Intelligent monitoring and diagnosis method and system for safety state of steam turbine

Abstract

This invention relates to the field of equipment condition monitoring technology, specifically to a method and system for intelligent monitoring and diagnosis of the safety status of steam turbines. The method includes the following steps: acquiring the voltage of the eddy current vibration sensor in the steam turbine bearing housing and performing analog-to-digital conversion; acquiring pulses from the speed probe to calculate the rotational frequency; segmenting the vibration amplitude sequence and performing frequency domain conversion; calculating the amplitude-frequency slope to identify the frequency of slope abrupt changes; and calculating the frequency and rotational frequency to identify the resonance frequency for alarm. In this invention, a vibration amplitude sequence is formed by voltage signal conversion and converted into a pulse signal. A correspondence is then established according to the sampling time, creating a unified temporal correlation between the vibration state and the rotor motion state. The associated frequency positions and corresponding frequency component amplitudes are organized according to frequency, presenting the energy distribution of the frequency positions. Combined with the rotational frequency, frequency harmonic judgment is performed and the resonance frequency is extracted. The number of frequency positions reaching the alarm condition is counted, expanding risk assessment from single-point alarm to multi-frequency point distribution judgment, thus enhancing the ability to locate anomalies and assess operating status.

Description

Technical Field

[0001] This invention relates to the field of equipment condition monitoring technology, and in particular to a method and system for intelligent monitoring and diagnosis of the safety status of steam turbines. Background Technology

[0002] The field of equipment condition monitoring technology involves a technical system for continuously observing and judging the operating status of rotating machinery. This technology mainly revolves around the acquisition of key physical quantities during equipment operation, the recording of operating status, and methods for determining status. By deploying speed sensors, vibration sensors, axial displacement sensors, and differential expansion measurement devices at key parts of the equipment, the speed, vibration amplitude, axial displacement, and rotor differential expansion changes generated during equipment operation are continuously collected. The collected operating parameters are recorded and compared using on-site monitoring devices. Based on the relationship between the equipment operating parameters and predetermined operating limits, the equipment status is monitored and judged. Traditional steam turbine safety condition monitoring refers to monitoring the rotor speed, vibration amplitude, axial displacement, and rotor differential expansion changes during steam turbine operation. This technical approach monitors and assesses the status of key operating parameters such as shaft vibration, axial displacement, and rotor differential expansion. Addressing the issue of abnormal operating parameters caused by equipment aging, component wear, and changes in operating conditions during long-term turbine operation, the traditional method involves installing eddy current vibration sensors at the turbine bearing housing to collect shaft vibration signals, installing speed probes at the shaft end to collect rotor speed, installing axial displacement sensors at the thrust bearing to measure rotor axial displacement, and arranging differential expansion measurement devices at the high-pressure and low-pressure cylinders to obtain rotor differential expansion data. Simultaneously, the monitoring system records the collected speed values, vibration amplitude, axial displacement, and differential expansion changes in real time, and determines the turbine's operating status based on pre-set alarm limits.

[0003] Current technologies, in actual operation, focus more on recording rotational speed, vibration amplitude, axial displacement, and differential expansion values ​​in parallel, and then directly comparing the recorded results with predetermined limits. While this approach can complete routine alarms, it lacks detailed correlation between parameters. In particular, when vibration data and rotor rotation status do not form a stable correspondence, on-site personnel often only see that the vibration is higher at a certain moment, but find it difficult to determine which frequency the anomaly occurs at, or to distinguish whether the amplitude change is caused by mechanical excitation, structural amplification, or fluctuations in operating conditions. When the equipment enters a complex load fluctuation range, the same alarm phenomenon may be driven by multiple frequency bands, and directly viewing the total amplitude record can easily mask local issues. Peaks in vibration can cause early signs of resonance to be masked by a wide range of background vibrations. For example, a significant spike may occur near a certain frequency, but the overall amplitude may not have exceeded the limit for an extended period. On-site assessments often tend to maintain observation, but as operation continues, local anomalies may expand into a risk of vibration across the entire machine. At the same time, traditional recording methods focus more on single instances of exceeding the limit, lacking detailed descriptions of the number of frequency locations, the distribution range of anomalies, and the boundaries of changes. During maintenance, it is difficult to quickly pinpoint key areas based on monitoring results, and maintenance arrangements tend to rely on experience, resulting in a delayed response time. This not only increases the cost of shutdown and troubleshooting but may also cause bearing housings, rotors, and related connecting structures to be subjected to repeated excitations for a long time, thereby amplifying the risk of wear, loosening, and fatigue accumulation. Summary of the Invention

[0004] To achieve the above objectives, the present invention adopts the following technical solution: a method for intelligent monitoring and diagnosis of the safety status of steam turbines, comprising the following steps: S1: Collect the voltage signal of the eddy current vibration sensor of the turbine bearing housing and convert it from analog to digital to generate a vibration amplitude sequence; collect the pulse signal of the turbine rotor speed probe and calculate the rotational frequency data; sort the vibration amplitude sequence according to the sampling time and associate it with the rotational frequency data to generate a turbine operation monitoring dataset. S2: Extract the vibration amplitude sequence of the turbine operation monitoring dataset, divide the vibration amplitude sequence into segments according to fixed time intervals, perform frequency domain transformation on each segment of the vibration amplitude sequence to calculate the associated frequency position and the corresponding frequency component amplitude, associate the associated frequency position and the corresponding frequency component amplitude with the vibration amplitude and sort them by frequency to generate the turbine vibration amplitude-frequency relationship sequence; S3: Extract the vibration amplitude changes of adjacent frequencies in the turbine vibration amplitude-frequency relationship sequence and calculate the amplitude-frequency slope value. Determine the amplitude-frequency slope value change range. When the amplitude-frequency slope value change exceeds the preset slope threshold, record the frequency position and generate a set of vibration slope change frequency abrupt changes. S4: Read the candidate frequency data of the vibration slope change frequency set, read the turbine rotor speed probe to calculate the rotational frequency data, calculate the doubling relationship between the candidate frequency data and the rotational frequency data and determine the integer doubling interval, record the frequencies that satisfy the doubling relationship, and generate the turbine structure resonance frequency sequence. S5: Extract the frequency position and frequency component amplitude data of the resonant frequency sequence of the turbine structure, compare the frequency component amplitude data with the turbine bearing housing vibration alarm threshold and record the frequency position that reaches the alarm condition, count the number of frequency positions that reach the alarm condition, and generate the number of times the turbine vibration alarm is triggered.

[0005] As a further aspect of the present invention, the turbine operation monitoring dataset includes a sampling time index, vibration amplitude identifier, and rotational frequency identifier; the turbine vibration amplitude-frequency relationship sequence includes a frequency sorting identifier, frequency component amplitude identifier, and amplitude-frequency correspondence identifier; the vibration slope abrupt change frequency set includes a slope abrupt change frequency identifier, slope change level identifier, and frequency candidate number identifier; the turbine structural resonance frequency sequence includes a resonance frequency identifier, harmonic relationship identifier, and structural response frequency identifier; and the turbine vibration alarm trigger count includes an alarm frequency count identifier, alarm distribution range identifier, and alarm trigger statistics identifier.

[0006] As a further aspect of the present invention, the segmented vibration amplitude sequence according to a fixed time interval includes dividing the vibration amplitude sequence sequentially according to a time window with a continuous time length of 5 seconds, and forming multiple continuous vibration amplitude subsequences according to the time sequence, each vibration amplitude subsequence containing no less than 256 vibration amplitude sampling points.

[0007] As a further aspect of the present invention, the step of performing frequency domain transformation to calculate the associated frequency position and corresponding frequency component amplitude for each vibration amplitude sequence includes performing a discrete Fourier transform operation on each vibration amplitude subsequence to obtain the associated frequency position and corresponding frequency component amplitude within the corresponding frequency range, and arranging the associated frequency position and corresponding frequency component amplitude in ascending order of frequency to form a frequency sequence.

