Method and device for monitoring deflection of main beam

Abstract

The invention provides a main beam deflection monitoring method and device, the device comprises a main beam, a plurality of monitoring members, a counterweight mechanism, a moving mechanism and a distance measuring member, the main beam is erected on the ground; the plurality of monitoring pieces are arranged at different positions of the main beam and are used for detecting modulus values of different positions of the main beam; the moving mechanism is arranged on the main beam and can linearly move in the length extending direction of the main beam. The counterweight mechanism is suspended on the moving mechanism and is used for simulating deflection changes to the main beam under different loads; the distance measuring part is arranged on the moving mechanism and used for measuring the relative distance between the moving mechanism and the end of the main beam, and efficient and accurate monitoring of deflection changes of the main beam of the crane is achieved by combining the monitoring device and the prediction model. The deflection of the main beam is calculated by introducing a measurement modulus value, a modulus value coefficient and position information, so that the deformation degree of the main beam after being stressed can be more accurately estimated, the method can be suitable for deflection monitoring of cranes of different models, an accurate deflection measurement formula is ensured, the monitoring accuracy and reliability are improved, and the working efficiency is improved. And powerful support is provided for safety evaluation of the crane main beam structure.

Description

Technical Field

[0001] The present invention relates to the technical field of main beam disturbance monitoring, and in particular to a main beam disturbance monitoring method and device. Background Art

[0002] The main beam structure may experience changes in its deflection due to various factors during long-term use. To address the problem of main beam deflection, a variety of repair devices can be used, such as prestressing method, flame correction method, repeated welding method, cutting method, and local cutting and padding method. When testing the main beam deflection, measurements can be taken using methods such as pulling steel wire, hanging hook, or magnet.

[0003] Publication number CN108584703B discloses an online measurement system for the lifting weight and main beam deflection of a double-beam bridge crane and a measurement method thereof. The measurement system hardware includes a hook group equipped with a washer-type force sensor, a laser ranging sensor installed at the mid-span position of the bottom plate under the main beam, a cable drum, a ground iron, a vehicle control cabinet, a wireless transmitter, a wireless receiver, a main controller, a display, a relay and a cable power supply; the measurement system software includes a wireless communication module, a data acquisition and processing module, and a measurement control module; the washer-type force sensor can directly measure the real-time lifting weight, and the laser ranging sensor is used to measure the two distance data from the bottom plate of the main beam to the ground iron when the crane is unloaded and loaded, and the two distance data are wirelessly transmitted to the ground controller through a radio frequency signal, and then the main beam deflection corresponding to the real-time lifting weight is calculated by a related formula.

[0004] Currently, disturbance monitoring is required for real-time online monitoring. However, each different device is affected by its various model parameters, and its disturbance calculation formula will be different. The above existing scheme cannot be applied to different models of cranes for disturbance monitoring. The traditional calculation method ignores the inaccurate deflection estimation caused by the combined influence of multiple factors, and cannot ensure the accurate disturbance calculation formula, thereby reducing its detection accuracy. Summary of the invention

[0005] In view of this, the present invention proposes a main beam disturbance monitoring method and device, which can be applicable to different types of cranes for disturbance monitoring, can ensure accurate disturbance measurement formula, and improve its detection accuracy.

[0006] The technical solution of the present invention is implemented as follows: In the first aspect, the present invention provides a monitoring device for main beam disturbance, comprising a main beam, a plurality of monitoring components, a counterweight mechanism, a moving mechanism and a distance measuring component, wherein:

[0007] The main beam is erected on the ground;

[0008] Several monitoring components are arranged at different positions on the main beam to detect the modulus values ​​at different positions of the main beam;

[0009] The moving mechanism is arranged on the main beam and can move linearly along the length extension direction of the main beam;

[0010] The counterweight mechanism is suspended on the moving mechanism to simulate the disturbance changes on the main beam under different loads;

[0011] The distance measuring component is arranged on the moving mechanism and is used for measuring the relative distance between the moving mechanism and the end of the main beam.

