CN101509604B - Method and device for detecting and assessing deposit in metal pipe - Google Patents

Abstract

The invention provides a method and device for detecting and assessing deposits in a metal pipe. The specific steps of the method are: 1) Tap the excitation part of the metal pipe with an excitation device to excite an impulse response characteristic that can reflect the structural body. 2) use a microphone and a data acquisition card to transmit and collect the excited sound wave signal; 3) record the sound with acquisition software and store it in the computer; 4) perform wavelet packet analysis, time domain analysis and/or spectrum analysis on the signal analysis, and calculate the characteristic parameters such as the energy value of each frequency band, the time domain signal duration and/or the maximum resonance frequency amplitude of the spectrogram in the wavelet packet analysis; 5) input the characteristic parameters into the neural network expert diagnosis system for identification, and determine whether there The accumulation is qualitative and quantitative, and the clogging degree in the tube is evaluated according to the quantitative results. The device used includes an excitation device, a microphone, and a computer with a data acquisition card. The microphone is fixed on a metal tube, and the microphone and the data acquisition card are respectively connected with the computer host.

Description

A method and device for detecting and evaluating deposits in metal pipes

technical field

The invention relates to a method for detecting and evaluating deposits in a metal pipe and a device used therein, and belongs to the technical field of non-destructive detection inside a pipe.

Background technique

With the development of industrial technology, various metal pipes and equipment are increasing day by day. To ensure its safe operation, frequent inspections are required. Due to the limitations of the pipeline and equipment operating environment, it is difficult to conduct internal inspection directly, which leads to various safety accidents. Peeling and clogging lead to a reduction in the flow area of the tubes, making the lower elbow of the furnace tubes in a state of overheating for a long time, resulting in a boiler tube burst accident. Therefore, it is urgent to use appropriate methods to accurately detect and evaluate the accumulation inside the elbow.

At present, there are roughly two methods for detecting deposits in metal pipes: one is to use radiography, because metal pipes and deposits in pipes are two substances with completely different structures. Absorbed dose is varied and thus will reflect on the negative film the state of accumulation of deposits. The method of using radiography is more intuitive, but its sensitivity is low, and it is difficult to identify a small number of accumulations from the image, especially when the thickness of the tube wall increases, the problem is more prominent, and it is easy to miss detection. For thin layers still attached to the pipe wall, especially homogeneous coatings, radiographic methods are almost impossible to detect. At the same time, due to the limitation of the use of radiation hazards, the radiographic flaw detection method affects other maintenance work, which is not conducive to shortening the maintenance period and affecting the economic benefits of the enterprise. Moreover, due to the limitation of the detection space, it is difficult to achieve comprehensive detection. The other is the industrial endoscopy method, which can effectively diagnose and predict failures of pipelines and equipment, and obtain real video images or image detection information inside the pipeline, which is more accurate and intuitive than the ray detection method. However, due to the complexity of the pipe structure and the size of the pipe diameter, the endoscope cannot detect every part; for those without an endoscope to enter the channel, pipe cutting inspection is required, which belongs to the category of destructive inspection.

Contents of the invention

In order to make up for the deficiencies of traditional ray detection and endoscopic detection methods, the present invention provides a method for detection and evaluation of deposits in metal elbows, and provides a device used in the method. The method and device can effectively solve the problem of Equipment failure or safety accidents caused by accumulations inside metal pipes, especially metal bends.

The technical solution for realizing the object of the present invention is: a method for detecting and assessing deposits in metal pipes, the specific steps of which are as follows:

1) Hit the excitation part of the metal pipe with the excitation device to excite the sound wave that can reflect the impulse response characteristics of the structure body;

2) Use a microphone and a data acquisition card to transmit and collect the excited sound wave signal;

3) Acquisition software records the sound and saves it in the computer;

4) Carry out wavelet packet analysis, time-domain analysis and/or spectrum analysis on the signal, and calculate the characteristic parameters such as the energy value of each frequency band, time-domain signal duration and/or maximum resonance frequency amplitude of the spectrogram in the wavelet packet analysis;

5) Enter the characteristic parameters into the neural network expert diagnosis system for identification, qualitatively and quantitatively determine whether there is accumulation in the pipe, and evaluate the degree of blockage in the pipe according to the quantitative results.

