, 1962, p.255-258. This describes the device ACTUS developed at AEI, for example "Problems of starting large steam turbine" on pages 255-256.
Section describes turbine speed-up operation.
Raising the turbine speed by monitoring for shaft eccentricity, vibrations, temperature differences between the steam and the metal outside of normal limits, and increasing the speed of the turbine at whatever speed it was at first. The method for controlling the speed to be maintained is clarified. This is the above-mentioned retention control method for abnormalities,
As already made clear, this is not necessarily a good idea. The conventional technology has been described above, but the controls based on the results of abnormality diagnosis include (1) switching to manual control, (2) maintaining the current rotation speed, and (3) tripping. Diagnosis is performed using an average value signal, that is, a DC signal. Although using an average value signal for signal processing is a simple method, it has the disadvantage that diagnosis cannot be made unless the signal changes as an average value. That is, in a case where a certain specific frequency component increases and another specific frequency component decreases, if the average value does not change as a result, it may be determined that no state change has occurred. However, even if the average value does not actually change,
There are many cases in which an increase in a certain frequency component must be considered a problem. Moreover, we know from experience that even if this itself does not immediately lead to an abnormality, a symptom of a certain abnormality appears. Therefore, when diagnosis is made using the average value of all frequency components of the vibration signal, there is a drawback that diagnosis cannot be made accurately and faithfully to the phenomenon. The present invention has been made with attention to this point, and its object is to perform safe operation control of a turbine generator, particularly starting operation control, based on an appropriate diagnosis that suits the actual situation. The following will be explained in order. The present invention relates to a method for controlling large rotating machines such as steam turbines and generators in thermal and nuclear power plants, and in particular detects vibrations during startup or steady operation, analyzes the detected vibration signals, and determines whether the machine is in normal operating condition. Alternatively, the present invention relates to a control method and apparatus for determining whether an abnormal state exists, detecting or predicting the cause of the abnormal state, and performing operational control such as speeding up, speeding down, or stopping a rotating machine. The main purpose of the present invention is to monitor predetermined vibration frequency components and diagnose the operating state of a rotating body, particularly a turbine generator. One object of the present invention is to perform operation control for controlling the speed of a turbine generator by monitoring at least one predetermined vibration frequency component in a plurality of predetermined operating rotational speed regions at the time of startup to diagnose the condition. We are here to provide you with a method. Another object of the present invention is to monitor specific frequency components, particularly at the time of startup, to perform predictive diagnosis of abnormalities and perform operational control such as speeding up, speeding down, and holding. A feature of the present invention is to diagnose the state of the rotating body by monitoring the state of at least one vibration frequency component that has a predetermined relationship with the rotation speed component of the turbine generator from among the detected vibration signals. It's about doing. Another feature of the invention is that the detected vibration signal:
The object of the present invention is to monitor and diagnose the conditions of a turbine and a generator based on the mutual correlation between a plurality of frequency components having a predetermined relationship with respect to rotational speed components. Another feature of the present invention is that a frequency component having a predetermined relationship with the rotational frequency component is generated before and after a critical speed region determined from the natural vibration of the rotating body from the start of the rotating body to the rated rotation speed. The purpose is to diagnose the condition and control the operation of the rotating body. Another feature of the present invention is that the rotational frequency component of the rotating body, its 1/2 harmonic component, odd harmonic component,
The purpose is to diagnose the state of harmonic components and control the starting operation of a rotating body. Another feature of the present invention is to perform spectrum analysis of all frequency components of the detected vibration signal to diagnose the rotational state of the turbine generator. Another feature of the present invention is that the vehicle is operated in accordance with a driving pattern consisting of four areas: a speed-up region, a holding driving region, a driving region in which the speed decreases to a speed below the critical speed closest to the relevant driving speed, and a trip region. There is a particular thing. Hereinafter, the present invention will be explained in detail with reference to Examples.
