A steel pipe bolt tightness state prediction method and system based on ultrasonic recognition

Abstract

The present application relates to steel pipe bolt loose state prediction technology field, especially in kind based on ultrasonic identification steel pipe bolt loose state prediction method and system. Content includes: obtaining multichannel ultrasonic reflection signal, and projecting to three-dimensional coordinate space, constructing ultrasonic energy attenuation distribution function;Based on ultrasonic energy attenuation distribution function, all multichannel ultrasonic reflection signal is mapped to four-dimensional space-time disturbance tensor field, and decoupling processing is carried out, and the disturbance energy offset rate is obtained;The disturbance energy offset rate is uniformly expressed as state vector, and a nonlinear time series recursive model is constructed to obtain the predicted state vector;Based on the predicted state vector, a double branch evaluation mechanism is introduced, state classification and pretightening force regression prediction are carried out. The traditional steel pipe bolt loose state prediction technology is not accurate for ultrasonic signal processing analysis, the high-dimensional space disturbance characteristics caused by bolt loosening cannot be obtained, which leads to low recognition accuracy and poor stability.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of steel pipe bolt tightness prediction, and particularly relates to a steel pipe bolt tightness prediction method and system based on ultrasonic identification. BACKGROUND

[0002] At present, in steel structure engineering, as a connecting key component, the pre-tightening force state of a bolt directly relates to the stress performance and operation safety of the overall structure. In particular, in a steel pipe structure, once the bolt is loose, the pre-tightening force is insufficient or the connection fails, it is easy to cause serious problems such as structural deformation, node cracking and even overall instability. Traditional bolt loosening detection methods mainly include artificial torque re-inspection, mechanical knocking and listening, dye penetration, magnetic powder detection and strain gauge installation and the like. However, these methods generally have problems such as dependence on manual operation, complicated operation, low detection efficiency, strong subjectivity of data and the like, and are difficult to cover high-altitude, closed or densely arranged areas, and cannot meet the needs of real-time, accuracy and automation level of complex structure systems. In recent years, with the development of non-destructive testing technology, ultrasonic detection has been gradually applied to the state identification of bolt connection parts due to its strong penetration, high sensitivity and non-contact advantages. However, the existing ultrasonic detection is mostly focused on the identification of defects such as weld cracks and thickness corrosion, and the identification of bolt connection loosening state mostly only stays at the level of simple features such as energy change or echo time difference, and a complete physical model between ultrasonic propagation and connection state has not been established, which is difficult to cope with complex ultrasonic signal interference under different structure forms and different working environments. In addition, the existing ultrasonic signal processing methods mostly use linear or low-dimensional means such as traditional filtering, Fourier transform and wavelet analysis, which cannot extract high-dimensional spatial disturbance features formed by bolt loosening state under multi-angle, multi-direction and multi-channel interference, limiting the recognition accuracy and system stability. At the same time, most of the existing state identification methods rely on static threshold judgment or shallow neural network classification, lack dynamic modeling ability of bolt connection state in time series, cannot realize the prediction of state evolution trend, and have weak generalization ability, which is difficult to adapt to complex scene deployment.

[0003] In summary, the traditional steel pipe bolt tightness prediction technology still has the technical problems of inaccurate ultrasonic signal processing and analysis, inability to obtain high-dimensional spatial disturbance features caused by bolt loosening, low recognition accuracy and poor stability. SUMMARY

[0004] The present application provides a steel pipe bolt tightness prediction method and system based on ultrasonic identification to solve the technical problems of inaccurate ultrasonic signal processing and analysis, inability to obtain high-dimensional spatial disturbance features caused by bolt loosening, low recognition accuracy and poor stability of the traditional steel pipe bolt tightness prediction technology.

[0005] The application discloses a steel pipe bolt tightness state prediction method and system based on ultrasonic recognition.

[0006] A steel pipe bolt tightness state prediction method based on ultrasonic recognition comprises the following steps.