[0008] As a further aspect of the present invention, the specific steps of S1 are as follows: S101: Acquire the output voltage signal of the eddy current vibration sensor of the turbine bearing housing and complete the analog-to-digital conversion to form a vibration amplitude sequence. At the same time, acquire the output pulse signal of the turbine rotor speed probe and calculate the rotation frequency data. Identify the period of the pulse signal and calculate the rotation frequency numerical sequence to obtain the vibration amplitude time sequence. S102: Extract the sampling time identifier according to the vibration amplitude time series and perform time order sorting. Arrange the frequency component amplitude data series by time index to form a time-sorted vibration amplitude sequence. At the same time, call the rotation frequency value sequence and perform time index matching to establish the association between the rotation frequency value and the corresponding sampling time position to obtain the vibration frequency association record table. S103: Based on the vibration frequency association record table, extract the vibration amplitude field and rotation frequency field and perform data pairing. Establish a correspondence between the frequency component amplitude record and the rotation frequency record according to the sampling time index. Perform sequence integration on all data corresponding to the sampling time index to obtain the turbine operation monitoring dataset.

[0009] As a further aspect of the present invention, the specific steps of S2 are as follows: S201: Extract the vibration amplitude sequence from the turbine operation monitoring dataset and segment it according to a fixed time interval. Call the vibration amplitude field and sampling time field of the turbine operation monitoring dataset, perform time index segmentation according to the preset time interval length, classify the continuous frequency component amplitude records into the corresponding time interval, and perform sequence recombination on the vibration records of the time interval to obtain the time period vibration amplitude sequence. S202: Extract vibration records for the time interval based on the vibration amplitude sequence of the time period, perform frequency domain transformation for each group of vibration records, calculate the frequency position associated with the frequency position and the corresponding frequency component amplitude, and perform sequence mapping between the frequency position identifier and the associated frequency position and the corresponding frequency component amplitude. Perform structured organization on the frequency records of the same time interval to obtain the sequence of associated frequency position and corresponding frequency component amplitude. S203: Extract the frequency position identifier based on the associated frequency position and the corresponding frequency component amplitude sequence, call the corresponding frequency component amplitude record of the time period vibration amplitude sequence, associate the associated frequency position and the corresponding frequency component amplitude with the vibration amplitude execution position, execute and arrange the associated records in frequency order to form a continuous frequency sorting structure, and generate the turbine vibration amplitude-frequency relationship sequence.

[0010] As a further aspect of the present invention, the specific steps of S3 are as follows: S301: Extract the vibration amplitude change relationship between adjacent frequency positions and calculate the amplitude-frequency slope value according to the turbine vibration amplitude-frequency relationship sequence. Call the frequency index field and vibration amplitude field of the turbine vibration amplitude-frequency relationship sequence, perform sequential retrieval on the frequency component amplitude records of adjacent frequency positions and calculate the amplitude change. Then, combine the frequency position interval to calculate the amplitude-frequency slope value, and sort the sequence according to the frequency index order to obtain the amplitude-frequency slope sequence. S302: Extract continuous amplitude-frequency slope records based on the amplitude-frequency slope sequence and perform change amplitude judgment, call the amplitude-frequency slope field and calculate adjacent slope change values, compare the slope change values ​​with the preset slope change threshold, extract the corresponding frequency index for slope records that meet the threshold conditions and perform position identification, sort all identified frequency indices in order to obtain abrupt frequency index sequence; S303: Extract all identified frequency index records according to the mutation frequency index sequence and perform set construction processing. Call the amplitude-frequency slope sequence frequency field to map the identified frequency index to the position. Perform sequence merging and order sorting on the mapped frequency records to form a candidate frequency set structure and generate a vibration slope mutation frequency set.

[0011] As a further aspect of the present invention, the specific steps of S4 are as follows: S401: Read candidate frequency data according to the vibration slope mutation frequency set, and at the same time obtain the rotational frequency data calculated by the turbine rotor speed probe. Call the frequency index field of the vibration slope mutation frequency set to extract candidate frequency records. Perform sequence sorting on the candidate frequency records to form a candidate frequency sequence. At the same time, collect the pulse signal of the turbine rotor speed probe and calculate the rotational frequency value. Perform time sequence sorting on the rotational frequency records to obtain a candidate frequency record table. S402: Extract candidate frequency records based on the candidate frequency record table and call the rotation frequency record. Perform frequency doubling relationship calculation on the candidate frequency and the rotation frequency. Perform integer frequency doubling interval judgment on the frequency records according to the frequency doubling relationship judgment rule. Mark the frequency index that meets the judgment rule and sort it in order to obtain the frequency doubling matching frequency sequence. S403: Extract the marked frequency index according to the frequency sequence of the frequency matching and call the frequency field of the candidate frequency record table to perform index mapping. Perform sequential arrangement and sequence sorting on the mapped frequency records. Organize all frequency records according to frequency order to generate the turbine structure resonance frequency sequence.

[0012] As a further aspect of the present invention, the specific steps of S5 are as follows: S501: Extract the frequency component amplitude data of the corresponding frequency position according to the resonant frequency sequence of the turbine structure, call the frequency index field of the resonant frequency sequence of the turbine structure to perform sequential retrieval, and perform position mapping in the frequency component amplitude record according to the frequency index. Perform sequence organization on the mapped frequency component amplitude record and form a frequency component amplitude record structure corresponding to the frequency, to obtain the resonant frequency amplitude record table. S502: Extract frequency component amplitude records based on the resonant frequency amplitude record table and call the turbine bearing housing vibration alarm threshold. Perform threshold comparison judgment on the vibration amplitude of the frequency position. Mark the frequency position of the frequency component amplitude records that meet the alarm judgment rules. Then, sort all the marked frequency positions in order to obtain the alarm frequency position sequence. S503: Extract all marked frequency position records according to the alarm frequency position sequence and perform quantity statistics. Calculate the count value for the frequency position marked records and sort the statistical records in order to generate the number of times the turbine vibration alarm is triggered.

[0013] The intelligent monitoring and diagnostic system for the safety status of steam turbines includes: The vibration acquisition module acquires the output voltage signal of the eddy current vibration sensor of the turbine bearing housing and completes analog-to-digital conversion to form a vibration amplitude sequence. At the same time, it acquires the output pulse signal of the turbine rotor speed probe and calculates the rotation frequency data. The vibration amplitude sequence is organized according to the sampling time sequence and a corresponding relationship is established with the rotation frequency data to obtain the turbine operation monitoring dataset. The amplitude-frequency construction module extracts the vibration amplitude sequence based on the turbine operation monitoring dataset and segments it according to a fixed time interval. It performs frequency domain transformation on the vibration amplitude sequence of the time interval to calculate the associated frequency position and the corresponding frequency component amplitude. It then organizes the associated frequency position and the corresponding frequency component amplitude with the corresponding vibration amplitude and arranges them according to frequency order to obtain the turbine vibration amplitude-frequency relationship sequence. The slope recognition module extracts the vibration amplitude change relationship between adjacent frequency positions and calculates the amplitude-frequency slope value based on the turbine vibration amplitude-frequency relationship sequence. It performs change amplitude judgment on continuous amplitude-frequency slope values. When the slope change exceeds the preset slope threshold, it records the corresponding frequency position and forms a candidate frequency set to obtain the vibration slope change frequency set. The frequency doubling determination module reads candidate frequency data based on the set of vibration slope abrupt change frequencies, and simultaneously obtains the rotational frequency data calculated by the turbine rotor speed probe. It performs frequency doubling relationship calculation on the candidate frequencies and rotational frequencies and determines whether they are in the integer frequency doubling range. The frequencies that satisfy the frequency doubling relationship are recorded to obtain the turbine structure resonance frequency sequence. The alarm statistics module extracts the frequency component amplitude data of the corresponding frequency position based on the resonant frequency sequence of the turbine structure, compares the frequency component amplitude data with the turbine bearing housing vibration alarm threshold, records the frequency position that meets the alarm condition, and performs statistical processing on the number of frequency positions that meet the alarm condition to obtain the number of times the turbine vibration alarm is triggered.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, a vibration amplitude sequence is formed by voltage signal conversion and pulse signal is calculated. Then, a corresponding relationship is established according to the sampling time, so that the vibration state and the rotor motion state form a unified temporal correlation. The associated frequency position and the corresponding frequency component amplitude and vibration amplitude are organized according to frequency to present the frequency position energy distribution. Combined with the rotation frequency, the frequency doubling judgment is performed and the resonance frequency is extracted. The number of frequency positions that meet the alarm conditions is counted, so that the risk assessment is expanded from single-point alarm to multi-frequency point distribution judgment, which enhances the ability of abnormal location and operation status assessment. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a schematic diagram of the steps of the present invention; Figure 2 This is a detailed schematic diagram of S1 of the present invention; Figure 3 This is a detailed schematic diagram of S2 of the present invention; Figure 4 This is a detailed schematic diagram of S3 of the present invention; Figure 5 This is a detailed schematic diagram of S4 of the present invention; Figure 6 This is a detailed schematic diagram of S5 of the present invention; Figure 7 This is a system module diagram of the present invention. Detailed Implementation