[0012] In a second aspect, the present invention further provides a method for monitoring main beam disturbance, which is implemented by using the above-mentioned main beam disturbance monitoring device, and comprises the following steps:

[0013] S1, install the corresponding number of monitoring components according to the length of the main beam;

[0014] S2, change the weight of the counterweight mechanism, and move it horizontally on the main beam through the moving mechanism, measure the modulus value corresponding to each monitoring point and the relative distance value of the moving mechanism to the end of the main beam according to the distance measuring component and the monitoring component, and establish a data sample set according to the corresponding counterweight weight;

[0015] S3, build a prediction model based on the improved informer network model, input the data sample set into the prediction model for training, and obtain the optimal prediction model;

[0016] S4, inputting the counterweight weight to be input and the relative distance of the moving mechanism to the end of the main beam into the optimal prediction model to obtain the compensation modulus value of the current position;

[0017] S5, calculating and obtaining a modulus value coefficient according to the compensation modulus value at the current position and the corresponding counterweight weight;

[0018] S6, calculating and obtaining the disturbance of the main beam according to the length of the main beam, the measured modulus value at the current position and the modulus value coefficient.

[0019] On the basis of the above technical solution, preferably, in step S1, a corresponding number of monitoring components 2 are installed according to the length of the main beam 1, wherein the number of monitoring components 2 arranged is several, a monitoring component 2 is provided at the center of the main beam 1, and several monitoring components 2 are equally divided and arranged on the main beam 1 according to the length of the main beam 1.

[0020] On the basis of the above technical solution, preferably, the monitoring element 2 is one of a strain sensor, a resistive strain sensor, a fiber Bragg grating strain sensor or an ultra-weak fiber Bragg grating strain sensor.

[0021] On the basis of the above technical solution, preferably, step S2 includes the following sub-steps:

[0022] S21, placing the mobile mechanism at a monitoring position at any end of the main beam, and making the counterweight mechanism 3 unloaded, and obtaining the modulus value of the current position as M0;

[0023] S22, moving the moving mechanism to the monitoring component at the center of the main beam under no-load condition, and obtaining the modulus value M1 of the center of the main beam under no-load condition;

[0024] S23, keeping the position unchanged, adding the weight of the counterweight block to the moving mechanism, collecting the weight of the added counterweight block, and obtaining the modulus value M2 of the center position of the main beam under this weight;

[0025] S24, keeping the weight unchanged, moving the moving mechanism from the center of the main beam to the other end of the main beam away from the initial end, when moving to the other end, adding the weight of the counterweight block to the moving mechanism while keeping the position unchanged, collecting the weight of the added counterweight block, and keeping the weight unchanged, moving the moving mechanism from the other end away from the initial end to the initial end side;

[0026] S25, making the moving mechanism move back and forth, and increasing the weight of the counterweight block in sequence, and obtaining the modulus value M when it runs to each position where the detection member is installed during the process i , the relative distance b between the moving mechanism and the end of the main beam i and the corresponding counterweight weight T i ;

[0027] S26, according to the modulus value M at each position during the movement of the moving mechanism i , weight of counterweight T i The relative distance b between the moving mechanism and the end of the main beam i Perform association and establish a data sample set.

[0028] On the basis of the above technical solution, preferably, step S3 includes the following sub-steps:

[0029] S31, preprocessing the data sample set, and calculating the relative distance b between each subset in the data sample set. i , weight of counterweight T i and modulus value M i Perform data cleaning, fill in missing values, process outliers, and amplify data through data enhancement to obtain a standard data set for model value prediction;

[0030] S32, divide the standard data set into a training set, a validation set and a test set, and the ratio of the training set, the validation set and the test set is 7:2:1;

[0031] S33, constructing a prediction model based on the network architecture of informer, wherein the prediction model includes an input layer, a multi-scale feature extraction layer, a convolutional fusion layer, two encoders, a fully connected layer and a decoder, wherein the output end of the input layer is connected to the input end of the multi-scale feature extraction layer, the output layer of the multi-scale feature extraction layer is residually connected to the input layer, and the output layer of the multi-scale feature extraction layer is connected to the input end of the convolutional fusion layer, the output end of the convolutional fusion layer is respectively connected to the input ends of the two encoders, the output ends of the two encoders are connected to the input end of the fully connected layer, the output end of the fully connected layer is connected to the input end of the decoder, and the decoder outputs the final prediction modulus value;

[0032] S34, using the ADAM optimization algorithm to iteratively train the prediction model, optimize the hyperparameters of the prediction model, and obtain the final prediction model.