[0007] S1. Obtain multi-channel ultrasonic reflection signals and project them to a three-dimensional coordinate space to quantize energy diffusion of ultrasonic pulses in space, and construct an ultrasonic energy attenuation distribution function; based on the ultrasonic energy attenuation distribution function, map all multi-channel ultrasonic reflection signals to a four-dimensional space-time disturbance tensor field to obtain ultrasonic disturbance energy amplitude, and perform decoupling processing to obtain a disturbance energy shift rate.

[0008] S2. Represent the disturbance energy shift rate as a state vector, construct a nonlinear time series recursive model, and obtain a predicted state vector; based on the predicted state vector, introduce a double-branch evaluation mechanism, perform state classification and pretightening force regression prediction, and realize steel pipe bolt tightness state prediction.

[0009] Preferably, S1 specifically comprises the following steps.

[0010] Signal emission and reception are performed on the target bolt by a phased array ultrasonic sensor group to obtain multi-channel ultrasonic reflection signals; the phased array ultrasonic sensor group is composed of ultrasonic sensing units.

[0011] Preferably, S1 specifically comprises the following steps.

[0012] Based on actual layout position coordinates of the ultrasonic sensing units, an ultrasonic energy attenuation distribution function is constructed in combination with an energy diffusion coefficient; based on the ultrasonic energy attenuation distribution function, all multi-channel ultrasonic reflection signals are superimposed and mapped to a four-dimensional space-time disturbance tensor field to obtain ultrasonic disturbance energy amplitude.

[0013] Preferably, S1 specifically comprises the following steps.

[0014] Based on the ultrasonic disturbance energy amplitude, a stationary energy reference state tensor is obtained, a direction modulation factor is introduced, and a signal period modulation term is combined to quantize the degree of echo energy deviation from a stable state in the main propagation direction to obtain the disturbance energy shift rate.

[0015] Preferably, S2 specifically comprises the following steps.

[0016] Based on the state vector, double-layer nonlinear transformation is performed, a nonlinear energy term of state disturbance is combined to construct a nonlinear time series recursive model, and a predicted state vector is obtained through nonlinear recursive rules.

[0017] Preferably, S2 specifically comprises the following steps.

[0018] The double-branch evaluation mechanism comprises a classification branch and a regression branch; in the classification branch, based on the predicted state vector, state label determination is performed through probability distribution matching; in the regression branch, a pretightening force prediction function is constructed through a nonlinear weighted residual fitting term and a gradient normalization term of a state evolution tensor residual, combined with the state vector and the predicted state vector, to obtain a predicted pretightening force value of the bolt.

[0019] Preferably, S2 specifically comprises:

[0020] In the construction process of the pretightening force prediction function, a residual term is generated based on the state vector and the predicted state vector, a nonlinear exponential compression term is combined to construct a nonlinear weighted residual fitting term, and the rate of change of the state vector is combined to predict the pretightening force value of the bolt.

[0021] A steel pipe bolt tightness state prediction system based on ultrasonic recognition comprises the following parts:

[0022] A signal sensing module, a signal projection mapping module, an intelligent decoupling module, a state vector construction module, a state vector prediction module, and a state classification prediction evaluation module;

[0023] The signal sensing module transmits and receives signals to the target bolt through a phased array ultrasonic sensor group, obtains multi-channel ultrasonic reflection signals, and sends the multi-channel ultrasonic reflection signals to the signal projection mapping module;

[0024] The signal projection mapping module projects the multi-channel ultrasonic reflection signals to a three-dimensional coordinate space, defines an ultrasonic energy attenuation distribution function in the three-dimensional coordinate space, and based on the ultrasonic energy attenuation distribution function, superimposes and maps all multi-channel ultrasonic reflection signals to a four-dimensional space-time disturbance tensor field to obtain an ultrasonic disturbance energy amplitude; the ultrasonic disturbance energy amplitude is sent to the intelligent decoupling module;

[0025] The intelligent decoupling module obtains a stationary energy benchmark state tensor based on the ultrasonic disturbance energy amplitude, introduces a direction modulation factor, combines a signal period modulation term, quantifies the degree of echo energy deviation from the stable state in the main propagation direction, and obtains a disturbance energy deviation rate; the disturbance energy deviation rate is sent to the state vector construction module;

[0026] The state vector construction module unifies the disturbance energy deviation rates in different directions into a state vector, takes the state vector as the basis for judging whether the current connection state is abnormal, and sends it to the state vector prediction module and the state classification prediction evaluation module;

[0027] The state vector prediction module designs a nonlinear time series recursive model based on the state vector to map the dynamic evolution of the state vector and obtain a predicted state vector; the predicted state vector is sent to the state classification prediction evaluation module;

[0028] The state classification prediction evaluation module introduces a double-branch evaluation mechanism based on the state vector and the predicted state vector, performs state classification and pretightening force regression prediction, and realizes bolt tightness state prediction.