First, the basics of abnormality diagnosis by analyzing frequency components of the present invention will be described. With turbine generators that tend to have large rotational speeds and especially large capacities, it is extremely important to diagnose abnormalities early and take appropriate countermeasures. The problem is what kind of signal should be used to diagnose the abnormality, but here we will first briefly describe a method for analyzing vibration signals from a detector mounted on the bearing. FIG. 1 shows an example of a large turbine generator, in which a high pressure turbine HP, an intermediate pressure turbine IP, a low pressure turbine LP, and a generator are directly connected. Reference numerals 1 to 6 each represent a bearing, and in most cases, a vibration detector is attached to this bearing portion. An example of how the vibration detector is installed is shown in Fig. 2. 11 is a rotating body, 12 is a vibration detector, 13 is a sample hold circuit, 14 is an analog/digital converter (A/D converter), and 15 is an FFT (Fast
Fourier Transform) transformer. 101 of the signal lines are vibration amplitude analog signals, and 10
2 is a digital signal of vibration amplitude, and 103 is a vibration frequency full spectrum signal.The vibration detector in Fig. 2 is shown as a contact type, but if it detects the vibration of equipment, it can be a non-contact type. Of course, it may be an expression. By the way, FIG. 2 shows the case of FFT, which is suitable for cases where multiple factors are intricately related to each other, such as when diagnosing an abnormality in rotational speed. The reason for this is that converting a time-series signal into the frequency domain makes it easier to extract the features of the main cause, and recently there have been attempts to use fast Fourier transform offline for factory tests, installation adjustments, or when trouble occurs. It has been done. For example, in Japan Patent Office Publication No. 1411 of 1970, "Fast Fourier Transform Data Processing Apparatus" (assigned by IBM, corresponding U.S. Patent Application No. 52332, filed on July 6, 1970), Among numerical problem solving methods using discrete Fourier transforms, this paper specifically describes the hardware implementation of various fast Fourier transform algorithms for discrete Fourier transform calculations. This description has nothing to do with abnormal rotation speed diagnosis, but details of FFT are described. The details of FFT are referred to in this document, and will be briefly described below. As is well known, in the discrete Fourier transformation equation (1), the calculation is accelerated by row permutation of the Fourier matrix F and Good's factorization formula. The Fourier coefficient a is a = 1/NF. a ^T = [a ₀ , a ₁ , a ₂ , ...a _N-1 ] ...(2) F _i+1 , l=W ^il・ΔW ⁱ W=e×p(−j2π/N) ΔW= e x p (jπ/N) ...(3) where T: transpose i: frequency or harmonic number i, l = 0, 1, 2, 3, ..., (N-1) Time series signal X ^T is X ^T = [x ₀ , x ₁ , x ₂ , x ₃ , . . _. ΔW is a phase correction operator for analysis at the center of the sampling point. FFT converters tend to be commonly used because the entire frequency spectrum of a time-series signal can be easily obtained. However, when performing online diagnosis of turbines and generators, continuous constant monitoring is required, regardless of whether they are normal or abnormal. In order to be useful in responding to emergencies, it is required that the analysis results are directly linked to the analysis of the causes of the abnormality. In other words, in order to employ the conventional FFT algorithm for continuous monitoring and abnormality diagnosis,
Although it has the advantage that the spectrum of all frequencies of a time-series signal can be easily obtained as described above, it has the following disadvantages. The first is W(=exp(-
It is necessary to calculate the multiplication of j2π/N) 1/2Nlog ₂ N times, which is a heavy burden for the control computer to perform the calculation in real time. (Here, N is the number of samples.) Second, although the entire frequency spectrum can be obtained, it is necessary for the operator to make a judgment based on the distribution contents and analyze the factors, which is not possible in an emergency. To address the first drawback, there is a method to increase the speed by using FFT configured with dedicated hardware, but this method has the drawback of being more expensive than using software. If the hardware becomes cheaper, it will be more useful. In contrast, Fast Hadamard Transformer analyzes time-series input signals using a rectangular wave as a reference wave.
There is a method of performing frequency analysis (correctly, alternation number analysis) using FHT (hereinafter abbreviated as FHT). FHT is Walsh-Hadamard Transformer
(abbreviated as WHT). For example, regarding this frequency analysis method itself,
“BIFORE or
Hadamard Transform” (IEEE Transaction
on Audis and Electroacoustics, September
1971, p. 225-234). Figure 3 shows a block diagram when FHT is used. There are 16 FHT converters. 17 is a band average frequency converter 8 is a vibration diagnosis device. The input signal is a signal 104 with an alternating number coefficient A obtained by FHT-transforming a time-series digital signal, and a band frequency spectrum signal is obtained based on a predetermined linear conversion coefficient K, or a band frequency spectrum signal is obtained based on an average linear conversion coefficient. An average frequency spectrum signal 105 is obtained. K will be described further below. Here I will briefly discuss FHT. The FHT method that processes time-series input signals is
Output the alternation number coefficient A. This alternation coefficient A is
It is defined as below. A=N ^-1/2 T [G ₁ ] [G ₂ ]...[G _o ]X...(6) n=log ₂ N Here, alternation number coefficient: A ^T = [A ₁ , A ₂ , ..., A _N ] where T: transpose Permutation matrix: T [G ₁ ]=E ^(l-1) H ⁽¹⁾ E ⁽ⁿⁱ⁾ where indicates the Kronecker product (PRODUCT). Diagonal matrix: E ^(U) = 1 Hadamard matrix: H ⁽¹⁾ = 1 1 1 -1 By the way, Walsh and Fourier Transfer (hereinafter abbreviated as WFT) transforms alternating number coefficient A into Fourier coefficient a using equation (7). There is a method to convert it to . a=K·A...(7) Here, the linear conversion coefficient K is given by equation (8). K=1/NF・H...(8) As is clear from equation (6), only addition and subtraction operations are required, so the calculation process is the same as that of FFT, which requires complex number multiplication of sine or cosine. Comparing the calculation time, it is less than 1/10. This is particularly suitable for online abnormality diagnosis of rotating machines. For example, it is particularly effective in diagnosing turbine startup. F and H are N-order Fourier matrices and Hadamard matrices. That is, according to the FHT method, the alternation coefficient output A can be converted into Fourier coefficients representing frequency components by using the linear transformation coefficient K. Based on these basic matters, the present invention will be explained by examples. First, we will discuss the analysis of detected vibration signals and abnormality diagnosis. The present invention is based on the use of the above-mentioned FHT, but in order to achieve high speed, instead of finding all frequency coefficients a, an arbitrary band average frequency _p is found according to equation (9). It is something. Expressed numerically, it is as follows. _p = _p・A...(9) Here, the average linear transformation coefficient _p = [ _p , ₀ , _p , ₁ ,..., _p , _j ,...,
k _p , _N-1 ] _p , _j = 1/m _p 〓 ⁱ k _i , _j (i, j = 1, 2, . . . , N) However, m _p can be arbitrarily selected as Σ. The above-mentioned average linear conversion coefficient _p takes various values depending on the target event and conditions. Table 1 shows an example of selecting the average linear conversion coefficient _p for abnormal vibrations of rotating machinery that we have found empirically.