[0017] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0018] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0019] Please see Figure 1 This invention provides an intelligent monitoring and diagnosis method for the safety status of steam turbines, comprising the following steps: S1: Acquire the output voltage signal of the eddy current vibration sensor of the turbine bearing housing and complete the analog-to-digital conversion to form a vibration amplitude sequence. At the same time, acquire the output pulse signal of the turbine rotor speed probe and calculate the rotation frequency data. Organize the vibration amplitude sequence according to the sampling time sequence and establish a corresponding relationship with the rotation frequency data to obtain the turbine operation monitoring dataset. S2: Extract the vibration amplitude sequence from the turbine operation monitoring dataset and process it into segments according to fixed time intervals. Perform frequency domain transformation on the vibration amplitude sequence of the time interval to calculate the associated frequency position and the corresponding frequency component amplitude. Organize the associated frequency position and the corresponding frequency component amplitude with the corresponding vibration amplitude and arrange them in frequency order to obtain the turbine vibration amplitude-frequency relationship sequence. S3: Extract the vibration amplitude change relationship between adjacent frequency positions and calculate the amplitude-frequency slope value based on the turbine vibration amplitude-frequency relationship sequence. Perform change amplitude judgment on continuous amplitude-frequency slope values. When the slope change exceeds the preset slope threshold, record the corresponding frequency position and form a candidate frequency set to obtain the vibration slope change frequency set. S4: Read candidate frequency data based on the set of vibration slope change frequency, and at the same time obtain the rotational frequency data calculated by the turbine rotor speed probe. Perform harmonic relationship calculation on the candidate frequency and the rotational frequency and determine whether they are in the integer harmonic range. Record the frequencies that satisfy the harmonic relationship to obtain the turbine structure resonance frequency sequence. S5: Extract the frequency component amplitude data of the corresponding frequency position according to the resonant frequency sequence of the turbine structure, compare the frequency component amplitude data with the turbine bearing housing vibration alarm threshold and record the frequency position that reaches the alarm condition, and perform statistical processing on the number of frequency positions that reach the alarm condition to obtain the number of times the turbine vibration alarm is triggered.

[0020] The turbine operation monitoring dataset includes a sampling time index, vibration amplitude identifier, and rotational frequency identifier; the turbine vibration amplitude-frequency relationship sequence includes a frequency sorting identifier, frequency component amplitude identifier, and amplitude-frequency correspondence identifier; the vibration slope change frequency set includes a slope change frequency identifier, slope change level identifier, and frequency candidate number identifier; the turbine structural resonance frequency sequence includes a resonance frequency identifier, harmonic relationship identifier, and structural response frequency identifier; and the turbine vibration alarm trigger count includes an alarm frequency count identifier, alarm distribution range identifier, and alarm trigger statistics identifier.

[0021] Please see Figure 2 The specific steps of S1 are as follows: S101: Acquire the output voltage signal of the eddy current vibration sensor of the turbine bearing housing and complete the analog-to-digital conversion to form a vibration amplitude sequence. At the same time, acquire the output pulse signal of the turbine rotor speed probe and calculate the rotational frequency data. Perform analog-to-digital conversion on the voltage signal sampling sequence to form a frequency component amplitude data sequence. Perform period identification on the pulse signal and calculate the rotational frequency numerical sequence to obtain the vibration amplitude time sequence. First, an eddy current vibration sensor is fixed to the outside of the turbine bearing housing. The sensor output continuously generates an analog voltage signal, with the voltage range set to 0 volts to 10 volts, corresponding to a vibration displacement range of 0 mm to 2 mm. The acquisition device continuously reads the voltage signal at a sampling frequency of 10,000 Hz, forming a voltage sampling sequence every second. Then, analog-to-digital conversion is performed to convert the analog voltage into a digital vibration amplitude. The analog-to-digital conversion resolution is set to 16 bits, with the voltage range of 0 volts to 10 volts corresponding to a digital range of 0 to 65535. For example, if the voltage at a certain moment is 2.5 volts, the digital vibration amplitude is calculated as 16384 using a voltage ratio. If the voltage is 5 volts, the digital value is 32768. The 10,000 digital values ​​per second are arranged chronologically to form a vibration amplitude sequence. Simultaneously, the turbine rotor speed probe outputs pulse signals. The speed probe is installed beside the rotor toothed disc, which has 60 teeth; one pulse signal is generated for each tooth passed. The acquisition device counts the number of pulses within a 0.1-second time interval and converts this count to the rotational frequency. For example, if 150 pulse signals are recorded within 0.1 seconds, the rotor rotation count for that time interval is calculated by dividing the number of pulses by the number of teeth, resulting in 2.5 revolutions per second (RPS), or 25 RPS, and a rotational frequency of 25 Hz. If the number of pulses in the next time interval is 180, the corresponding RPS is 3, and the corresponding frequency is 30 Hz. After statistical analysis of consecutive time intervals, a rotational frequency numerical sequence is formed. The voltage signal sampling sequence undergoes analog-to-digital conversion and data processing, recording the frequency component amplitude data for consecutive time intervals according to timestamps. For example, sampling times are recorded as 0.0001 seconds, 0.0002 seconds, 0.0003 seconds, etc. The frequency component amplitude data is arranged in timestamp order. For example, the frequency component amplitude records for a certain time interval are 16384, 16420, 16350, and 16290, corresponding to sampling times from 0.0001 seconds to 0.0004 seconds. By combining all sampling times with the corresponding vibration amplitudes, a complete vibration amplitude time series is obtained.

[0022] Table 1. Raw data from sensor sampling 0.1 2.5 16384 150 25 0.2 2.6 17039 156 26 0.3 2.8 18349 162 27 0.4 3.0 19660 168 28 0.5 3.1 20316 180 30 As shown in Table 1, the sampling device records the vibration voltage and the number of pulses over a continuous time period. The vibration amplitude is obtained by voltage ratio conversion, and the rotation frequency data is obtained by pulse statistics, forming a vibration amplitude time series.