[0033] On the basis of the above technical scheme, preferably, step S33 includes inputting the training set into the input layer, the input layer outputs the input data to the multi-scale feature extraction layer for feature extraction, the multi-scale feature extraction layer uses one-dimensional depth-separable convolution kernels of 1×3, 1×5 and 1×7 to extract features respectively, and uses zero padding to make the output feature map size of all convolution operations consistent; the features obtained by the multi-scale feature extraction layer are superimposed, and residual connection is introduced to fuse the superimposed features with the data of the input layer, the convolution fusion layer uses a 1×1 convolution kernel to perform a convolution operation on the fused data, and inputs them into each encoder respectively to capture the long-term dependency of the input data; the data output by the two encoders are input into the fully connected layer for feature splicing, and the spliced ​​data are input into the decoder, and the decoder outputs the final prediction modulus value.

[0034] On the basis of the above technical solution, preferably, the ADAM optimization algorithm is used in step S34 to iteratively train the prediction model, optimize the hyperparameters of the prediction model, and obtain the final prediction model, wherein:

[0035] The improved Bayesian optimization algorithm is used to optimize the hyperparameters of the prediction model. The calculation expression is:

[0036]

[0037] Let γ = P(y <y * ) is used to divide g(x) and h(x) and construct the following function:

[0038] P(x)=∫P(x|y)P(y)=γg(x)+(1-γ)h(x);

[0039] The formula is simplified to get the expression:

[0040]

[0041] The simplified expression is:

[0042]

[0043] In the formula, is the expected loss when the hyperparameter is x; γ is a weight parameter used to weight between g(x) and h(x); y* is a threshold used to divide the value range of y; P(x|y) is the probability of hyperparameter x when the loss is y; g(x) represents the loss function when the loss of f(x) is less than the threshold y*; h(x) represents the loss function when the loss of f(x) is greater than or equal to the threshold y*; P(x) represents the probability distribution of random variable x, and P(y) represents the probability distribution of random variable y;

[0044] Adopt ADAM optimizer with learning rate of 1e-5, use Adam optimizer to optimize the weight of prediction model, input training set into prediction model for iterative training, during training, Adam optimizer will update the weight of model according to gradient information until loss function converges;

[0045] The prediction model was tested using the test set, and the MSE and RMSE evaluation indicators were used to evaluate the prediction model. The model was evaluated and screened based on the calculation results to determine the optimal prediction model.

[0046] On the basis of the above technical solution, preferably, step S5 includes the following steps:

[0047] According to the compensation modulus value obtained by the prediction model and the corresponding counterweight weight and the absolute distance between the moving mechanism and the end of the main beam, a functional relationship is constructed, and the expression is:

[0048] M i,z =f(T i , b i )

[0049] Where M i,z is the compensation modulus value corresponding to the counterweight weight at the position of the i-th detection part, and f() is the function equation;

[0050] According to the compensation modulus value and the corresponding counterweight weight, the modulus value coefficient of the counterweight at the corresponding position is calculated, and the expression is:

[0051]

[0052] In the formula, K i is the relationship coefficient between the i-th compensation modulus value and the corresponding counterweight weight.

[0053] On the basis of the above technical solution, preferably, the disturbance of the main beam is calculated according to the length of the main beam, the measured modulus value at the current position and the modulus value coefficient in step S6, and the expression is:

[0054]

[0055] Among them, y max Indicates the maximum deflection of the counterweight corresponding to the current position; M i Indicates the modulus value of the counterweight corresponding to the current position detected by the detection part, K i is the modulus coefficient of the weight corresponding to the current position, b i is the absolute distance between the current position and the end of the main beam, l represents the length of the main beam, and a represents a constant.

[0056] The main beam disturbance monitoring method and device of the present invention have the following beneficial effects compared with the prior art:

[0057] (1) By combining the monitoring device and the prediction model, efficient and accurate monitoring of the disturbance changes of the crane main beam is achieved. The introduction of the measured modulus value, modulus value coefficient and position information to calculate the disturbance of the main beam can more accurately estimate the deformation degree of the main beam after being subjected to force. It can be applied to different types of cranes for disturbance monitoring, ensuring an accurate disturbance measurement formula, improving the accuracy and reliability of monitoring, and providing strong support for the safety assessment of the crane main beam structure.