[Table] R: Rotation frequency component R _c : Natural frequency Table 1 shows the average linear conversion coefficients of turbines and generators in thermal nuclear power plants, which are high-speed rotating bodies.
An example of _p is shown. The causes of abnormal vibration are:
It is roughly divided into five types: bearing misalignment,
These are loosening of the coupling ring, self-excited vibration due to bearing oil film characteristics, thermal bending due to rubbing, and unbalanced rigidity. When we analyzed the frequencies of abnormal vibrations based on each of these factors, we found that each of them takes a specific frequency range. That is, it has been found that frequency components (spectrums) are generated depending on factors, such as subharmonic vibrations and harmonic vibrations. Therefore, by finding frequency components according to fractional harmonic components or harmonics, it is possible to know the state of the factors. Table 1 shows these relationships. I have already mentioned that FHT can significantly reduce processing time compared to conventional FHT. This is because it almost disappears. Next, the relationship between the FHT converter 16 and the band average frequency converter 17 will be explained in more detail. Fourth
The figure shows the vibration amplitude characteristics of an event, where the horizontal axis is time t and the vertical axis is vibration amplitude X. In the figure, time t ₀ ,
X ₀ for 8 sample times of t ₁ , ..., t ₇ ,
We are considering the number of samples of 8, X ₁ , ..., X ₇ . The number of samples and the vibration amplitude X are taken in through a sample hold circuit 13 and converted into a digital signal by an A/D converter 14. Figure 5 shows
3 is a diagram showing a detailed relationship between an A/D converter 14, an FHT converter 16, and a band average frequency converter 17. FIG. The band average frequency converter 17 includes a linear converter 70, a memory 71, a multiplier M, and an adder.
Consists of add. In FIG. 5, add indicates an adder and sub indicates a subtracter. The symbol in the subtracter means that the signal on this signal line is subtracted. The symbol □× M is a multiplier. The detailed configuration of the FHT converter 16 is shown in FIG. A/D converted vibration amplitude X
is stored in a memory (not shown) and then operates on element u. This calculation of element u is performed according to the following formula. u(0)=X ₀ +X ₄ u(1)=X ₁ +X ₅ u(2)=X ₂ +X ₆ 〓 〓 u(6)=X ₂ −X ₆ u(7)=X ₃ −X ₇ ... ...(10) Next, element u' is calculated based on u obtained by equation (10). The calculation of this element u' is performed by the following equation (11). u′(0)=u(0)+u(2) u′(1)=u(1)−u(3) 〓 u′(6)=u(4)−u(6) u′(7) =u(5)-u(7)...(11) By repeating the above calculation n times (n=log ₂ N, where N is the number of samples), the alternation number coefficient A(0), A
(1), ..., A(7) (general formula A(k)) is found. In the figure, add indicates an addition section and Sub indicates a subtraction section. The output A(0) of the FHT converter 16 in FIG. 6 above,
A(1), ..., A(7) are Fourier coefficients a ₀ , a ₁ , b ₁ , ...
…, b To convert to ₄ , use the conversion coefficient. An example of this conversion coefficient is shown in Table 2.

