Search

CN-121977804-A - Vibration energy monitoring method and system for large steel structure rotary machine

CN121977804ACN 121977804 ACN121977804 ACN 121977804ACN-121977804-A

Abstract

The invention relates to the field of computer technology and large-scale machinery monitoring intersection, in particular to a vibration energy monitoring method and a vibration energy monitoring system for a large-scale steel structure rotating machine, comprising the following steps of S1, carrying out fusion acquisition of multi-source coupling vibration signals; S2, constructing and calculating local vibration energy of the rotating component and the steel structure respectively, analyzing and determining a coupling coefficient, quantifying coupling transfer energy, introducing energy attenuation coefficient to correct to obtain total vibration energy of the system, establishing a model parameter dynamic updating mechanism, S3, extracting time domain, frequency domain and energy domain characteristics of coupling vibration, reducing influence of environmental interference on signal quality, S4, constructing a dynamic energy base line, updating in real time, extracting multidimensional abnormal characteristics, inputting a fault diagnosis model to judge fault types and severity, and outputting corresponding level early warning information. The invention solves the problems of insufficient traditional monitoring coupling characterization, unbalanced precision real-time property and the like, and ensures the reliable operation of equipment.

Inventors

  • WANG LIDONG
  • DU YUE
  • WANG XINJIE

Assignees

  • 河西学院

Dates

Publication Date
20260505
Application Date
20260121

Claims (10)