Among them, the specific steps of neural network expert diagnosis system identification are: 1) collecting training samples, including the time domain signal duration, the maximum resonance frequency amplitude of the spectrogram and/or the energy value and blockage degree of each frequency band analyzed by the wavelet packet; 2) constructing And train the network, input the sample data in step 1) into the constructed BP neural network, set the training parameters and train the network, when the mean square error of the training meets the requirements, the training will be automatically terminated; 3) adopt The trained network is used for diagnostic testing, that is, the characteristic parameters of the pipe to be tested are input to obtain the degree of blockage in the metal pipe.

The present invention also provides the device used in the detection and evaluation method of the accumulation in the above metal pipe. The device includes an excitation device, a microphone, and a computer with a data acquisition card. Acquisition software for signal recording is installed in the computer. Signal processing software for wavelet packet analysis, time domain analysis and/or frequency spectrum analysis and calculation of characteristic parameters and neural network expert diagnosis system, the microphone is fixed on the metal tube, and the microphone and data acquisition card are respectively connected to the host computer.

It can be seen from the above-mentioned technical scheme that the detection and evaluation method of deposits in metal pipes according to the present invention is an acoustic-vibration detection method, which is to measure the sound signal characteristics of the vibration by exciting the tested piece to generate mechanical vibration (sound wave). Non-destructive testing technology for judging the detected object. Due to the rapid development of computer and signal processing technology, the scientificity and accuracy of detecting and evaluating the deposits in metal pipes with vibroacoustic method can reach a very high level.

Compared with the prior art, the detection and evaluation method of deposits in metal pipes provided by the invention has the following main advantages:

(1) It can quickly determine whether there is accumulation in metal pipes, including metal bends, and quantitatively evaluate the degree of blockage in the pipes.

(2) It is not limited by the detection space and does not need to destroy the pipeline, which is better than the ray detection method and the endoscopy method.

(3) The operation is simple and the detection is accurate, which effectively solves the problem of equipment failure or safety accidents caused by accumulations inside the metal elbow. It has strong practicability and broad prospects.

Description of drawings

Fig. 1 is a schematic diagram of a detection device provided by the present invention.

Fig. 2 is a schematic diagram of the exciting part of the present invention.

Fig. 3 is a flow chart of diagnosing the blockage degree in the tube by using the BP neural network according to the present invention.

Detailed ways

The present invention will be described in further detail below in conjunction with the examples, but the present invention is not limited to the examples provided. The embodiments all take the metal elbow which is difficult to detect at present as the tested object.

Embodiment 1: as shown in Figure 1, detection device provided by the present invention comprises excitation device 1, microphone 2, computer 3 with data acquisition card, is installed with the acquisition software for signal recording in computer 3, is used for carrying out wavelet to signal Packet analysis, time domain analysis and/or spectrum analysis, and signal processing software and neural network experts who can calculate the characteristic parameters of wavelet packet analysis such as the energy value of each frequency band, time domain signal duration and/or maximum resonance frequency amplitude of spectrogram diagnostic system. The microphone 2 is fixed on the metal bent pipe and is located at the forward incident position of the sound wave, and the microphone 2 and the data acquisition card are respectively connected with the host computer. In order to excite the sound wave that reflects the impulse response characteristics of the structure body in the metal elbow, the excitation device in this embodiment 1 can use a steel hammer. The excitation part on the metal pipe should be the part that is easy to block the elbow. The lower part A of the lower elbow of the metal pipe is preferred. If the detection space is limited, the side B or upper part C of the lower elbow can be selected for excitation (see Figure 2).