[Table] When the Fourier coefficients a and b are obtained using such conversion coefficients, the following equation (12) is obtained. In equation (12), a ₀ is a DC component, a ₁ , a ₂ , a ₃ , and a ₄ are sine wave components, and b ₁ , b ₂ , b ₃ , and b ₄ are cosine wave components. still,
In Table 2, even if the numbers after the fifth decimal point are rounded off, the values are practically the same. As a result, the conversion coefficients used in equation (12) are 1.3066, 0.5412, 1.4142,
There are five numbers: 0.5412 and 1.3066. Therefore, these five transform coefficients are stored, and when calculating the Fourier coefficients a and b, the Fourier coefficients a and b are calculated using equation (12). The above process of finding Fourier coefficients from alternating number coefficients is a combination of conventional FHT and linear transformation.In the present invention, instead of finding individual Fourier coefficients according to equation (12), It was decided to obtain the Fourier coefficient as an average. Therefore, the fifth
Referring back to the drawings, an embodiment of the present invention will be described. As a frequency band, DC component, average of sine wave,
Let's consider the average of the cosine waves. As is clear from the above explanation, the conversion coefficient of the DC component _a0 is 1, so _a0 =A(0)...(13). The average value of the sine wave is = a ₁ + a ₂ + a ₃ + a ₄ (here, a ₄ = 0) = 0.7654 A (2) + 1.4142 (4) + 1.8478 A (6) ... (14) . The average value of the cosine wave is: =b ₁ +b ₂ +b ₃ +b ₄ =1.8478A(1)+1.4142A(3) +0.7654A(5)+A(7) (15). However, 1.8478=1.3066+0.5412 0.7654=1.3066+0.5412 1.4142=1.4142...(16) Three pieces of data, 1.8478, 0.7654, and 1.4142, are required to obtain equations (14) and (15). These three pieces of data serve as conversion coefficients for the band average. In the band average frequency converter 17 shown in FIG. 5, the above three pieces of data are stored in the memory 71. Furthermore, the linear conversion unit 70 multiplies the data in the memory 71 and the number of alternations, then performs addition, and calculates equations (14) and (15) to obtain . According to this embodiment, the average frequency spectrum can be detected with a small memory capacity. The above description relates to sine and cosine components as average values, but average values of various harmonic components are also possible. Furthermore, it goes without saying that as the number of samples N increases, more accurate band averaging becomes possible. Generally, N=256, 512, 1024,...
etc. are practical. The general formula for converting the alternating number coefficient A into the Fourier coefficient a representing the frequency component (in the case of the number of samples N) is formula (17). In equation (17), R: rotational frequency component,
Rc: Indicates the natural frequency component. Further, to convert to a frequency spectrum, calculation is performed according to equation (18). ｜a _i ｜=10log ₁₀ √ ² _i + ² _i ……(18) For example, 1/2 rotation speed component P _1/2R is becomes. According to the embodiments described above, it has become possible to obtain the frequency characteristics as a band for the vibration factors shown in Table 1, so that the vibration factors can be properly grasped and
and countermeasures became possible. Next, the present invention can also be achieved by a computer.
A flowchart at that time is shown in FIG. First of all,
In the flow 501, the time series signal X ^T = [X ₁ , X ₂ , ...,
X _N ]. Next, in flow 502, the alternation number coefficient A is determined by A=N ^-1/2 T[G ₁ ]...[G _o ]·X. Next, based on the average conversion coefficient, the band average Fourier coefficient ^a _p is determined as ^a _p =K _p ·A. This process is the same as the process in the above embodiment.
Next, move on to check. That is, in flow 504, the band Fourier coefficient ^a _p reaches the allowable value L _p (abnormality determination level)
A check is made to see if it is within the range, and if it is satisfied, the process moves to flow 506, and a check is made to see if scanning of all the sample numbers has been completed, and if the scanning has not been completed, the process moves to flow 507. If it has finished, it waits for the next sample to be processed. flow 504
If you are not satisfied with the current state, follow flow 5.
05 to display. The checking process in the above process can also be applied to the case shown in FIG. Next, another embodiment of the present invention will be described. 8th
The figure shows an example of a case in which the conventional FHT and linear transformation described above are combined. Therefore, the conversion unit 17 is configured to perform the calculation of equation (12).
The greatest feature of this embodiment is that a comparing section 19 is provided between converters 16 and 17. Comparison section 19
has a number of comparators corresponding to the number of alternating coefficients. Each comparator receives as input the corresponding alternation number coefficient and a preset tolerance value A for each alternation number coefficient, and compares the two. The allowable value A _s of the alternation number coefficient is a reference value that can be recognized as abnormal if the value exceeds it depending on the cause of the abnormal vibration. The comparator compares each alternation number coefficient A _k with the tolerance value A _sk (A _s0 to A _s7 in this case), and if the difference from the tolerance value is smaller than a predetermined width M _p No output is output, and if it is large, it is considered abnormal and the alternation number coefficient A _k at that time is output. The converter 17 uses the alternation number coefficient determined to be abnormal to obtain Fourier coefficients related thereto. The result of this Fourier coefficient is checked again to see if it is abnormal. That is, in the above embodiment, a first abnormality check is performed on the alternation number coefficient, and if it is determined to be abnormal, the Fourier coefficient is determined and a second abnormality check is performed. Note that when only one alternation number coefficient is determined to be abnormal, other alternation number coefficients are also required for calculation in order to obtain the Fourier coefficient (a ₀ is an exception). Therefore,
The Fourier coefficients must also be obtained by taking in those alternating number coefficients. Although not shown in the figure, a control circuit checks the results of each of the comparators and checks whether the output of the comparator should be provided. FIG. 9 shows a flowchart of an embodiment implemented by a computer. flow 5
In 08, the deviation between the allowable value A _sk and each alternation number coefficient A _k is calculated. Next, in flow 509, it is checked whether the deviation ε is less than or equal to the value M _p .