[0023] S102: Extract the sampling time identifier from the vibration amplitude time series and perform time order sorting. Arrange the frequency component amplitude data series by time index to form a time-sorted vibration amplitude sequence. At the same time, call the rotation frequency value sequence and perform time index matching to establish the association between the rotation frequency value and the corresponding sampling time position to obtain the vibration frequency association record table. The sampling time field is read line by line, and all sampling records are sorted in ascending order of time. The sampling time precision is set to 0.0001 seconds, and the time records are generated by the internal clock of the acquisition controller. For example, if the recorded data contains times of 0.001 seconds, 0.002 seconds, 0.003 seconds, 0.005 seconds, and 0.004 seconds, the sorting operation will adjust them to 0.001 seconds, 0.002 seconds, 0.003 seconds, 0.004 seconds, and 0.005 seconds. After sorting, the frequency component amplitude data is arranged according to the time index to form a time-sorted vibration amplitude sequence. Then, the rotational frequency numerical sequence is called, and time index matching is performed. The rotational frequency data recording time interval is 0.1 seconds, while the frequency component amplitude recording time interval is 0.001 seconds, so a time alignment operation is performed. Specifically, all frequency component amplitude records within each 0.1-second interval are grouped into the same rotational frequency record. For example, the time interval from 0 to 0.1 seconds corresponds to a rotation frequency of 25 Hz, and all 1000 frequency component amplitude records within this interval are labeled with a frequency of 25 Hz. The time interval from 0.1 to 0.2 seconds corresponds to a rotation frequency of 26 Hz, and all frequency component amplitude records within this interval are labeled with a frequency of 26 Hz. Then, a correspondence is established between the frequency component amplitude data and the rotation frequency values. A rotation frequency field is added to each frequency component amplitude record. For example, time 0.035 seconds corresponds to a vibration amplitude of 17020, and a rotation frequency of 25 Hz is recorded. Time 0.125 seconds corresponds to a vibration amplitude of 17560, and a rotation frequency of 26 Hz is recorded. Frequency association is completed through time matching.

[0024] Table 2 Correlation Record of Vibration Amplitude and Rotation Frequency 0.035 17020 25 0.036 17050 25 0.125 17560 26 0.126 17590 26 0.205 18100 27 Table 2 lists the time matching results of vibration amplitude and rotation frequency. Frequency correlation records are constructed by aligning time intervals.

[0025] S103: Extract the vibration amplitude field and rotation frequency field from the vibration frequency association record table and perform data pairing. Establish a correspondence between the frequency component amplitude record and the rotation frequency record according to the sampling time index. Perform sequence integration on all data corresponding to the sampling time index to obtain the turbine operation monitoring dataset. Read all sampling time index records from the associated record table, and combine the vibration amplitude and corresponding rotation frequency according to the same time index. For example, a sampling time of 0.035 seconds corresponds to a vibration amplitude of 17020 and a rotation frequency of 25 Hz; these two data points form a paired record. Then, perform sequence integration on all data corresponding to the sampling time indexes. During the integration process, records are read in chronological order, and the vibration amplitude and rotation frequency data are written sequentially into the running monitoring dataset. For example, there are 10 records within the time series from 0.0350 seconds to 0.0360 seconds. For instance, the following data is recorded in a certain time period: time 0.035 seconds, vibration amplitude 17020, rotation frequency 25 Hz; time 0.036 seconds, vibration amplitude 17050, rotation frequency 25 Hz; time 0.037 seconds, vibration amplitude 17010, rotation frequency 25 Hz; time 0.038 seconds, vibration amplitude 17080, rotation frequency 25 Hz. This sequential integration forms a continuous monitoring dataset. This dataset records the vibration amplitude and rotation frequency at each sampling time point. For example, if the monitoring duration is 60 seconds and the sampling frequency is 10,000 Hz, the number of frequency component amplitude records reaches 600,000, and the number of rotational frequency records is 600. After time matching and integration, each vibration record corresponds to one rotational frequency record, resulting in a complete turbine operation monitoring dataset.

[0026] Please see Figure 3 The specific steps of S2 are as follows: S201: Extract the vibration amplitude sequence from the turbine operation monitoring dataset and segment it according to a fixed time interval. Call the vibration amplitude field and sampling time field of the turbine operation monitoring dataset, perform time index segmentation according to the preset time interval length, assign the continuous frequency component amplitude records to the corresponding time interval, and perform sequence recombination on the vibration records of the time interval to obtain the time period vibration amplitude sequence. First, the vibration amplitude and sampling time fields are read from the monitoring dataset, with the time interval length set to 1 second. Then, all data is divided into intervals according to the sampling time. For example, data with sampling times from 0 to 1 second is assigned to the first time interval, and data from 1 to 2 seconds is assigned to the second time interval. If the sampling frequency is 10000 Hz, each time interval contains 10000 vibration records. Next, a time indexing operation is performed to classify the frequency component amplitude records according to the time interval. For example, the first time interval contains 10000 data records such as 16384, 16420, and 16350. The second time interval contains 10000 data records such as 17010, 17020, and 17030. Finally, sequence reconstruction is performed on the vibration records within the time intervals. Specifically, the frequency component amplitude records are arranged in the sampling order and a continuous array structure is created. For example, the vibration amplitude sequence for the first time interval is represented as arrays 16384, 16420, 16350, 16290, etc. The vibration amplitude sequence for the second time interval is represented as 17010, 17020, 17030, 17050, etc. For instance, when the monitoring dataset records for 60 seconds, 60 time-interval vibration amplitude sequences are obtained by dividing the dataset into time intervals, with each sequence containing 10,000 vibration records. After segmentation, the time-interval vibration amplitude sequence is obtained.

[0027] S202: Extract vibration records for time intervals based on the vibration amplitude sequence of the time period, perform frequency domain transformation for each group of vibration records, calculate the frequency position associated with the frequency position and the corresponding frequency component amplitude, and perform sequence mapping between the frequency position identifier and the associated frequency position and the corresponding frequency component amplitude. Perform structured organization on the frequency records of the same time interval to obtain the sequence of associated frequency position and corresponding frequency component amplitude. Frequency domain transformation is performed on the vibration sequence for each time interval. The vibration amplitude sequence is read and formed into continuous data segments according to the sampling order. For example, the vibration sequence of the first time interval contains 10,000 frequency component amplitude records. Then, frequency component calculation is performed on the sequence. Specifically, a Fast Fourier Transform (FFT) is performed on the time series data to calculate the vibration intensity corresponding to different frequency positions. For example, the sampling frequency is 10,000 Hz, and the frequency analysis range is 0 Hz to 5,000 Hz. The calculation results yield the vibration component value corresponding to each frequency position. For example, after transformation, the vibration data of the first time interval yields the following associated frequency positions and corresponding frequency component amplitudes: vibration component 1200 at 10 Hz, vibration component 1500 at 20 Hz, vibration component 2100 at 30 Hz, and vibration component 1700 at 40 Hz. Then, a sequence mapping is performed between the frequency position identifier, the associated frequency position, and the corresponding frequency component amplitude. Specifically, a frequency position field is created and the corresponding vibration component is recorded. For example, a frequency of 10 Hz corresponds to component 1200, and a frequency of 20 Hz corresponds to component 1500. The frequency records for the same time interval are structured and arranged in order of frequency, forming a sequence of associated frequency positions and corresponding frequency component amplitudes.

[0028] S203: Extract the frequency position identifier based on the associated frequency position and the corresponding frequency component amplitude sequence, call the corresponding frequency component amplitude record of the vibration amplitude sequence of the time period, associate the associated frequency position and the corresponding frequency component amplitude with the vibration amplitude execution position, execute and arrange the associated records in frequency order to form a continuous frequency sorting structure, and generate the turbine vibration amplitude-frequency relationship sequence. The process reads all frequency position fields from the associated frequency position and corresponding frequency component amplitude sequence. For example, frequencies of 10 Hz, 20 Hz, 30 Hz, and 40 Hz. Then, it retrieves the corresponding frequency component amplitude records from the time period vibration amplitude sequence, associating the associated frequency position with the corresponding frequency component amplitude and the vibration amplitude execution position. Specifically, it reads the peak vibration amplitude corresponding to the frequency position. For example, a frequency of 30 Hz corresponds to a peak vibration amplitude of 2100. The associated records are then sorted according to frequency order, arranged from smallest to largest frequency. For example, 10 Hz, 20 Hz, 30 Hz, and 40 Hz. For instance, the frequency records for a certain time interval are as follows: 10 Hz vibration amplitude 1200, 20 Hz vibration amplitude 1500, 30 Hz vibration amplitude 2100, and 40 Hz vibration amplitude 1700. This sorting forms a continuous frequency structure. The resulting continuous frequency sorting structure generates a turbine vibration amplitude-frequency relationship sequence.