  1. 1. A vibration energy monitoring method for a large steel structure rotating machine, comprising: S1, fusion acquisition of multisource coupling vibration signals, namely adopting a layered sensor array dotting scheme, combining high-precision clock synchronization and sensor self-adaptive configuration strategies, comprehensively acquiring vibration, rotating speed, stress and non-contact vibration data of a rotating part and a steel structure, and realizing synchronous fusion of multisource signals; S2, constructing a rotating part-steel structure coupling vibration energy calculation model, namely respectively calculating local vibration energy of the rotating part and the steel structure, determining a coupling coefficient through cross-correlation analysis, quantifying coupling transfer energy, introducing energy attenuation coefficient correction to obtain total vibration energy of the system, and building a model parameter dynamic update mechanism; S3, self-adaptive anti-interference signal processing, namely preprocessing, coupling characteristic enhancement and self-adaptive noise suppression are carried out on the acquired multi-source signals, the time domain, frequency domain and energy domain characteristics of coupling vibration are extracted, and the influence of environmental interference on signal quality is reduced; s4, based on the abnormality diagnosis and early warning of the energy characteristics, constructing a dynamic energy baseline, updating in real time, extracting multidimensional abnormal characteristics related to energy amplitude, trend and distribution, inputting a fault diagnosis model to judge the fault type and severity, and outputting corresponding level early warning information according to a preset threshold.
  2. 2. The method for monitoring vibration energy of a large steel structure rotating machine according to claim 1, wherein in S1, the specific steps of fusion acquisition of multisource coupled vibration signals are as follows: the method comprises the steps of installing a contact type high-precision acceleration sensor and a rotating speed sensor at key positions of a rotating part, installing a contact type acceleration sensor and a strain sensor at key force transmission paths of a steel structure, deploying a laser Doppler non-contact vibration meter and a machine vision sensor at a region inconvenient to install the contact type sensor, configuring a high-precision clock module for each sensor, synchronously recording a time stamp when signals are collected, carrying out real-time calibration on edge computing nodes through GPS/Beidou time service signals, and automatically matching the types, sampling frequencies and the collection duration of the sensors according to the types, the operation conditions and the monitoring precision requirements of the equipment.
  3. 3. The method for monitoring vibration energy of a large steel structure rotating machine according to claim 1, wherein in S2, the specific steps of constructing a coupled vibration energy calculation model are as follows: S21, calculating local vibration kinetic energy of the rotating part based on the vibration acceleration signal and the mass of the rotating part, calculating elastic strain energy and vibration kinetic energy of the steel structure by combining the strain signal and the steel structure material parameter, and summing to obtain local total energy of the steel structure; S22, calculating the correlation between the phase difference and the amplitude through a cross-correlation analysis algorithm by using vibration signals of the connecting part of the rotating part and the steel structure, determining a coupling coefficient based on the phase difference and the amplitude, and calculating coupling transmission energy according to an energy transmission efficiency formula; S23, combining local energy and transmitted energy, obtaining initial total energy by superposing and subtracting repeated calculation parts, and introducing energy attenuation coefficient to correct so as to obtain final system total vibration energy; S24, collecting equipment operation parameters in real time through a core processing module, and establishing a mapping relation between the parameters, the coupling coefficient and the attenuation coefficient based on a machine learning algorithm to realize real-time dynamic adjustment of the coefficients.
  4. 4. A vibration energy monitoring method for a large steel structure rotating machine according to claim 3, wherein the calculation formula of the local vibration kinetic energy of the rotating member is: ; wherein T is the acquisition time length, m is the mass of the rotating part, Is a vibration acceleration signal; the calculation formula of the elastic strain energy of the steel structure is as follows: ; In the formula, In order to monitor the volume of the region, As a stress signal, the stress signal is, Is a strain signal; The coupling coefficient calculation formula is: ; In the formula, For the amplitude dependence of the vibration signal, Is a phase difference; The total vibration energy correction formula of the system is as follows: ; In the formula, In order to couple the energy to be transferred, Is the energy attenuation coefficient.
  5. 5. The method for monitoring vibration energy of a large steel structure rotating machine according to claim 1, wherein in S3, the specific steps of adaptive anti-interference signal processing are as follows: S31, adopt Removing abnormal values by a criterion, removing direct current drift components by high-pass filtering, normalizing sensor signals of different measuring levels to the same dimension, and finishing signal preprocessing; s32, decomposing an original vibration signal into a plurality of intrinsic mode functions IMFs through a variational mode decomposition VMD, screening effective IMFs according to the matching degree of the center frequency of the IMFs and the characteristic frequency of the rotating part, and extracting the time domain, the frequency domain and the energy domain characteristics of coupled vibration through further decomposition of empirical wavelet transformation EWT; S33, adopting an improved convolutional neural network CNN-long-short-term memory network LSTM hybrid model, taking the preprocessed signal as input, taking a laboratory calibrated noiseless signal as a label training model, adaptively adjusting network parameters according to the signal noise intensity, and combining a feedback mechanism to realize noise suppression.
  6. 6. The method for monitoring vibration energy of a large steel structure rotating machine according to claim 1, wherein in S4, the specific steps of dynamic energy baseline construction and abnormal feature extraction are as follows: The normal operation data of the equipment, which are collected in the system initialization stage and are not less than 72 hours, are combined with rated parameters of the equipment, and a statistical analysis method is adopted to calculate the normal vibration energy range under different working conditions and serve as an initial dynamic energy baseline; in the running process of the equipment, the newly acquired normal data are fused into the baseline calculation through an incremental learning algorithm, so that the baseline is updated in real time; extracting a deviation rate of total energy from a baseline based on the processed vibration energy data Multi-dimensional abnormal characteristics such as energy peak value, energy fluctuation variance, energy change slope, energy accumulation quantity, coupling transmission energy duty ratio, energy duty ratio change rate of each frequency band and the like, Is the baseline energy value.
  7. 7. The method for monitoring vibration energy of a large steel structure rotating machine according to claim 1, wherein in S4, the specific steps of fault diagnosis and early warning are as follows: constructing a fault diagnosis model based on the combination of gradient lifting tree XGBoost and an attention mechanism, taking multidimensional energy abnormal characteristics as input, taking equipment fault type and severity as output, and training the model by simulating fault data and field historical fault data in a laboratory; Setting a multi-stage early warning threshold, respectively judging to be normal, primary early warning, secondary early warning and tertiary early warning according to the deviation rate of total energy and a base line and the abnormal trend, and pushing early warning information, wherein the multi-stage early warning threshold comprises And the abnormal trend is not normal, Or the slight abnormal trend is first-level early warning, Or the obvious abnormal trend is a secondary early warning, Or the detection of the definite fault features is three-level early warning.
  8. 8. A vibration energy monitoring system for large-scale steel structure rotating machinery is characterized by comprising a sensing layer, a transmission layer, a processing layer and an application layer, wherein the sensing layer, the transmission layer, the processing layer and the application layer are used for realizing full-flow vibration energy monitoring through cooperative work, the sensing layer consists of a layered sensor array, a high-precision clock module and a GPS/Beidou time service unit and is used for multi-source coupling vibration signal acquisition and clock synchronization, the transmission layer adopts a wired and wireless mixed transmission mode and comprises an edge computing node, a transmission module and a local buffer unit and is used for data preprocessing, real-time transmission and network disconnection buffering, the processing layer comprises a CPU (Central processing unit) server, a GPU (graphic processing unit) server and a relational and time sequence mixed database, the CPU server is responsible for system control and data management, the GPU server is used for vibration energy calculation, feature extraction and fault diagnosis, and the application layer comprises an interactive terminal, an alarm device and an application system and is used for monitoring data visualization, fault early warning and historical data analysis.
  9. 9. The vibration energy monitoring system for the large-scale steel structure rotary machine according to claim 8, wherein the sensor of the sensing layer has the characteristics of dust prevention, water prevention, high temperature resistance and electromagnetic interference resistance, the local buffer unit of the transmission layer can realize data buffer in a broken network state, automatic reissue after network recovery, the hybrid database of the processing layer supports efficient storage and quick query of massive monitoring data, and the visualization function of the application layer comprises the steps of displaying real-time change trends of energy of the rotary part, energy of the steel structure and total energy in a line graph, displaying energy distribution in a thermodynamic diagram, and displaying early warning level, fault position and processing state information in a billboard.
  10. 10. The vibration energy monitoring system for the large-scale steel structure rotating machine according to claim 8, wherein system software adopts a three-layer architecture of an edge end, a service end and a client end, the edge end is developed based on Python and used for data preprocessing and light weight calculation, the service end is developed based on JavaSpringBoot frames and used for data management and system control, an algorithm engine is developed based on Python+ TensorFlow and used for core algorithm realization, and the client end is developed based on Vue and used for visual interaction.