Embodiment 2: The specific steps of detecting and assessing the accumulation in the metal elbow with the detection device of Embodiment 1 are as follows:

1) Hit the lower part A of the metal elbow with a steel hammer to excite a sound wave that can reflect the impulse response characteristics of the structure body;

2) Use a microphone and a data acquisition card to transmit and collect the excited sound wave signal;

3) Use the acquisition software to record the sound and store it in the computer;

4) Carry out wavelet packet analysis, time-domain analysis and/or spectrum analysis on the signal, and calculate the characteristic parameters such as the energy value of each frequency band, time-domain signal duration and/or maximum resonance frequency amplitude of the spectrogram in the wavelet packet analysis;

5) Enter the characteristic parameters into the neural network expert diagnosis system for identification, qualitatively and quantitatively determine whether there is accumulation in the pipe, and evaluate the degree of blockage in the pipe according to the quantitative results.

The flow chart of neural network expert diagnosis system identification is shown in Figure 3, and its specific steps are as follows:

1) Collection of training samples: including the time domain signal duration, the maximum resonance frequency amplitude of the spectrogram, the energy value of each frequency band analyzed by the wavelet packet and the degree of blockage.

2) Construct and train the network, input the sample data in step 1) into the constructed BP neural network, set the training parameters and train the network, and the training will be automatically terminated when the training mean square error meets the requirements.

3) The trained network is used for diagnostic testing, that is, the characteristic parameters of the tube to be tested are input, and the corresponding clogging degree in the tube can be obtained from the output vector value.

The BP neural network constructed in the present invention is a multi-layer feed-forward neural network, which can realize any nonlinear mapping from input to output. The BP neural network expert diagnosis system can be realized by using the existing MATLAB software platform, as long as the characteristic parameters of the tube to be tested are input, the corresponding blockage degree in the tube can be obtained from the output vector value.

The neural network expert diagnosis system in the present invention can use the energy value of each frequency band obtained by wavelet packet decomposition, the time domain signal duration obtained by time domain analysis and/or the maximum resonance frequency amplitude of the spectrogram obtained by spectrum analysis as characteristic parameters Diagnosis and identification, the energy value of each frequency band obtained by wavelet packet decomposition, the time domain signal duration, and the maximum resonance frequency amplitude of the spectrogram, these three characteristic parameters can be used for diagnosis and identification respectively, and can also be used in each frequency band obtained by wavelet packet decomposition Based on the energy value of the energy value as the characteristic parameter for diagnosis and identification, the time domain signal duration is used as the characteristic parameter for diagnosis and identification and/or the maximum resonance frequency amplitude of the spectrogram is used as the characteristic parameter for diagnosis and identification to achieve the purpose of improving accuracy. In this embodiment, only the energy value of each frequency band obtained by wavelet packet decomposition is used as the characteristic parameter as an example, and the BP neural network expert diagnosis system is used for diagnosis and identification.

In the present invention, 8 frequency band energies decomposed by the 3-layer wavelet packet are used as input nodes; the degree of pipeline blockage: smooth pipeline, small amount of blockage, more blockage, and severe blockage is used as output nodes to organize the neural network. See the table below for the selection of output nodes.

output unit 1 output unit 2 meaning 0 0 The pipeline is smooth 0 1 A small amount of blockage 1 0 more clogged 1 1 severe blockage

The samples for training the network are a set of data with known congestion status. When training the network, after setting the input vector of the training sample and the corresponding output vector, run the program in MATLAB. The BP network adopts the Levenberg-Marquardt optimization algorithm, and trainlm is used as Training function, the network stops automatically after reaching the set accuracy during training, and finally uses the trained network to test the samples to be tested.

The specific data of this embodiment are as follows:

Use 30 groups of known congestion levels to train the network, and test the collected 12 groups of signals with unknown congestion levels.