[0029] The technical scheme of the present application has the following beneficial effects:

[0030] 1. By constructing a four-dimensional space-time disturbance tensor field and introducing a stationary energy benchmark state tensor, a disturbance energy deviation rate is generated to quantify the deviation influence caused by the bolt loosening state on the propagation path of ultrasonic energy in a specific direction, thereby enhancing the sensitivity to abnormal propagation patterns at the data feature level, realizing the ability to automatically extract directional disturbance changes from continuous ultrasonic echo fields, and enhancing the robustness of state perception.

[0031] 2. The disturbance energy deviation rate is constructed as a multi-dimensional time state vector, and a nonlinear time series recursive model with time dynamics and state memory capability is constructed through double-layer nonlinear transformation and nonlinear energy items of state disturbance, which can capture the continuous evolution trend of bolt connection state at multiple time points and spatial directions, breaking through the limitation of traditional static identification model that can only make point-in-time judgment, and having the ability to provide early warning for structural state changes.

[0032] 3. A double-branch evaluation mechanism that integrates static residual and dynamic gradient is constructed, the classification branch is matched with the conditional probability of the probability density function of each state category pre-trained, and the state category with the maximum posterior probability is selected as the current bolt state label according to the Bayesian decision criterion, which has good interpretability and stability; the regression branch adopts a pretightening force prediction function constructed based on disturbance prediction residual and state change rate double variables, which couples the physical residual characteristics and prediction trend into a mathematical model, not only improving the prediction accuracy, but also quantifying the bolt connection loosening degree, directly serving the structural safety evaluation and maintenance decision. BRIEF DESCRIPTION OF DRAWINGS

[0033] Figure 1 A steel pipe bolt tightness state prediction system structure diagram based on ultrasonic identification is provided.

[0034] Figure 2 A steel pipe bolt tightness state prediction method flow chart based on ultrasonic identification is provided. DETAILED DESCRIPTION

[0035] In order to further illustrate the technical means and effects taken by the present application to achieve the predetermined inventive purpose, the technical solutions in the embodiments of the present application will be described clearly and completely below in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0036] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0037] The specific scheme of the steel pipe bolt tightness state prediction method and system based on ultrasonic recognition provided by the present application will be specifically described below in combination with the drawings.

[0038] Referring to the drawings Figure 1 It shows a steel pipe bolt tightness state prediction system structure diagram based on ultrasonic recognition provided by an embodiment of the present application, which includes the following parts:

[0039] The signal sensing module, the signal projection mapping module, the intelligent decoupling module, the state vector construction module, the state vector prediction module and the state classification prediction evaluation module;

[0040] The signal sensing module transmits and receives signals to the target bolt through the phased array ultrasonic sensor group, obtains multi-channel ultrasonic reflection signals, and sends the multi-channel ultrasonic reflection signals to the signal projection mapping module;

[0041] The signal projection mapping module projects the multi-channel ultrasonic reflection signals to a three-dimensional coordinate space, defines an ultrasonic energy attenuation distribution function in the three-dimensional coordinate space, and based on the ultrasonic energy attenuation distribution function, superimposes and maps all multi-channel ultrasonic reflection signals to a four-dimensional space-time disturbance tensor field to obtain an ultrasonic disturbance energy amplitude; the ultrasonic disturbance energy amplitude is sent to the intelligent decoupling module;