[0029] Please see Figure 4 The specific steps of S3 are as follows: S301: Extract the vibration amplitude change relationship between adjacent frequency positions and calculate the amplitude-frequency slope value based on the turbine vibration amplitude-frequency relationship sequence. Call the frequency index field and vibration amplitude field of the turbine vibration amplitude-frequency relationship sequence, perform sequential retrieval on the frequency component amplitude records of adjacent frequency positions and calculate the amplitude change. Then, combine the frequency position interval to calculate the amplitude-frequency slope value, and sort the sequence according to the frequency index order to obtain the amplitude-frequency slope sequence. First, the frequency index field and vibration amplitude field of the vibration amplitude-frequency relationship sequence are read, and all frequency records are read one by one in ascending order. For example, if the frequency sequence records within a certain time interval are 10 Hz, 20 Hz, 30 Hz, 40 Hz, and 50 Hz, the corresponding frequency component amplitude records are 1200, 1500, 2100, 1700, and 1600 vibration amplitude units. Then, a sequential search operation is performed on the frequency component amplitude records of adjacent frequency positions, reading subsequent records one by one starting from the first frequency record to form adjacent frequency pairs. After the adjacent frequency pairs are formed, the change in vibration amplitude is calculated. The vibration amplitude at the next frequency position is read and the vibration amplitude at the previous frequency position is subtracted. For example, the amplitude change between a 10 Hz vibration amplitude of 1200 and a 20 Hz vibration amplitude of 1500 is 300; the amplitude change between a 20 Hz vibration amplitude of 1500 and a 30 Hz vibration amplitude of 2100 is 600; the amplitude change between a 30 Hz vibration amplitude of 2100 and a 40 Hz vibration amplitude of 1700 is -400; and the amplitude change between a 40 Hz vibration amplitude of 1700 and a 50 Hz vibration amplitude of 1600 is -100. Then, the frequency interval between adjacent frequency records is read. In the current implementation data, the frequency interval is 10 Hz. The amplitude-frequency slope value is calculated by performing a ratio between the amplitude change and the frequency interval. For example, a vibration amplitude change of 300 corresponds to a frequency interval of 10 Hz, resulting in an amplitude-frequency slope of 30 per Hz; a vibration amplitude change of 600 corresponds to a frequency interval of 10 Hz, resulting in an amplitude-frequency slope of 60 per Hz; a vibration amplitude change of -400 corresponds to a frequency interval of 10 Hz, resulting in an amplitude-frequency slope of -40 per Hz; and a vibration amplitude change of -100 corresponds to a frequency interval of 10 Hz, resulting in an amplitude-frequency slope of -10 per Hz. These slope results are then arranged in frequency index order to form an amplitude-frequency slope sequence. For example, within the range of 10 Hz to 50 Hz, the amplitude-frequency slope sequences are 30, 60, -40, and -10. If the frequency range within the monitoring period is extended to 0 Hz to 500 Hz, and frequency positions are recorded at 10 Hz intervals, then 50 amplitude-frequency slope records are obtained for each time interval. All slope records are then sequenced according to frequency order and written into the amplitude-frequency slope sequence record structure. The above calculations yield a complete amplitude-frequency slope sequence, which serves as the basis for subsequent frequency mutation identification.

[0030] S302: Extract continuous amplitude-frequency slope records based on the amplitude-frequency slope sequence and perform change amplitude judgment, call the amplitude-frequency slope field and calculate the adjacent slope change value, compare the slope change value with the preset slope change threshold, extract the corresponding frequency index for the slope record that meets the threshold condition and perform position identification, sort all the identified frequency indices in order to obtain the mutation frequency index sequence; All slope records in the amplitude-frequency slope sequence are read, one by one in frequency order. For example, records such as 30, 60, -40, -10, 15, and 20 are read. Then, the difference between adjacent slope records is calculated to obtain the slope change value. Specifically, the next slope value is read and the previous slope value is subtracted. For example, the change between 60 and 30 is 30, the change between -40 and 60 is -100, the change between -10 and -40 is 30, and the change between 15 and -10 is 25. The slope change value is then compared with a preset slope change threshold. The slope change threshold is obtained through statistical analysis of historical operating data. 120 sets of monitoring data from the stable operation phase of the steam turbine are selected, and the slope change value is calculated for each set of data, with the average change value statistically analyzed. The statistical results show an average change value of 38 vibration amplitude units per hertz. Subsequently, a safety margin adjustment is made to this data, setting the threshold to 80 vibration amplitude units per hertz.

[0031] Table 3. Statistical Table of Historical Running Slope Changes 1 42 2 35 3 50 4 38 5 44 6 33 7 41 8 39 As shown in Table 3, the maximum slope change values ​​in the historical operating data are all below 60, so the threshold is increased to 80 vibration amplitude units per Hertz.

[0032] A threshold comparison operation is then performed. When the absolute value of the slope change is greater than 80, it is marked as a slope abrupt change location. For example, when the slope change is -100, its absolute value is 100, which is greater than the threshold of 80, so the corresponding frequency position is marked. If this change occurs in the 20 Hz to 30 Hz range, then 30 Hz is recorded as the abrupt change frequency index. The same judgment is performed on all slope change records, recording the frequency indices that meet the threshold condition one by one. Finally, all identified frequency indices are sorted in ascending order. For example, the recorded results are 30 Hz, 90 Hz, and 150 Hz. This yields the abrupt change frequency index sequence.

[0033] S303: Extract all identified frequency index records based on the mutation frequency index sequence and perform set construction processing. Call the amplitude-frequency slope sequence frequency field to map the identified frequency index to the position, and perform sequence merging and order sorting on the mapped frequency records to form a candidate frequency set structure and generate a vibration slope mutation frequency set. Read all frequency records in the mutation frequency index sequence. For example, the mutation frequency index sequence obtained during a certain monitoring period is 30 Hz, 90 Hz, and 150 Hz. Then, the frequency field in the amplitude-frequency slope sequence is called to perform position mapping, and the amplitude record of the frequency component corresponding to each mutation frequency is read. For example, 30 Hz corresponds to an amplitude of 2100 vibration amplitude units, 90 Hz corresponds to an amplitude of 2400 vibration amplitude units, and 150 Hz corresponds to an amplitude of 2600 vibration amplitude units. Then, a sequence merging operation is performed on the mapped frequency records. During the merging process, mutation frequency indices generated from multiple time intervals are read and uniformly organized. For example, if a 30 Hz mutation is detected in the first time interval and another 30 Hz mutation is detected in the second time interval, only one 30 Hz record is retained in the set structure. If a 90 Hz mutation is detected in the third time interval, a new record of 90 Hz is added. This merging operation is continued for mutation frequencies in all time intervals. For example, if five consecutive monitoring intervals yield abrupt frequency records of 30 Hz, 30 Hz, 90 Hz, 150 Hz, and 90 Hz, a unique frequency set of 30 Hz, 90 Hz, and 150 Hz is obtained after merging. Then, a sequential sorting operation is performed, sorting all frequency records in the set according to their frequency magnitude. For example, the sorted result is 30 Hz, 90 Hz, and 150 Hz. The sorted data is written into a candidate frequency set structure. This forms a candidate frequency set and generates a vibration slope abrupt frequency set. This set records the frequency locations where vibration amplitude-frequency slope abruptly changes during operation and monitoring, and serves as candidate frequency input data for the subsequent resonance frequency identification process.