The 30 sets of data used for training are:

30.0495 18.4179 14.6217 9.604 1.2107 2.2465 12.8384 3.0631;

28.2124 16.9243 18.6037 9.3183 1.1545 2.2111 16.9503 3.1863;

30.6729 18.5736 15.4031 9.3168 1.2411 2.2405 13.7539 2.841;

31.5199 16.1455 8.6163 7.4801 1.0547 1.8589 7.2009 2.0322;

29.3098 16.693 20.538 9.1304 1.1516 2.1648 18.7914 3.1577;

33.1014 16.9648 5.8163 6.8013 1.062 1.7588 3.9724 1.7494;

34.1644 15.4385 9.8281 7.1821 0.9769 1.7512 8.5037 2.0858;

23.6146 12.0159 9.6093 7.2466 0.8723 1.6928 8.3176 2.6076;

22.1163 11.351 12.1397 6.7117 0.8908 1.5488 11 2.566;

22.7624 11.6882 12.5572 7.9192 0.8946 1.7671 11.2059 2.9013;

22.7271 1.612 10.9272 7.1139 0.8482 1.7065 9.7939 2.4573;

21.6667 11.113 11.095 7.3857 0.868 1.694 9.9033 2.7154;

19.0865 11.8117 10.6225 8.8218 0.8411 2.0432 9.3954 3.0273;

22.4018 11.6643 10.7664 7.1807 0.8929 1.7129 9.6013 2.6466;

18.4965 9.3241 6.894 4.9953 0.6971 1.2059 5.9303 1.7506;

20.6543 11.2943 8.6274 5.8514 0.8424 1.4129 7.4631 2.019;

20.3756 11.482 7.9329 5.9992 0.8106 1.4381 6.6095 2.1217;

18.7483 9.9208 8.626 5.6166 0.7624 1.2935 7.5942 2.045;

19.4512 10.5363 8.8685 5.2879 0.7718 1.3214 7.9413 1.8085;

20.0309 10.7921 7.8509 5.4841 0.8183 1.3939 6.7776 2.0888;

19.443 11.1617 7.6568 5.0871 0.7642 1.3334 6.5187 1.7238;

17.1874 9.8051 8.8757 4.5728 0.7267 1.1347 7.8838 1.7536;

16.9546 9.5353 7.6742 5.2076 0.7476 1.2378 6.7484 1.9521;

16.1874 9.2767 7.5292 4.7548 0.7098 1.2115 6.6599 1.7472;

16.6872 9.4724 6.3118 4.835 0.7002 1.1473 5.3364 1.6791;

17.526 10.0279 7.2432 5.27 0.7397 1.2601 6.2707 1.8005;

16.2897 9.2727 6.4617 4.9421 0.6802 1.1712 5.5427 1.6939;

16.3995 9.0821 7.4974 4.9787 0.7113 1.1775 6.6992 1.7883;

18.7414 10.4255 6.815 5.0967 0.7771 1.191 5.6654 1.6386;

16.8314 9.6273 6.9508 4.245 0.6977 1.0531 5.9724 1.5462

The 12 sets of data used for testing are:

29.6227 17.6801 19.46551 0.2675 1.2334 2.4326 17.6177 3.5355;

31.5177 18.698 15.7807 9.8632 1.2486 2.3671 14.071 3.2618;

29.766 18.0832 16.4785 9.5238 1.2275 2.288 14.8059 3.2531;

26.2162 13.9517 10.3812 7.1666 0.9572 1.6761 8.9609 2.5129;

25.6382 14.1059 8.1882 8.1704 0.9905 1.935 6.7521 2.8516;

22.6363 11.2832 7.2127 6.3094 0.8145 1.4425 5.9081 2.1353;

18.2451 9.5414 6.9163 4.9789 0.6904 1.1664 5.9753 1.6873;

20.387 10.4255 8.2459 5.0596 0.8114 1.248 7.2317 1.8474;

18.4716 9.6129 7.8849 4.9423 0.6834 1.1595 6.9823 1.6983;

16.2074 9.1581 8.0225 4.708 0.6868 1.177 7.0942 1.8232;

15.4307 8.5706 6.1938 4.4521 0.6819 1.1381 5.3196 1.6066;

15.3517 8.6671 6.112 4.6578 0.6668 1.0989 5.3329 1.5494

After the neural network diagnosis, the output of the neural network is:

Result analysis:

Comparing the obtained results with the actual state in the pipeline during the test, it was found that the fourth group of tests was wrongly diagnosed, the actual situation was "a small amount of blockage", and the neural network misjudged it as "unblocked". The results obtained in other states are consistent with the actual situation. It can be seen that the diagnosis and recognition accuracy of the method of the present invention is very high.