If it is below, the process moves to flow 506, and if it is above, the process moves to flow 510 to obtain Fourier coefficients.
Next, it is displayed in flow 511. These embodiments also apply to the case of band averaging. Also at this time, a comparator 19 is provided between converters 16 and 17. Above, the processing of detected vibration amplitude signals has been mainly described, and next, the operational control of the turbine and generator based on the diagnostic results obtained by processing these signals will be described. FIG. 10 shows the overall block diagram.
This is an example in which a high pressure turbine 51 designated HP, an intermediate pressure turbine IP designated, a low pressure turbine 53 designated LP, and a generator 53 designated G are directly connected. 39 is a rotation speed detector, and 40 to 45 are vibration amplitude detectors each attached to a bearing. Reference numeral 48 denotes a vibration diagnosis device for diagnosing the rotating machine from vibration amplitude value signals, and 47 is a target rotation speed setting device, which converts the schedule signal from the start schedule device 46 for speed control into the vibration signal (detector) of the rotating body. 40-45
The correction is made according to the diagnostic results obtained from the signal detected by the sensor. 55 is a turbine speed control device that controls the speed of the turbine based on the setting signal 47; is the motor that operates the valve MSV-SV64, and MSV-SV is the motor that operates the auxiliary valve of the main blocking valve.
indicates the main blocking valve, and CV indicates the control valve. The turbine speed control device controls the main steam blocking bypass valve 64 according to the deviation between the signal 203 from the rotation speed detector 39 and the signal from the setting device 47, and controls the turbine 51 to reach a predetermined rotation speed. Controls the amount of steam inflow. I mentioned earlier that the signal from the startup scheduler is corrected according to the diagnosis results, but
Alternatively, it may be set in priority to the schedule signal to control the speed of the turbine. FIG. 11 shows a specific example of the configuration of the vibration diagnosis device 48. 216, 217 are band filters, 218
is a multiplexer, 219 is a sample holder,
220 is an A/D converter, 221 is a timing generation circuit that generates data in synchronization with rotation, 222 is a Fourier transformer, 223 is a discriminator that determines the state of the rotating body based on Fourier-transformed signals, and diagnostic device 224 is a Fourier transformer. It is composed of a converter 222 and a discriminator 223. The signal 204 is an analog shaft vibration signal, the signal 205 is a digital shaft vibration signal, the signal 206 is a harmonic spectrum signal, and the signal 207 is a target value setting signal based on the diagnosis result, which are input to the target rotation speed setting device 47. A signal source 208 is a switching signal source for a multiplexer, and the multiplexer is switched by this signal to sequentially switch and acquire a plurality of vibration detection signals. FIG. 12 shows an example of the waveform. Figure 1a shows an example of a shaft vibration signal detected by a vibration detector, such as signal 201 in Figure 11. Figure b shows the signal after passing through the bandpass filter, and it can be seen that harmonic components have been removed. For example, signal 204 in FIG. 11 corresponds to this. Further, c in the same figure is an output signal of the sample and hold circuit, and the signal 204' in FIG. 11 corresponds to this. Further, a in the same figure is a digital shaft vibration signal converted by an A/D converter, and the signal 205 in FIG. 11 corresponds to this. d shows an example in which the output signal 204' of the sample and hold circuit is converted into a digital signal of 11 bits (MSB is a sign bit). As mentioned above, these time-series digital signals are, for example, X ₀ in Fig. 6 in the case of FHT,
X ₁ , X ₂ , _...
Find (7). And the Fourier coefficients a ₀ , a ₁ , b ₁ ,...
... _b4 , for example, use the conversion coefficients shown in Table 2. Here, for simplicity, the number of samples N
In the case of =8 (X ₀ , X ₁ , X ₂ , ...X ₇ ),
Actually, as mentioned above, 256, 512, 1024, etc. are adopted. FIG. 13a shows the vibration amplitude situation at the time of startup of the turbine generator. It can be seen that there are several critical speed regions before the rated speed is reached, and the vibration amplitude increases in these regions. Generally, the primary critical speed is around 1000 rpm, the secondary critical speed is 2000 rpm, and the 3rd, 4th, and 5th critical speeds are around 3000 rpm.
It is around 3400rpm. FIG. 13b shows a typical speed increase pattern at startup. A to D correspond to A to D in a of the figure, respectively. In the present invention, frequency analysis of the vibration signal is performed in these retreat speed regions, and the status of the turbine generator is diagnosed based on the results of the full frequency spectrum analysis, or a certain relationship with the rotational speed component is determined as shown in Table 1. The system analyzes and diagnoses the situation of specific frequency components, and performs operational control. For example, FIG. 14 shows a case where diagnosis is performed by setting a threshold value. The horizontal axis is the rotational speed component, which is the ratio to the vibration amplitude of 100μ (peak to peak) [dB]
There are various ways to express it, such as in dB as a ratio to the over all frequency components, or as an absolute value. In this case, the vertical axis represents a frequency component that is 1/2 of the rotational speed component, but like the horizontal axis, it is often expressed in [dB] or as an absolute value. In FIG. 14, U is a lifting operation area, H is a holding operation area at the rotation speed, D is an operation area at a lower retracting speed, and T is a trip area for stopping operation. In other words, if it is in the U area, for example, the 13th
It is sufficient to continue increasing the speed according to the startup schedule (or speed-up schedule) shown in FIG. b. The activation schedule signal is provided, for example, from the activation schedule device 46 shown in FIG. If it is in the H region, holding operation is performed at that rotational speed for a predetermined holding time. If the engine is still in the H range even after a certain holding time, the engine speed is lowered to a lower, lower retraction speed and the engine is operated.