[0034] Please see Figure 5 The specific steps of S4 are as follows: S401: Read candidate frequency data based on the vibration slope mutation frequency set, and simultaneously obtain the rotational frequency data calculated by the turbine rotor speed probe. Call the frequency index field of the vibration slope mutation frequency set to extract candidate frequency records. Perform sequence organization on the candidate frequency records to form a candidate frequency sequence. At the same time, collect the pulse signal of the turbine rotor speed probe and calculate the rotational frequency value. Perform time sequence organization on the rotational frequency records to obtain a candidate frequency record table. The system reads all frequency records from the set and extracts them one by one. For example, the candidate frequency set records are 30 Hz, 90 Hz, 150 Hz, and 210 Hz. Then, the candidate frequency records are sorted sequentially, arranged in ascending order to form a candidate frequency sequence. Simultaneously, the rotational frequency data calculated by the turbine rotor speed probe is acquired. The speed probe is installed at the end of the turbine main shaft, on a toothed disk with 60 teeth. The acquisition device counts the number of pulses output by the probe within 0.1 seconds and converts them to the rotational frequency. For example, if 150 pulse signals are recorded in a certain time period, the rotor speed is 150 divided by 60, which equals 2.5 revolutions per second (RPM). Within 0.1 seconds, this corresponds to 25 RPM, so the recorded rotational frequency is 25 Hz. If 156 pulse signals are recorded in the next time period, the speed is 2.6 RPM, corresponding to a rotational frequency of 26 Hz. Then, all rotational frequency records are arranged in chronological order. For example, the recorded rotational frequency sequence for a certain operating phase is 25 Hz, 26 Hz, 27 Hz, 28 Hz, and 29 Hz. Next, the candidate frequency records and rotational frequency records are uniformly organized and written into the candidate frequency record table structure. For example, candidate frequency records are 30 Hz, 90 Hz, 150 Hz, and 210 Hz, while corresponding rotational frequency records are 25 Hz, 26 Hz, and 27 Hz. This organization forms the candidate frequency record table and provides the data foundation for subsequent frequency harmonic matching calculations.

[0035] S402: Extract candidate frequency records from the candidate frequency record table and call the rotating frequency record. Perform frequency doubling relationship calculation on the candidate frequency and the rotating frequency. Perform integer frequency doubling interval judgment on the frequency records according to the frequency doubling relationship judgment rule. Mark the frequency index that meets the judgment rule and sort it in order to obtain the frequency doubling matching frequency sequence. First, the frequency values ​​in the candidate frequency sequence are read one by one. For example, 30 Hz, 90 Hz, 150 Hz, and 210 Hz are read. Then, the rotation frequency records are read, such as 25 Hz, 26 Hz, and 27 Hz. Next, the harmonic relationship calculation is performed. Specifically, the candidate frequency is read and its ratio is calculated with the rotation frequency to determine if it is close to an integer harmonic. For example, the ratio of candidate frequency 150 Hz to rotation frequency 25 Hz is 6, therefore it is determined to be a 6th harmonic relationship; the ratio of candidate frequency 90 Hz to rotation frequency 30 Hz is 3, therefore it is determined to be a 3rd harmonic relationship. To ensure recognition stability, a harmonic determination interval is set. The harmonic determination interval is determined through statistical analysis of running data. Statistical results show that the normal harmonic deviation range is within ±1 Hz, therefore the harmonic determination interval is set to the integer harmonic frequency ±1 Hz. For example, when the rotation frequency is 25 Hz, the theoretical value of the 2nd harmonic is 50 Hz, so the determination interval is 49 Hz to 51 Hz. If the candidate frequency is 50 Hz, it is determined to be a harmonic matching frequency; if the candidate frequency is 52 Hz, the matching condition is not met. This judgment is then performed on all candidate frequencies. For example, 150 Hz and 25 Hz satisfy a 6th harmonic relationship, 90 Hz and 30 Hz satisfy a 3rd harmonic relationship, while the ratio of 210 Hz to 25 Hz is 8.4, which does not satisfy the integer harmonic condition. The frequencies that satisfy the harmonic determination rules are recorded, marked, and arranged in frequency order to obtain the harmonic matching frequency sequence.

[0036] S403: Extract the marked frequency index based on the frequency sequence of the frequency harmonic matching and call the frequency field of the candidate frequency record table to perform index mapping. Perform sequential arrangement and sequence sorting on the mapped frequency records, organize all frequency records according to frequency order, and generate the turbine structure resonance frequency sequence. Read all frequency records from the harmonic matching frequency sequence. For example, the matching frequency records are 50 Hz, 75 Hz, and 150 Hz. Then, read the vibration amplitude field and time index field from the candidate frequency record table and perform position mapping according to the frequency index. For example, read the frequency component amplitude record corresponding to the 50 Hz frequency (2200 vibration amplitude units), the frequency component amplitude record corresponding to the 75 Hz frequency (2500 vibration amplitude units), and the frequency component amplitude record corresponding to the 150 Hz frequency (2700 vibration amplitude units). Then, perform a sequential sorting operation on the mapped frequency records. The sorting method is based on the frequency value from smallest to largest. For example, the sorted frequency sequence is 50 Hz, 75 Hz, and 150 Hz. Next, write all frequency records into the resonance frequency sequence structure. If the same harmonic frequency is detected repeatedly in multiple time intervals, only one unique record is retained through merging. For example, if the 50 Hz frequency is detected in three consecutive time intervals, only one 50 Hz record is retained in the resonance frequency sequence. All the sorted frequency records are organized in frequency order to generate a turbine structure resonance frequency sequence.

[0037] Please see Figure 6 The specific steps of S5 are as follows: S501: Extract the frequency component amplitude data at the corresponding frequency position according to the turbine structure resonance frequency sequence, call the frequency index field of the turbine structure resonance frequency sequence to perform sequential retrieval, and perform position mapping in the frequency component amplitude record according to the frequency index. Perform sequence organization on the mapped frequency component amplitude record and form the frequency component amplitude record structure corresponding to the frequency, to obtain the resonance frequency amplitude record table. Read all frequency records from the resonant frequency sequence. For example, the resonant frequency sequence records are 50 Hz, 75 Hz, and 150 Hz. Then, call the frequency component amplitude record dataset and perform position mapping based on the frequency index. Specifically, retrieve the corresponding frequency position record in the vibration amplitude-frequency relationship sequence. For example, read the frequency component amplitude record 2200 vibration amplitude units at the 50 Hz position, 2500 vibration amplitude units at the 75 Hz position, and 2700 vibration amplitude units at the 150 Hz position. Then, perform sequence sorting on the mapped frequency component amplitude records. During sorting, the records are arranged in frequency order. For example, the sorted result is 50 Hz corresponding to vibration amplitude 2200, 75 Hz corresponding to vibration amplitude 2500, and 150 Hz corresponding to vibration amplitude 2700. The sorted records are written into the resonant frequency amplitude record table structure, with record fields including frequency position and frequency component amplitude data. The resonant frequency amplitude record table is then obtained.

[0038] S502: Extract frequency component amplitude records based on the resonance frequency amplitude record table and call the turbine bearing housing vibration alarm threshold. Perform threshold comparison judgment on the vibration amplitude of the frequency position. Mark the frequency position of the frequency component amplitude records that meet the alarm judgment rules. Then, sort all the marked frequency positions in order to obtain the alarm frequency position sequence. First, all frequency component amplitude records in the resonance frequency amplitude record table are read. For example, records show a vibration amplitude of 2200 units for 50 Hz, 2500 units for 75 Hz, and 2700 units for 150 Hz. Next, the turbine bearing housing vibration alarm threshold is read. The alarm threshold is obtained through statistical analysis of equipment operating standards and historical monitoring data. The statistical results show that the bearing housing vibration amplitude is below 1800 units during stable operation; therefore, the alarm threshold is set to 2000 units. Then, a threshold comparison operation is performed. When the vibration amplitude is greater than 2000 units, frequency position marking is performed. For example, a 50 Hz vibration amplitude of 2200 is greater than the threshold of 2000, so 50 Hz is marked as the alarm frequency position; a 75 Hz vibration amplitude of 2500 is also greater than the threshold of 2000, so this frequency is also marked; a 150 Hz vibration amplitude of 2700 is also greater than the threshold of 2000, so marking is performed. Then, all marked frequency positions are sorted in frequency order, for example, the sorted result is 50 Hz, 75 Hz, and 150 Hz. This yields the alarm frequency position sequence.