[0058] (2) The training set data is received through the set input layer, and the multi-scale feature extraction layer is combined with the depth-separable convolution kernel and zero-padding technology to effectively reduce the amount of calculation and the number of parameters, while enhancing the feature extraction capability and keeping the output feature map size consistent. By introducing residual connection and 1×1 convolution kernel fusion features, the model training efficiency and performance are improved; the two encoders capture long-term dependencies, the fully connected layer further processes the features, and the final decoder generates an accurate prediction modulus value, achieving efficient and accurate structural deflection prediction;

[0059] (3) The improved Bayesian optimization algorithm can find the hyperparameter combination that optimizes the performance of the prediction model, thereby improving the prediction accuracy and generalization ability of the model. In addition, the improved Bayesian optimization algorithm can find a more stable and reliable hyperparameter combination by considering the mutual influence and constraints between the hyperparameters, thereby enhancing the stability of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0061] Figure 1 A three-dimensional diagram of a device for monitoring main beam disturbance of the present invention;

[0062] Figure 2 It is a simplified schematic diagram of a device for monitoring main beam disturbance of the present invention;

[0063] Figure 3 It is a flow chart of the main beam disturbance monitoring method of the present invention;

[0064] Figure 4 A schematic diagram of the steps for building a prediction model for the main beam disturbance monitoring method of the present invention. DETAILED DESCRIPTION

[0065] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0066] like Figure 1-4 As shown, the present invention provides a monitoring device for main beam disturbance, comprising a main beam 1, a plurality of monitoring components 2, a counterweight mechanism 3, a moving mechanism 4 and a distance measuring component 5, wherein the main beam 1 is erected on the ground; the plurality of monitoring components 2 are arranged at different positions on the main beam 1, for detecting the modulus values ​​at different positions of the main beam 1; the moving mechanism 4 is arranged on the main beam 1, and can move linearly along the length extension direction of the main beam 1; the counterweight mechanism 3 is suspended on the moving mechanism 4, for simulating the disturbance changes caused to the main beam 1 under different loads; the distance measuring component 5 is arranged on the moving mechanism 4, for measuring the relative distance of the moving mechanism 4 relative to the end of the main beam 1.

[0067] It should be noted that the main beam 1 is firmly erected on the ground to ensure that it can remain stable under various test conditions. The monitoring components 2 are evenly distributed at different positions of the main beam 1 to detect the modulus values ​​of the corresponding counterweights at different positions of the main beam 1. The monitoring components 2 use strain sensors, and can also be resistive strain sensors, fiber grating strain sensors, ultra-weak fiber grating strain sensors, etc. The counterweight mechanism 3 includes a base plate and four vertical rods. The four vertical rods are all fixed on the base plate and are symmetrically arranged at the center point. The weight of the counterweight mechanism 3 is changed by arranging counterweight blocks on the vertical rods. When adding or reducing weight, the weight blocks on the four vertical rods need to be subtracted at the same time. The counterweight block makes the bottom plate on the counterweight mechanism 3 in a uniformly stressed state, and the counterweight mechanism 3 is hung on the bottom of the mobile mechanism 4 by a cable and is located in the center; the mobile mechanism 4 is a walking trolley, which can accurately move the adjustable counterweight mechanism 3 and the distance measuring component 5 to any position on the main beam 1; the distance measuring component 5 adopts a laser ranging sensor, and ultrasonic ranging or optical ranging and other measurement methods can also be used to measure the absolute or relative distance of the mobile mechanism 4 relative to the end of the main beam 1. Through the setting of the monitoring component 2, the counterweight mechanism 3, the mobile mechanism 4 and the distance measuring component 5, comprehensive and accurate monitoring of the disturbance change of the main beam 1 is achieved.

[0068] In this embodiment, a plurality of monitoring components 2 are evenly arranged at equal intervals along the length direction of the main beam 1 , and a monitoring component 2 is arranged at the center of the main beam 1 .

[0069] It should be noted that the several monitoring components 2 of the main beam disturbance monitoring device are evenly arranged at equal intervals along the length direction of the main beam 1, ensuring that the deformation of the main beam 1 at different positions and loads can be fully and evenly captured, providing strong data support for subsequent disturbance calculations.