[0042] The intelligent decoupling module obtains a stationary energy reference state tensor based on the ultrasonic disturbance energy amplitude, introduces a direction modulation factor, combines a signal period modulation term, quantifies the degree of echo energy deviation from the stable state in the main propagation direction, and obtains a disturbance energy deviation rate; the disturbance energy deviation rate is sent to the state vector construction module;

[0043] The state vector construction module uniformly represents the disturbance energy deviation rates in different directions as a state vector, takes the state vector as the basis for judging whether the current connection state is abnormal, and sends it to the state vector prediction module and the state classification prediction evaluation module;

[0044] The state vector prediction module designs a nonlinear time-series recursive model based on the state vector to map the dynamic evolution of the state vector. It obtains the state vector at the next time step, i.e., the predicted state vector, through the nonlinear recursive rule mapping. The predicted state vector is then sent to the state classification prediction and evaluation module.

[0045] The state classification prediction and evaluation module, based on the state vector and the predicted state vector, introduces a dual-branch evaluation mechanism to perform state classification and preload regression prediction, thereby realizing the prediction of the tightness state of steel pipe bolts.

[0046] See attached document Figure 2 The diagram illustrates a flowchart of a method for predicting the tightness of steel pipe bolts based on ultrasonic identification, according to an embodiment of the present invention. The method includes the following steps:

[0047] S1. Acquire multi-channel ultrasonic reflection signals and project them onto a three-dimensional coordinate space. Quantify the energy diffusion of ultrasonic pulses as they propagate in space and construct an ultrasonic energy attenuation distribution function. Based on the ultrasonic energy attenuation distribution function, map all multi-channel ultrasonic reflection signals onto a four-dimensional spatiotemporal perturbation tensor field and perform decoupling processing to obtain the perturbation energy offset rate.

[0048] The target bolt is subjected to signal transmission and reception using a phased array ultrasonic sensor array, resulting in multi-channel ultrasonic reflection signals, denoted as... , indicating the first In the ultrasonic sensing unit, the first For the transmit-receive channel combination at time The collected ultrasonic reflection signals; the phased array ultrasonic sensor group is composed of Composed of several ultrasonic sensing units, deployed by professional technicians near key connection nodes of the steel pipe; furthermore, to give the subsequent analysis physical spatial meaning, the multi-channel ultrasonic reflection signals are projected into a three-dimensional coordinate space to form continuous propagation. Specifically, an ultrasonic energy attenuation distribution function is defined in the three-dimensional coordinate space to simulate the energy diffusion caused by factors such as medium loss, contact interface reflection, and structural geometric transformations when the ultrasonic pulse propagates in space, as defined below:

[0049] ,

[0050] in, It is the ultrasonic energy attenuation distribution function, which is a Gaussian distribution function in three-dimensional space. It is used to describe how the influence of ultrasonic energy on each point decreases with distance when ultrasonic waves propagate in space. At spatial coordinate points Above, by the first In the ultrasonic sensing unit, the first The influence factor of ultrasonic energy formed by the combination of transmitting and receiving channels; represent the actual layout position coordinates of the first ultrasound sensing unit; represent the actual layout position coordinates of the first channel direction, used to control the diffusion radius of the Gaussian distribution function, capable of reflecting the scattering characteristics of directional ultrasonic energy during spatial propagation, determined based on the actual reference scene combined with the beam broadening model, with a reference value range of ;

[0051] Based on the above ultrasonic energy attenuation distribution function, all multi-channel ultrasonic wave reflection signals are superimposed and mapped to a four-dimensional space-time disturbance tensor field , which is represented as:

[0052] ,

[0053] wherein, represents the ultrasonic disturbance energy amplitude at spatial point and time , and serves as the original input for subsequent energy shift feature extraction; represents the number of transmit-receive channel combinations;

[0054] After the four-dimensional space-time disturbance tensor field is constructed, in order to reveal the disturbance energy spatial path shift caused by bolt loosening, it is necessary to decouple the disturbance shift caused by the bolt loosening state to the propagation path from the four-dimensional space-time disturbance tensor field. By introducing the disturbance energy shift rate, the degree of echo energy deviation from the stable state in a certain main transmission direction is quantified, which can reflect the influence of the bolt joint interface state on the ultrasonic propagation characteristics in that direction. The specific implementation formula is as follows:

[0055] ,

[0056] wherein, is the shift rate of the ultrasonic energy field at time relative to the stable state energy field of the structure in the main propagation direction angle , i.e. the disturbance energy shift rate; is the stationary energy reference state tensor, obtained by integrating and averaging all time ultrasonic disturbance energy amplitudes at spatial point ; is the main propagation direction angle, representing the configuration angle of the transmission channel, which is the incident central angle of the phased array ultrasonic sensor group; is the direction modulation factor, used to adjust the projection weight of different main propagation directions in the integral, embodying the direction selectivity, obtained through ultrasonic simulation, measurement fitting, etc., with a reference value range of ; is a complete cycle of ultrasonic wave, and is the inverse of the frequency of the emitted signal; is a signal cycle modulation term, which is used to extract the response characteristics of periodic disturbance; is a very small positive number, which is used to prevent division by zero, and can be ; is the difference between the current four-dimensional space-time disturbance tensor field and the stable energy reference state tensor, i.e. the disturbance term; By adjusting the directionality weight of different main propagation directions, the integral is more concerned about the disturbance distribution along the specified direction, which is used to control the direction selection accuracy; By integrating in three-dimensional space, the overall distribution of disturbance energy in a certain direction in space is obtained; the denominator part represents the total amount of the stable energy reference state tensor, which is used as a normalization factor to make the disturbance energy shift rate comparable and to eliminate the absolute difference of energy shift between different structural scales or measuring points.

[0057] S2. The disturbance energy shift rate is uniformly represented as a state vector, and a nonlinear time series recursive model is constructed to obtain a predicted state vector; based on the predicted state vector, a double-branch evaluation mechanism is introduced to perform state classification and pretightening force regression prediction, thereby realizing the prediction of the bolt tightness state of the steel pipe.

[0058] The disturbance energy shift rates in different directions are uniformly represented as a state vector , i.e.

[0059] ,

[0060] wherein, represents the disturbance energy shift rate at time in the main propagation direction angle ; is the number of selected main propagation directions, which is determined according to specific application requirements, such as , then it can be selected once every 45 degrees; represents transposition; the state vector reflects the multi-directional energy disturbance distribution at a certain time, and is the basis for judging whether the current connection state is abnormal;

[0061] Further, in order to predict the evolution trajectory of the above state vector over time, a nonlinear time series recursive model is constructed to map the dynamic evolution of the state vector; the nonlinear time series recursive model maps the state vector at the next time through a nonlinear recursive rule to obtain a predicted state vector, and the specific implementation formula is as follows:

[0062] ,

[0063] wherein, is the next time a state vector, i.e., a predicted state vector; is a first layer linear transformation weight matrix of the disturbance state mapping, used for mapping the current state vector to an intermediate feature space, extracting the coupling relationship between the disturbance features, obtained by back propagation learning, the first layer linear transformation weight matrix The reference value range of each element in ; is a disturbance offset vector (bias term), which can reflect the basic disturbance offset in different directions, obtained by back propagation learning, the disturbance bias vector The reference value range of each element in ; is a second layer disturbance response synthesis weight matrix, obtained by back propagation learning, the second layer disturbance response synthesis weight matrix The reference value range of each element in ; is a disturbance energy convergence degree adjustment factor, used for adjusting the nonlinear feedback strength of the total disturbance energy , which is analogous to a damping factor, and the reference value range is ; is a disturbance direction state response prediction main term; is a nonlinear energy term of state disturbance, which is obtained by taking the logarithm of the square of the L2 norm of the state vector, and inhibits the increase of large energy disturbance, so as to enhance the robustness; the above back propagation learning method is a well-known technical means for those skilled in the art, and will not be described here;

[0064] Further, in order to finally identify and quantify the bolt state, a double branch evaluation mechanism is introduced to perform state classification and pretightening force regression prediction; the double branch evaluation mechanism includes a classification branch and a regression branch; in the classification branch, the existing probability distribution matching is used to realize state label judgment, and the specific implementation manner is that the predicted state vector is brought into the probability density function of each state category which has been trained, the likelihood value of its occurrence under each state category, i.e., the conditional probability, is calculated, the prior probability value of the corresponding category is combined, the posterior probability is solved according to the Bayes formula, and finally the state category with the maximum posterior probability is selected as the current bolt state label according to the Bayes decision criterion, so as to realize the multi-classification judgment based on probability matching; the methods used in the classification branch are well-known technical means for those skilled in the art, and will not be described here;