[0039] S503: Extract all marked frequency position records according to the alarm frequency position sequence and perform quantity statistics. Calculate the count value for the frequency position marked records and sort the statistical records in order to generate the number of times the turbine vibration alarm is triggered. First, all frequency records in the alarm frequency location sequence are read. For example, 50 Hz, 75 Hz, and 150 Hz are read. Then, a statistical operation is performed on each frequency record. This statistical operation is completed by reading all time interval alarm records in the operational monitoring dataset. For example, the detection results over 10 consecutive monitoring time intervals are as follows: 50 Hz alarmed 3 times, 75 Hz alarmed 4 times, and 150 Hz alarmed 2 times. Next, the statistical results are sorted sequentially. The sorted results are arranged in ascending order of frequency and the number of alarms is recorded. For example, 50 Hz alarmed 3 times, 75 Hz alarmed 4 times, and 150 Hz alarmed 2 times. Finally, the statistical results are written into the alarm trigger record structure, generating a turbine vibration alarm trigger count record, which represents the cumulative number of vibration alarms triggered at different resonant frequency locations within the monitoring period.

[0040] Please see Figure 7 The intelligent monitoring and diagnostic system for the safety status of steam turbines includes: The vibration acquisition module acquires the output voltage signal of the eddy current vibration sensor of the turbine bearing housing and completes analog-to-digital conversion to form a vibration amplitude sequence. At the same time, it acquires the output pulse signal of the turbine rotor speed probe and calculates the rotation frequency data. The vibration amplitude sequence is organized according to the sampling time sequence and a corresponding relationship is established with the rotation frequency data to obtain the turbine operation monitoring dataset. The amplitude-frequency construction module extracts the vibration amplitude sequence from the turbine operation monitoring dataset and segments it according to a fixed time interval. It performs frequency domain transformation on the vibration amplitude sequence of the time interval to calculate the associated frequency position and the corresponding frequency component amplitude. It then organizes the associated frequency position and the corresponding frequency component amplitude with the corresponding vibration amplitude and arranges them according to frequency order to obtain the turbine vibration amplitude-frequency relationship sequence. The slope recognition module extracts the vibration amplitude change relationship between adjacent frequency positions and calculates the amplitude-frequency slope value based on the turbine vibration amplitude-frequency relationship sequence. It performs change amplitude judgment on continuous amplitude-frequency slope values. When the slope change exceeds the preset slope threshold, it records the corresponding frequency position and forms a candidate frequency set, thus obtaining the vibration slope change frequency set. The frequency doubling determination module reads candidate frequency data based on the set of vibration slope abrupt change frequencies, and simultaneously obtains the rotational frequency data calculated by the turbine rotor speed probe. It performs frequency doubling relationship calculation on the candidate frequencies and rotational frequencies and determines whether they are in the integer frequency doubling range. The frequencies that satisfy the frequency doubling relationship are recorded to obtain the turbine structure resonance frequency sequence. The alarm statistics module extracts the frequency component amplitude data of the corresponding frequency position based on the resonant frequency sequence of the turbine structure, compares the frequency component amplitude data with the turbine bearing housing vibration alarm threshold, records the frequency positions that meet the alarm conditions, and performs statistical processing on the number of frequency positions that meet the alarm conditions to obtain the number of times the turbine vibration alarm is triggered.

[0041] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for intelligent monitoring and diagnosis of the safety status of steam turbines, characterized in that, Includes the following steps: S1: Collect the voltage signal of the eddy current vibration sensor of the turbine bearing housing and convert it from analog to digital to generate a vibration amplitude sequence; collect the pulse signal of the turbine rotor speed probe and calculate the rotational frequency data; sort the vibration amplitude sequence according to the sampling time and associate it with the rotational frequency data to generate a turbine operation monitoring dataset. S2: Extract the vibration amplitude sequence of the turbine operation monitoring dataset, divide the vibration amplitude sequence into segments according to fixed time intervals, perform frequency domain transformation on each segment of the vibration amplitude sequence to calculate the associated frequency position and the corresponding frequency component amplitude, associate the associated frequency position and the corresponding frequency component amplitude with the vibration amplitude and sort them by frequency to generate the turbine vibration amplitude-frequency relationship sequence; S3: Extract the vibration amplitude changes of adjacent frequencies in the turbine vibration amplitude-frequency relationship sequence and calculate the amplitude-frequency slope value. Determine the amplitude-frequency slope value change range. When the amplitude-frequency slope value change exceeds the preset slope threshold, record the frequency position and generate a set of vibration slope change frequency abrupt changes. S4: Read the candidate frequency data of the vibration slope change frequency set, read the turbine rotor speed probe to calculate the rotational frequency data, calculate the doubling relationship between the candidate frequency data and the rotational frequency data and determine the integer doubling interval, record the frequencies that satisfy the doubling relationship, and generate the turbine structure resonance frequency sequence. S5: Extract the frequency position and frequency component amplitude data of the resonant frequency sequence of the turbine structure, compare the frequency component amplitude data with the turbine bearing housing vibration alarm threshold and record the frequency position that reaches the alarm condition, count the number of frequency positions that reach the alarm condition, and generate the number of times the turbine vibration alarm is triggered.

2. The intelligent monitoring and diagnosis method for the safety status of a steam turbine according to claim 1, characterized in that: The turbine operation monitoring dataset includes a sampling time index, vibration amplitude identifier, and rotational frequency identifier; the turbine vibration amplitude-frequency relationship sequence includes a frequency sorting identifier, frequency component amplitude identifier, and amplitude-frequency correspondence identifier; the vibration slope abrupt change frequency set includes a slope abrupt change frequency identifier, slope change level identifier, and frequency candidate number identifier; the turbine structural resonance frequency sequence includes a resonance frequency identifier, harmonic relationship identifier, and structural response frequency identifier; the turbine vibration alarm trigger count includes an alarm frequency count identifier, alarm distribution range identifier, and alarm trigger statistics identifier.

3. The intelligent monitoring and diagnosis method for the safety status of a steam turbine according to claim 1, characterized in that: The segmented vibration amplitude sequence according to a fixed time interval includes dividing the vibration amplitude sequence sequentially according to a time window with a continuous time length of 5 seconds, and forming multiple continuous vibration amplitude subsequences according to the time sequence. Each vibration amplitude subsequence contains no less than 256 vibration amplitude sampling points.

4. The intelligent monitoring and diagnosis method for the safety status of a steam turbine according to claim 1, characterized in that: The step of performing frequency domain transformation to calculate the associated frequency position and corresponding frequency component amplitude for each vibration amplitude sequence includes performing a discrete Fourier transform operation on each vibration amplitude subsequence to obtain the associated frequency position and corresponding frequency component amplitude within the corresponding frequency range, and arranging the associated frequency position and corresponding frequency component amplitude in ascending order of frequency to form a frequency sequence.

5. The intelligent monitoring and diagnosis method for the safety status of a steam turbine according to claim 1, characterized in that, The specific steps of S1 are as follows: S101: Acquire the output voltage signal of the eddy current vibration sensor of the turbine bearing housing and complete the analog-to-digital conversion to form a vibration amplitude sequence. At the same time, acquire the output pulse signal of the turbine rotor speed probe and calculate the rotation frequency data. Identify the period of the pulse signal and calculate the rotation frequency numerical sequence to obtain the vibration amplitude time sequence. S102: Extract the sampling time identifier according to the vibration amplitude time series and perform time order sorting. Arrange the frequency component amplitude data series by time index to form a time-sorted vibration amplitude sequence. At the same time, call the rotation frequency value sequence and perform time index matching to establish the association between the rotation frequency value and the corresponding sampling time position to obtain the vibration frequency association record table. S103: Based on the vibration frequency association record table, extract the vibration amplitude field and rotation frequency field and perform data pairing. Establish a correspondence between the frequency component amplitude record and the rotation frequency record according to the sampling time index. Perform sequence integration on all data corresponding to the sampling time index to obtain the turbine operation monitoring dataset.