[0070] The present invention also provides a main beam disturbance monitoring method, which is implemented by using the main beam disturbance monitoring device mentioned above, and comprises the following steps:

[0071] S1, install a corresponding number of monitoring components 2 according to the length of the main beam 1.

[0072] There are several monitoring components 2 arranged, and the monitoring components 2 are equally divided and arranged on the main beam 1 according to the length of the main beam 1 .

[0073] It should be noted that several monitoring components 2 are evenly distributed on the main beam 1 at equal intervals, which can ensure that each monitoring component can independently capture the deformation information of its position, thereby avoiding monitoring blind spots or data redundancy caused by excessive concentration or sparseness of monitoring points. In this embodiment, the number of monitoring components 2 is 5, and one of the monitoring components 2 is installed in the middle position of the main beam 1. Because the middle position is usually the area where the main beam 1 deforms most significantly when it is under load, in order to more accurately capture the deformation of this area, a monitoring component 2 is set in the middle position; the other four monitoring components 2 are evenly spaced according to the middle position of the main beam 1 and the length of the main beam 1.

[0074] According to the arrangement of the monitoring components 2 in this embodiment, it is ensured that accurate deformation data can be obtained at each key position, thereby improving the efficiency and accuracy of the entire monitoring system.

[0075] S2, change the weight of the counterweight mechanism 3, and move it horizontally on the main beam 1 through the moving mechanism 4, measure the modulus value corresponding to each monitoring point and the relative distance value of the moving mechanism 4 to the end of the main beam 1 according to the distance measuring part 5 and the monitoring part 2, and establish a data sample set according to the corresponding counterweight weight.

[0076] Wherein, step S2 includes the following sub-steps:

[0077] S21, the mobile mechanism 4 is placed statically at the position of the monitoring member 2 at any end of the main beam 1, and the counterweight mechanism 3 is placed in an unloaded state, and the modulus value of the current position is obtained as M0;

[0078] S22, moving the moving mechanism 4 to the monitoring component 2 at the center of the main beam 1 under no-load condition, and obtaining the modulus value M1 at the center of the main beam 1 under no-load condition;

[0079] S23, keep the position unchanged, add the weight of the counterweight block on the moving mechanism 4, collect the weight of the added counterweight block, and obtain the modulus value M2 of the center position of the main beam 1 under this weight;

[0080] S24, keeping the weight unchanged, moving the moving mechanism 4 from the center position of the main beam 1 to the other end of the main beam 1 away from the initial end, when moving to the other end, adding the weight of the counterweight block to the moving mechanism 4 while keeping the position unchanged, collecting the weight of the added counterweight block, and keeping the weight unchanged, moving the moving mechanism 4 from the other end away from the initial end to the initial end side;

[0081] S25, making the moving mechanism 4 move back and forth, and increasing the weight of the counterweight block in sequence, and obtaining the modulus value M when it runs to each position where the detection member 2 is installed during the process i , the relative distance b between the moving mechanism 4 and the end of the main beam 1 i and the corresponding counterweight weight Ti ;

[0082] S26, according to the modulus value M at each position during the movement of the moving mechanism 4 i , weight of counterweight T i The relative distance b between the moving mechanism 4 and the end of the main beam 1 i Perform association and establish a data sample set.