Here, the lower low-speed evacuation area is, for example, Fig. 13a.
If the vehicle is in the (C) range, this means driving at a reduced speed to the (B) speed range. This means that even in the H region, if the holding operation continues for a certain predetermined period of time, a speed-down operation similar to the operating state in the D region will be performed. As mentioned above, D is the 14th area in (C) area in Figure 13.
When reaching D in the diagram, the speed should be reduced to the region (B) in Figure 13. FIG. 14 shows the relationship between the rotational speed component and the 1/2 rotational speed component, and even if the 1/2 rotational speed component is small, if the rotational speed component increases, a trip may occur. As is clear from Table 1, this is a diagnosis that an abnormality due to vibration causes different from bearing misalignment, such as thermal bending due to rubbing, has occurred. 1st
Figure 4 shows the relationship between the rotation speed component and its 1/2 rotation speed component, but the vertical axis is the 1/3 rotation speed component, the critical speed component,
By monitoring the twice the rotational speed component, each vibration factor can be monitored and diagnosed. In summary, the operation control method based on abnormality diagnosis of the present invention will be explained with reference to the flow diagram shown in FIG. 15. Input processing step 301 inputs the rotational speed (rpm) of the turbine and generator and the shaft vibration signal X measured at each bearing (number of samples N x number of channels). Next, in the initial symptom extraction step 302,
The alternation number coefficient (A ₀ to _{A (N-1)} ) is determined according to equation (6). Next, in the abnormality determination step 303, (20)
Obtain the alternation number spectrum |A| according to the formula, and obtain the alternation number spectrum in normal operating condition |
The deviation from Anormal| is taken and the magnitude relationship with a predetermined small value L is determined. If |A|-|Anormal|L, the pattern speed at that time is A in Figure 13.
~Step 304~ Determine in which speed range of B
Determination is made in step 307. A to D are 1st, 2nd, 3rd...
This is a predetermined speed area that avoids a certain range around the 5th critical speed. Then, in steps 308-311, patterns I~ corresponding to each rotational speed region are selected. An example of the pattern is shown in Figure 15c. The pattern also has the same horizontal and vertical axes, but U, H, D,
The setting level of T will be different. Pattern I is
As shown in FIG. 15D, the vertical axis is 1/2R and 1/3R. In the case of pattern I, R _C does not need to be a problem because the speed is below the first critical speed. A pattern is a pattern, (FIG.15C)
The 2R component is added to the vertical axis in the case of . Assuming that the pattern speed is now in region B of FIG. 13, a pattern is selected (step 30).
9). In the next step 316 (FIG. 15B), the alternation coefficients subjected to the WFT process according to equation (17) are converted into Fourier coefficients a representing frequency components. Since the vertical axes at this time are 1/2R, 1/3R, and R _C as is clear from FIG. For example, calculate the elements a ₀ , b _1/3R , a _1/3R , b _1/2R , a _1/2R , b _Rc , a _Rc , and use the same formula for the rotational frequency component on the horizontal axis in FIG. 15C. (17) in b _R , a _R
It is sufficient to perform calculations on the corresponding elements of . In the next steps 317 to 320, as a result, in which region of the operation pattern U, H, D, and T shown in FIG.