[0065] In the regression branch, the following pretightening force prediction function is constructed by a physical residual model:

[0066] ,

[0067] wherein, is the predicted current time bolt pre-tightening force value; , i.e. is the disturbance energy offset rate on the first main propagation direction angle ; is the disturbance energy offset rate on the first main propagation direction angle at the predicted time ; is the main term fitting coefficient, and the reference value range is through finite element physical calibration, and the finite element physical calibration is a technical means familiar to those skilled in the art, which will not be described here; is the predicted state weight adjustment factor, which is used to control the relative influence between the current state and the predicted state, and is determined through cross-validation optimization, and the reference value range is , and the cross-validation optimization method is a technical means familiar to those skilled in the art, which will not be described here; is the exponential compression modulation factor, which is used to control the exponential suppression intensity of the predicted state, and is determined through gain response fitting method, and the reference value range is , and the gain response fitting method is a technical means familiar to those skilled in the art, which will not be described here; is the secondary term fitting coefficient, which is used to control the contribution of the disturbance gradient to the pre-tightening force, and is determined according to the expert experience method, and the reference value range is ; is the vector first-order difference (discrete gradient), which is used to measure the change rate of the state vector in the direction dimension; is the Euclidean norm, which is used to calculate the total amplitude of the change of the state vector; is the residual term, which represents the residual square between the disturbance energy offset rate on the first main propagation direction between the current time and the next time (predicted); is the nonlinear exponential compression term, which is used to adjust the weight of the above residual term when the predicted value is large; is the nonlinear weighted residual fitting term, which can reflect the "position deviation" between the current state of the structure and its disturbance trend, and focuses on static deviation and energy amplitude; is the gradient normalization term of the state evolution tensor residual, which can reflect the change speed in the spatial direction disturbance, and focuses on "dynamic instability".

[0068] In summary, a steel pipe bolt tightness state prediction method and system based on ultrasonic identification are completed.

[0069] The progressive nature of the specification and claims, with different embodiments described on different levels of generality, does not require readiness of each and every embodiment recited on a lower level of generality to qualify as an embodiment of the invention. Nor does it require that a single embodiment described on a higher level of generality be described on a lower level of generality.

[0070] The various embodiments in the specification are described in progressive order, and each embodiment can refer to the other embodiments. Each embodiment is distinguished from the other embodiments by the different points of the embodiment.

[0071] The above embodiments are only used to illustrate the technical solutions of the present application, but not limit the present application; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A method for predicting the tightness of steel pipe bolts based on ultrasonic recognition, characterized in that, Includes the following steps: S1. Acquire multi-channel ultrasonic reflection signals and project them onto a three-dimensional coordinate space. Quantify the energy diffusion of the ultrasonic pulse during propagation in space and construct an ultrasonic energy attenuation distribution function. Based on the ultrasonic energy attenuation distribution function, superimpose all multi-channel ultrasonic reflection signals and map them onto a four-dimensional spatiotemporal perturbation tensor field. The amplitude of the ultrasonic disturbance energy is obtained, specifically expressed as: ， in, Indicates a point in space and time The amplitude of ultrasonic disturbance energy at the location; Indicates the first In the ultrasonic sensing unit, the first For the transmit-receive channel combination at time The collected ultrasonic reflection signals; Indicates the number of transmit-receive channel combinations; Indicates the number of ultrasonic sensing units; It is the ultrasonic energy attenuation distribution function; At spatial coordinate points Above, by the first In the ultrasonic sensing unit, the first The influence factor of ultrasonic energy formed by the combination of transmitting and receiving channels; Based on the amplitude of ultrasonic disturbance energy, a steady energy reference state tensor is obtained. A direction modulation factor is introduced and combined with a signal period modulation term to quantify the degree of deviation of the echo energy from the steady state in the main propagation direction. The disturbance offset caused by the bolt loosening state on the propagation path is decoupled from the four-dimensional spatiotemporal disturbance tensor field to obtain the disturbance energy offset rate. S2. The disturbance energy offset rate is uniformly represented as a state vector. Based on the state vector, a two-level nonlinear transformation is performed, and a nonlinear time-series recursive model is constructed by combining the nonlinear energy term of the state disturbance. The predicted state vector is obtained by mapping through nonlinear recursive rules. Based on the predicted state vector, a dual-branch evaluation mechanism is introduced, which includes a classification branch and a regression branch: In the classification branch, the state label is determined by probability distribution matching based on the predicted state vector; In the regression branch, the preload prediction function is constructed by combining the state vector and the predicted state vector through the nonlinear weighted residual fitting term and the gradient normalization term of the state evolution tensor residual, and the preload regression prediction is performed to realize the prediction of the tightness state of the steel pipe bolt.