6. The intelligent monitoring and diagnosis method for the safety status of a steam turbine according to claim 1, characterized in that, The specific steps of S2 are as follows: S201: Extract the vibration amplitude sequence from the turbine operation monitoring dataset and segment it according to a fixed time interval. Call the vibration amplitude field and sampling time field of the turbine operation monitoring dataset, perform time index segmentation according to the preset time interval length, classify the continuous frequency component amplitude records into the corresponding time interval, and perform sequence recombination on the vibration records of the time interval to obtain the time period vibration amplitude sequence. S202: Extract vibration records for the time interval based on the vibration amplitude sequence of the time period, perform frequency domain transformation for each group of vibration records, calculate the frequency position associated with the frequency position and the corresponding frequency component amplitude, and perform sequence mapping between the frequency position identifier and the associated frequency position and the corresponding frequency component amplitude. Perform structured organization on the frequency records of the same time interval to obtain the sequence of associated frequency position and corresponding frequency component amplitude. S203: Extract the frequency position identifier based on the associated frequency position and the corresponding frequency component amplitude sequence, call the corresponding frequency component amplitude record of the time period vibration amplitude sequence, associate the associated frequency position and the corresponding frequency component amplitude with the vibration amplitude execution position, execute and arrange the associated records in frequency order to form a continuous frequency sorting structure, and generate the turbine vibration amplitude-frequency relationship sequence.

7. The intelligent monitoring and diagnosis method for the safety status of a steam turbine according to claim 1, characterized in that, The specific steps for S3 are as follows: S301: Extract the vibration amplitude change relationship between adjacent frequency positions and calculate the amplitude-frequency slope value according to the turbine vibration amplitude-frequency relationship sequence. Call the frequency index field and vibration amplitude field of the turbine vibration amplitude-frequency relationship sequence, perform sequential retrieval on the frequency component amplitude records of adjacent frequency positions and calculate the amplitude change. Then, combine the frequency position interval to calculate the amplitude-frequency slope value, and sort the sequence according to the frequency index order to obtain the amplitude-frequency slope sequence. S302: Extract continuous amplitude-frequency slope records based on the amplitude-frequency slope sequence and perform change amplitude judgment, call the amplitude-frequency slope field and calculate adjacent slope change values, compare the slope change values ​​with the preset slope change threshold, extract the corresponding frequency index for slope records that meet the threshold conditions and perform position identification, sort all identified frequency indices in order to obtain abrupt frequency index sequence; S303: Extract all identified frequency index records according to the mutation frequency index sequence and perform set construction processing. Call the amplitude-frequency slope sequence frequency field to map the identified frequency index to the position. Perform sequence merging and order sorting on the mapped frequency records to form a candidate frequency set structure and generate a vibration slope mutation frequency set.

8. The intelligent monitoring and diagnosis method for the safety status of a steam turbine according to claim 1, characterized in that, The specific steps of S4 are as follows: S401: Read candidate frequency data according to the vibration slope mutation frequency set, and at the same time obtain the rotational frequency data calculated by the turbine rotor speed probe. Call the frequency index field of the vibration slope mutation frequency set to extract candidate frequency records. Perform sequence sorting on the candidate frequency records to form a candidate frequency sequence. At the same time, collect the pulse signal of the turbine rotor speed probe and calculate the rotational frequency value. Perform time sequence sorting on the rotational frequency records to obtain a candidate frequency record table. S402: Extract candidate frequency records based on the candidate frequency record table and call the rotation frequency record. Perform frequency doubling relationship calculation on the candidate frequency and the rotation frequency. Perform integer frequency doubling interval judgment on the frequency records according to the frequency doubling relationship judgment rule. Mark the frequency index that meets the judgment rule and sort it in order to obtain the frequency doubling matching frequency sequence. S403: Extract the marked frequency index according to the frequency sequence of the frequency matching and call the frequency field of the candidate frequency record table to perform index mapping. Perform sequential arrangement and sequence sorting on the mapped frequency records. Organize all frequency records according to frequency order to generate the turbine structure resonance frequency sequence.

9. The intelligent monitoring and diagnosis method for the safety status of a steam turbine according to claim 1, characterized in that, The specific steps of S5 are as follows: S501: Extract the frequency component amplitude data of the corresponding frequency position according to the resonant frequency sequence of the turbine structure, call the frequency index field of the resonant frequency sequence of the turbine structure to perform sequential retrieval, and perform position mapping in the frequency component amplitude record according to the frequency index. Perform sequence organization on the mapped frequency component amplitude record and form a frequency component amplitude record structure corresponding to the frequency, to obtain the resonant frequency amplitude record table. S502: Extract frequency component amplitude records based on the resonant frequency amplitude record table and call the turbine bearing housing vibration alarm threshold. Perform threshold comparison judgment on the vibration amplitude of the frequency position. Mark the frequency position of the frequency component amplitude records that meet the alarm judgment rules. Then, sort all the marked frequency positions in order to obtain the alarm frequency position sequence. S503: Extract all marked frequency position records according to the alarm frequency position sequence and perform quantity statistics. Calculate the count value for the frequency position marked records and sort the statistical records in order to generate the number of times the turbine vibration alarm is triggered.

10. A steam turbine safety status intelligent monitoring and diagnostic system, characterized in that, The system is used to implement the intelligent monitoring and diagnosis method for the safety status of a steam turbine as described in any one of claims 1-9, and the system includes: The vibration acquisition module acquires the output voltage signal of the eddy current vibration sensor of the turbine bearing housing and completes analog-to-digital conversion to form a vibration amplitude sequence. At the same time, it acquires the output pulse signal of the turbine rotor speed probe and calculates the rotation frequency data. The vibration amplitude sequence is organized according to the sampling time sequence and a corresponding relationship is established with the rotation frequency data to obtain the turbine operation monitoring dataset. The amplitude-frequency construction module extracts the vibration amplitude sequence based on the turbine operation monitoring dataset and segments it according to a fixed time interval. It performs frequency domain transformation on the vibration amplitude sequence of the time interval to calculate the associated frequency position and the corresponding frequency component amplitude. It then organizes the associated frequency position and the corresponding frequency component amplitude with the corresponding vibration amplitude and arranges them according to frequency order to obtain the turbine vibration amplitude-frequency relationship sequence. The slope recognition module extracts the vibration amplitude change relationship between adjacent frequency positions and calculates the amplitude-frequency slope value based on the turbine vibration amplitude-frequency relationship sequence. It performs change amplitude judgment on continuous amplitude-frequency slope values. When the slope change exceeds the preset slope threshold, it records the corresponding frequency position and forms a candidate frequency set to obtain the vibration slope change frequency set. The frequency doubling determination module reads candidate frequency data based on the set of vibration slope abrupt change frequencies, and simultaneously obtains the rotational frequency data calculated by the turbine rotor speed probe. It performs frequency doubling relationship calculation on the candidate frequencies and rotational frequencies and determines whether they are in the integer frequency doubling range. The frequencies that satisfy the frequency doubling relationship are recorded to obtain the turbine structure resonance frequency sequence. The alarm statistics module extracts the frequency component amplitude data of the corresponding frequency position based on the resonant frequency sequence of the turbine structure, compares the frequency component amplitude data with the turbine bearing housing vibration alarm threshold, records the frequency position that meets the alarm condition, and performs statistical processing on the number of frequency positions that meet the alarm condition to obtain the number of times the turbine vibration alarm is triggered.

Images