Determine R, R _C and 1/3R. And the corresponding speed control (steps 321 to 324)
Do this. Specifically, the target rotation speed setting device 47 in FIG. 10 gives priority to the starting schedule signal predetermined in the target rotation speed setting device 47 as a setting value of the speed control device 55. Priority level is T
There is a relationship of >D>H>U, and the highest among the selected ones is operated. And step 3
Display of 13 vibrational spectra (e.g. Cathod
Ray Tube display) processing, etc., and determines whether processing has been completed for all channels (step 314), and if not, selects another channel (step 315) and repeats the same processing. . Here, the channel refers to the vibration detectors 39 to 45 in FIG. Further, in the above description, a case has been described in which a frequency component for a predetermined frequency component, for example, 1/2R is calculated, but a method of calculating an average spectrum by calculating before and after that frequency component may also be used. In that case, b _(1/2R+1) , b _(1/2R-1) ,
It is sufficient to perform calculations on b _(1/2R) , a _(1/2R+1) , a _(1/2R-1) , and a _(1/2R) . Alternatively, a method may be used in which an average value is calculated for a certain region before and after a specific frequency component. For example 1/2R
Considering a constant frequency range r as a reference (1/2R±r)
It may also be a method of performing calculations on . That is, in the example of equation (17), calculations are performed on b _(1/2R±r) and a _(1/2R±r) . i = 29, 30, 31 in the case of p = 1 in Table 1 is the so-called (1/2R ± 1)
This is an example of the case. Here, since m ₁ =6 is selected, an example is shown in which calculations can also be performed in the case of (1/4R±1), that is, the case of i=14, 15, and 16. In this way, diagnosing the condition of a rotating machine using the average frequency component including a region before and after the selected frequency, or a region before and after the selected frequency,
This has the effect of avoiding erroneous diagnosis due to noise signals from the vibration detector and compensating for sampling phase shifts. This can be expressed as an example of a pattern as shown in FIG. 15E. In addition, in FIG. 15A step 303 |
Although it takes the deviation from Anormal｜, it has the effect of absorbing the variation caused by the rotating machine. FIGS. 16a to 16d are drawings for explaining one of the effects of the present invention. Assume that the vibration amplitude signal at the current operating speed in figure a is as shown in figure b. Since the vibration amplitude of this waveform is small, no abnormality is detected by the conventional average value method, and eventually an oil whirl phenomenon occurs as shown in FIG. On the other hand, if frequency analysis is performed and the spectrum is monitored as in the present invention, signs of oil whirl can be detected in advance at point ▽a. The spectrum marked with ● in d of the same figure represents this. It can be seen that when the natural vibration frequency of the experimental machine is 22.5 (Hz), the spectrum at this frequency is particularly large compared to other frequencies. This means that signs of an oil whirl are appearing. Therefore, according to the present invention, even if the vibration amplitude value is small or the vibration amplitude average value is small, it is possible to predict abnormal vibrations that may occur in the future. Therefore, abnormal vibrations can be detected early and safe operation control can be performed. The vertical axis in Figure 16d is the peak of the input vibration waveform.
The frequency spectrum normalized by the −to−peak value is shown. Further, FIG. 17 shows an example of comparison between the case where the present invention is applied and the case where starting operation is performed using the conventional technique.
The case where the present invention is applied is shown by a solid line, and when an abnormality is detected at point ▽a, the engine is operated in a low speed evacuation area and then speeds up again to the rated rotational speed. On the other hand, in the conventional technology (broken line), as shown by point ▽b, the average value of vibration amplitude enters the trip region due to the delay in detecting the abnormality, so the trip occurs immediately.
After determining the cause of the abnormality, restart the system. In this manner, the present invention can detect an abnormality at an early stage, so that unnecessary trips can be prevented. Alternatively, abnormality may be determined by comparing and monitoring the deviation e between the measured alternation number spectrum A and the standard value A with a predetermined tolerance value. These relationships are shown in equations (20) to (22). ｜A｜=√ ² _2K + ² _2K+1 ……(20) However, k=0~(N/2−1) ｜A _s ｜=√ ² _s2K + ² _s2K+1 ……(21) e=｜ Ａ｜−｜A _s ｜ ...(22) Rotating bodies such as turbines and generators have a certain degree of residual unbalance due to manufacturing and implementation, so this becomes an excitation force when rotating, and even under normal conditions the rotational speed will decrease. Shaft vibration is generated around the component.
Therefore, even if an abnormal phenomenon occurs, during the initial period,
This normal vibration component becomes a disturbance that is difficult to detect. Therefore, if the normal vibration characteristics are memorized in advance and the deviations from this are monitored, abnormal characteristics can be detected with high sensitivity.However, in conventional monitoring in the time domain, the normal vibration characteristics It is difficult to match the phase of the waveform with the vibration waveform measured each time. In this embodiment, since the coefficients are converted into alternating number coefficients and compared with normal characteristics, there is no need to consider the influence of phase. Furthermore, although the embodiments of the present invention have been described in terms of frequency analysis using digital signals, an analog system may be used. However, in that case, a filter is required that can automatically change the time constant of the analog filter according to the rotational speed of the rotating machine. This is because, for example, as shown in FIG. 15C, it is necessary to analyze frequency components for frequencies that are in a relative relationship with the rotational frequency. However, the time constant of the natural vibration frequency filter, which is determined by the characteristics of the rotating machine, may be fixed. However, if an abnormality diagnosis is to be performed during operation at the rated rotational speed, a dedicated filter may be provided for each.

[Brief explanation of drawings]

FIG. 1 shows the switch in the coupled state of a large turbine generator. FIG. 2 shows how the vibration amplitude detector is mounted on the bearing and a block diagram of its signal processing. Figure 3 shows a block diagram of the vibration signal analysis method using FHT. FIG. 4 shows an example of vibration amplitude characteristics where the horizontal axis is time t and the vertical axis is vibration amplitude f. FIG. 5 shows a detailed diagram of the A/D converter, FHT converter and band average frequency converter. FIG. 6 shows a detailed flowchart of calculations by the FHT converter. FIG. 7 shows a flowchart when the FHT conversion calculation and linear conversion shown in FIG. 6 are performed by a digital computer. Figure 8 is as mentioned above.