Description

Vibration energy monitoring method and system for large steel structure rotary machine Technical Field The invention relates to the technical field of computer technology and large-scale machine monitoring intersection, in particular to a vibration energy monitoring method and system for a large-scale steel structure rotary machine. Background The large-scale steel structure rotating machinery is used as core equipment in the key fields of metallurgy, wind power, chemical industry and the like, and bears important tasks of industrial production, energy supply and the like, the structural complexity is high, the operation load is large, the service environment is harsh, the steel structure is used as a support and conduction carrier, and the steel structure is coupled with vibration of a rotating part, once the vibration energy is abnormally accumulated, structural fatigue, component abrasion, even major accidents such as damage of the whole machine and the like are extremely easy to occur, namely, the coupling vibration of a large-scale wind turbine tower, a hub and a main shaft in the wind power field can cause the cracking of the tower and the breakage of blades, and the vibration coupling of the roller rotation of a large-scale steel rolling mill in the metallurgy field and the steel structure frame can influence the rolling precision of steel and shorten the service life of the equipment, so that accurate monitoring of the vibration energy is the key for guaranteeing the reliable operation of the equipment and reducing the economic loss and the safety risk. At present, the vibration monitoring technology of the large-scale rotary machine is mainly divided into two types, namely, contact type monitoring uses an acceleration sensor, a speed sensor and a displacement sensor as cores, vibration signals are collected through being directly installed on the surface of equipment, and the vibration monitoring technology is a main flow scheme of an industrial scene, and non-contact type monitoring is based on technologies such as laser Doppler vibration measurement, machine vision and the like, and is suitable for extreme environments such as high temperature, high pressure and high speed rotary parts and the like, in which the contact type sensor is inconvenient to install. In the aspect of signal processing, the prior art mainly adopts methods such as Fourier transformation, wavelet analysis, short-time Fourier transformation and the like to extract characteristic parameters such as vibration frequency, amplitude and the like, and aiming at rotating machinery containing large-scale steel structures, the prior monitoring system generally monitors the vibration of the steel structures and rotating parts separately, adopts a mode of multi-sensor point distribution and data fusion to initially sense a coupling state, and part of schemes are introduced into a Wireless Sensor Network (WSN) to solve the wiring problem or adopts a mode of finite element analysis and actual measurement data fusion to construct equipment dynamics model auxiliary analysis. The existing monitoring technology takes sensing acquisition, signal processing and diagnosis early warning as a core system, adopts a contact type and non-contact type mixed monitoring mode, data transmission is carried out to an edge node through a wired and wireless mixed architecture (industrial Ethernet, loRa and 5G modules), parameters such as a time domain peak value, a frequency domain characteristic frequency and the like are extracted after filtering and denoising pretreatment, signal analysis is based on Fourier transform and wavelet analysis, fault identification is realized by combining algorithms such as random forests and CNNs, the early warning mechanism is based on fixed threshold values, and a part of advanced schemes can construct dynamic threshold values through historical data to improve suitability. The prior art has a plurality of remarkable defects that firstly, vibration energy is not accurately depicted and the coupling effect is not considered sufficiently, the vibration energy is focused on single parameters such as vibration amplitude, frequency and the like, the vibration characteristics cannot be comprehensively depicted from the energy dimension, the generating, transmitting and accumulating processes of the energy cannot be accurately reflected, the complex coupling effect of a rotating part and a steel structure is only processed by adopting simple data splicing or weighting fusion, a coupling vibration energy transmission model is not established, the characteristic judgment deviation of the energy distribution is easily caused, and the misjudgment or omission judgment is caused; secondly, the monitoring precision and the real-time performance are difficult to be compatible, the contact type sensor is influenced by the environmental interference such as the installation position, the temperature, the electromagnetic noise and the like