The present invention can also separately use the duration of the time-domain signal or the maximum resonance frequency amplitude of the spectrogram as the characteristic parameter for diagnosis and identification, but the test proves that the diagnosis and identification accuracy of the two is not as high as the energy value of each frequency band obtained by wavelet packet decomposition. The accuracy of diagnosis and identification of characteristic parameters is high. Through experiments, it can be seen that on the basis of using the energy value of each frequency band obtained by wavelet packet decomposition as the characteristic parameter for diagnosis and identification, the time domain signal duration is used as the characteristic parameter for diagnosis and identification and/or the maximum resonance frequency amplitude of the spectrogram is used as the Diagnosis and recognition based on characteristic parameters can improve the accuracy of diagnosis and recognition.

Claims (7)

1. the detection of a deposit in metal pipe and assessment method, its concrete steps are following:

1) knock the excitation position of metal tube with exciting bank, motivate the sound wave of the impulse response characteristic that can reflect structural body, wherein exciting bank is the steel tup, and the excitation position is that metal tube is prone to block part;

2) with microphone and data collecting card the acoustic signals that motivates is spread out of and gathers;

3) use the acquisition software recorded voice, and deposit computer in;

4) signal is carried out wavelet packet analysis, time-domain analysis and/or frequency analysis, calculate each band energy value, time-domain signal endurance and/or these characteristic parameters of spectrogram maximum resonant frequency amplitude of wavelet packet analysis;

5) characteristic parameter input neural network expert diagnostic system is discerned, qualitative and quantitative to whether there being deposit to carry out in managing, and based on chocking-up degree in the quantitative result evaluation pipe.

2. based on the detection and the assessment method of the said deposit in metal pipe of claim 1; The concrete steps that it is characterized in that the identification of neutral net expert diagnostic system are: 1) gather training sample, comprise each band energy value, time-domain signal duration and/or the spectrogram maximum resonant frequency amplitude and the chocking-up degree of wavelet packet analysis; 2) structure and training network are input to the sample data in the step 1) in the good BP neutral net of structure, network are trained after configuring training parameter, and training will stop automatically when the mean square error of training reaches requirement; 3) adopt the network that trains to carry out diagnostic test, promptly import the characteristic parameter of pipe to be measured, can obtain chocking-up degree in the metal tube.

3. according to the detection and the assessment method of the said deposit in metal pipe of claim 1, it is characterized in that: chocking-up degree is divided into unimpeded, a small amount of obstruction of pipeline, stops up more and seriously stops up four ranks.

4. according to the detection and the assessment method of the said deposit in metal pipe of claim 1, it is characterized in that: the excitation position is the lower portion of elbow under the metallic conduit.

5. the used device of the detection of the said deposit in metal pipe of claim 1 and assessment method; The computer that comprises exciting bank, microphone, band data collecting card; The acquisition software that is used for signal recording, signal processing software and the neuron network expert diagnostic system that is used for signal is carried out wavelet packet analysis, time-domain analysis and/or frequency analysis and calculated characteristics parameter are installed in the computer; Microphone fixing is on metal tube, and microphone links to each other with computer main respectively with data collecting card.

6. device according to claim 5 is characterized in that: microphone is positioned at the forward entrance position of sound wave.

7. device according to claim 5 is characterized in that: said characteristic parameter comprises each band energy value, time-domain signal endurance and/or the spectrogram maximum resonant frequency amplitude of wavelet packet analysis.

Images