FIG. 7 is a calculation flow diagram of an example in which the FHT calculation and linear transformation of FIG. 6 are combined. FIG. 9 is a calculation flow diagram when the calculation of FIG. 8 is performed by a digital computer. FIG. 10 shows a block diagram in the case where diagnosis is performed based on vibration and actual speed control of the turbine is performed. Figure 11 is Fig.
10 shows a more specific block diagram of the vibration diagnosis device shown in FIG. FIG. 12 shows an example of signal waveforms at each part shown in FIG. 11. FIG. 13a shows the relationship between the rotational speed and vibration amplitude of the rotary machine and the retracting speed range, and FIG. 13b shows the relationship between the startup schedule and the retracting speed range. FIG. 14 shows an example of a driving pattern based on the analyzed frequency components. Figure 15A,
B shows a flow diagram for explaining the vibration signal frequency analysis and operation control method, and C, D, and E show an example of an operation pattern. Figure 16a shows an example of the vibration amplitude due to foil whir Lubing with respect to rotation speed, b
is an example of the vibration amplitude signal at point a in the above a,
c shows an example of a vibration amplitude signal at point ▽b due to foil whirl, and d shows an example of a frequency spectrum at points ▽a and ▽b of a. FIG. 17 shows a comparison example between the present invention and the conventional technology at startup. 1 to 6...Bearing, 11...Shaft, 1
2... Vibration detector, 13... Sample hold circuit, 14... A/D converter, 15... FFT converter, 16... FHT converter, 17... Band average frequency converter, 46... Turbine Starting schedule control device, 47...Target rotation speed setting device, 48...
...Vibration diagnostic device.

Claims (1)

[Scope of Claims] 1. A method of detecting a vibration signal in the operating state of a rotating machine and frequency-analyzing the detected vibration signal to control the operation of the rotating machine, comprising: storing a plurality of sets of operation patterns for each rotational frequency component of the rotating machine, each consisting of at least a speed-up region, a holding operation region, a speed-down region, and a trip region, which are determined from a rotational frequency and a rotational frequency component having a specific relationship; A driving pattern corresponding to the detected rotational frequency of the rotating machine is selected from among the set of driving patterns, and at least the amplitude of the frequency component obtained by the frequency analysis is different from any one of the selected driving patterns. A rotating machine characterized in that, depending on whether the rotating machine belongs to a region, any one of the corresponding speed-up, holding, speed-down, and trip operations is commanded, and the operation of the rotating machine is controlled in accordance with the command. Operation control method based on abnormality diagnosis. 2. In claim 1, the plurality of sets of operating patterns have a rotational frequency of 1/1 of the rotational frequency of the rotating machine.
Abnormality diagnosis of a rotating machine, comprising patterns corresponding to each of the rotational frequency components corresponding to 2 and 1/3 and 2 times, and selecting a corresponding operation pattern according to the rotational frequency of the rotating machine. Operation control method. 3. In claim 1, the plurality of sets of operating patterns include patterns corresponding to the rotational frequency component corresponding to the natural vibration frequency component of the rotating machine, and when the rotating machine is at the natural vibration frequency, An operation control method by abnormality diagnosis of a rotating machine, characterized by selecting the above-mentioned operation pattern. 4. In claim 1, the plurality of sets of operating patterns include patterns each corresponding to a range of 1/2 times ±r of the rotational frequency of the rotating machine, and the plurality of sets of operating patterns each include patterns corresponding to a range of 1/2 times ±r of the rotational frequency of the rotating machine. An operation control method based on abnormality diagnosis of a rotating machine, characterized in that a corresponding operation pattern is selected based on the operating pattern. (However, r = 1, 2, . . . natural number) 5. In claim 1, each of the plurality of sets of operating patterns corresponds to a range of 1/3±r of the rotational frequency of the rotating machine. Contains a pattern
An operation control method based on abnormality diagnosis of a rotating machine, characterized in that a corresponding operation pattern is selected according to the rotational frequency of the rotating machine. (However, r = 1, 2, ... natural number) 6. In claim 1, the plurality of sets of operating patterns are patterns corresponding to a range of twice the rotational frequency of the rotating machine ±r. 1. An operation control method by abnormality diagnosis of a rotating machine, the method comprising: selecting a corresponding operation pattern according to the rotational frequency of the rotating machine. (However, r = 1, 2, . . . natural number) 7. In claim 1, the plurality of sets of operating patterns include patterns respectively corresponding to before and after r of the natural vibration frequency of the rotating machine, An operation control method based on abnormality diagnosis of a rotating machine, which selects a corresponding operation pattern according to the rotational frequency of the rotating machine. (However, r = 1, 2, ... natural number) 8. In claim 1, the detected vibration signal is subjected to FHT processing to calculate an alternating number spectrum, and is compared with a normal value of the spectrum. An operation control method based on abnormality diagnosis for a rotating machine, characterized in that when a deviation exceeds a predetermined value, an operation pattern suitable for the rotating machine is selected.