2. The method for predicting the tightness of steel pipe bolts based on ultrasonic recognition according to claim 1, characterized in that, S1 specifically includes: The phased array ultrasonic sensor group transmits and receives signals to the target bolt to obtain multi-channel ultrasonic reflection signals; the phased array ultrasonic sensor group is composed of ultrasonic sensing units.

3. The method for predicting the tightness of steel pipe bolts based on ultrasonic recognition according to claim 2, characterized in that, S1 specifically includes: Based on the actual deployment coordinates of the ultrasonic sensing unit and the energy diffusion coefficient, an ultrasonic energy attenuation distribution function is constructed.

4. The method for predicting the tightness of steel pipe bolts based on ultrasonic recognition according to claim 1, characterized in that, S2 specifically includes: In the process of constructing the preload prediction function, a residual term is generated based on the state vector and the predicted state vector. A nonlinear weighted residual fitting term is constructed by combining the nonlinear exponential compression term and the rate of change of the state vector. The preload value of the bolt is then predicted.

5. A steel pipe bolt tightness state prediction system based on ultrasonic recognition, applied to the steel pipe bolt tightness state prediction method based on ultrasonic recognition as described in claim 1, characterized in that, Includes the following parts: The system comprises a signal sensing module, a signal projection mapping module, an intelligent decoupling module, a state vector construction module, a state vector prediction module, and a state classification, prediction, and evaluation module. The signal sensing module transmits and receives signals from the target bolt through a phased array ultrasonic sensor group, obtains multi-channel ultrasonic reflection signals, and sends the multi-channel ultrasonic reflection signals to the signal projection mapping module. The signal projection mapping module projects multi-channel ultrasonic reflection signals onto a three-dimensional coordinate space, defines an ultrasonic energy attenuation distribution function in the three-dimensional coordinate space, and maps all multi-channel ultrasonic reflection signals onto a four-dimensional spatiotemporal perturbation tensor field based on the ultrasonic energy attenuation distribution function to obtain the ultrasonic perturbation energy amplitude; the ultrasonic perturbation energy amplitude is then sent to the intelligent decoupling module. The intelligent decoupling module obtains the steady-state energy reference tensor based on the amplitude of ultrasonic disturbance energy, and introduces a direction modulation factor. Combined with the signal periodic modulation term, it quantifies the degree to which the echo energy in the main propagation direction deviates from the steady state, and obtains the disturbance energy offset rate. The disturbance energy offset rate is then sent to the state vector construction module. The state vector construction module uniformly represents the perturbation energy offset rate in different directions as a state vector, uses the state vector as the basis for judging whether there is an anomaly in the current connection state, and sends it to the state vector prediction module and the state classification prediction and evaluation module. The state vector prediction module designs a nonlinear time-series recursive model based on the state vector to map the dynamic evolution of the state vector and obtain the predicted state vector; the predicted state vector is then sent to the state classification prediction and evaluation module. The state classification prediction and evaluation module, based on the state vector and the predicted state vector, introduces a dual-branch evaluation mechanism to perform state classification and preload regression prediction, thereby realizing the prediction of the tightness state of steel pipe bolts.

Images