[0083] It should be noted that the mobile mechanism 4 is placed statically at the position of the monitoring part 2 at any end of the main beam 1, so that the counterweight mechanism 3 is in a no-load state, and the modulus value of the current position is M0 at this time; this value represents the initial state of the main beam 1 without any additional load; the mobile mechanism 4 is moved from the initial position to the center position of the main beam 1 under the no-load state, and the modulus value M1 of the center position of the main beam 1 under the no-load state is obtained at the monitoring part 2 at the center position, so as to understand the natural deformation of the main beam 1 under the no-load state; the mobile mechanism 4 is kept stationary at the center position, the weight of the counterweight block is added to the mobile mechanism 4, and the weight of the added counterweight block is collected to obtain the modulus value M1 under this weight. The modulus value M2 at the center position of the main beam 1 reflects the deformation of the main beam 1 under a specific load; keep the current weight unchanged, move the mobile mechanism 4 from the center position to the other end of the main beam 1 away from the initial end, and when it moves to the other end, keep the position unchanged, increase the weight of the counterweight block on the mobile mechanism 4 again, and collect the increased weight; keep the new weight unchanged, move the mobile mechanism 4 from the end away from the initial end to the initial end side, make the mobile mechanism 4 reciprocate on the main beam, and increase the weight of the counterweight block in turn during each movement, and obtain the current modulus value M each time it moves to the position where the monitoring component 2 is installed i , the relative distance b between the moving mechanism 4 and the end of the main beam 1 i And the corresponding weight T i , according to the modulus value M at each position during the movement of the moving mechanism 4 i , weight of counterweight T i The relative distance b between the moving mechanism 4 and the end of the main beam 1 i The associated data are then organized into a data sample set, which contains the deformation data of the main beam under different loads and positions, providing important data support for subsequent predictions.

[0084] S3, builds a prediction model based on the improved informer network model, inputs the data sample set into the prediction model for training, and obtains the optimal prediction model.

[0085] Wherein, step S3 includes the following sub-steps:

[0086] S31, preprocessing the data sample set, and calculating the relative distance b between each subset in the data sample set.i , weight of counterweight T i and modulus value M i Perform data cleaning, fill in missing values, process outliers, and amplify data through data enhancement to obtain a standard data set for model value prediction;

[0087] It should be noted that each subset in the data sample set is preprocessed, including the relative distance b i , weight of counterweight T i and modulus value M i Data cleaning, missing value filling and outlier processing can improve data quality and ensure the effectiveness and accuracy of model training. Data augmentation technology is used to amplify data to increase the diversity and richness of the data set, improve the generalization ability of the model, avoid overfitting, and finally obtain a standard data set for model value prediction.

[0088] S32, divide the standard data set into a training set, a validation set and a test set, and the ratio of the training set, the validation set and the test set is 7:2:1;

[0089] It should be noted that the training set is used to train the model, the validation set is used to adjust the model's hyperparameters, and the test set is used to evaluate the final performance of the model.

[0090] S33, constructing a prediction model based on the network architecture of informer, wherein the prediction model includes an input layer, a multi-scale feature extraction layer, a convolutional fusion layer, two encoders, a fully connected layer and a decoder, wherein the output end of the input layer is connected to the input end of the multi-scale feature extraction layer, the output layer of the multi-scale feature extraction layer is residually connected to the input layer, and the output layer of the multi-scale feature extraction layer is connected to the input end of the convolutional fusion layer, the output end of the convolutional fusion layer is respectively connected to the input ends of the two encoders, the output ends of the two encoders are connected to the input end of the fully connected layer, the output end of the fully connected layer is connected to the input end of the decoder, and the decoder outputs the final prediction modulus value;

[0091] It should be noted that the input layer receives data from the standard data set and then passes it to the multi-scale feature extraction layer. The multi-scale feature extraction layer extracts features of different scales and performs residual connection with the input layer to improve the efficiency and accuracy of feature extraction. The convolution fusion layer convolutionally fuses the output of the multi-scale feature extraction layer to generate a higher-level feature representation to enhance the nonlinear expression ability of the model and improve the expressiveness of the features. The two encoders encode the output of the convolution fusion layer and extract the dependencies in the time series. The fully connected layer further processes the output of the encoder and passes it to the decoder. The decoder generates the final prediction modulus value based on the output of the fully connected layer.

[0092] Specifically, the training set is input into the input layer, and the input layer outputs the input data to the multi-scale feature extraction layer for feature extraction. The multi-scale feature extraction layer uses one-dimensional depth-separable convolution kernels of 1×3, 1×5 and 1×7 to extract features respectively, and uses zero padding to make the output feature map size of all convolution operations consistent; the features obtained by the multi-scale feature extraction layer are superimposed, and residual connection is introduced to fuse the superimposed features with the data of the input layer. The convolution fusion layer uses a 1×1 convolution kernel to perform convolution operations on the fused data, and inputs them into each encoder respectively to capture the long-term dependency of the input data; the data output by the two encoders are input into the fully connected layer for feature splicing, and the spliced ​​data is input into the decoder, and the decoder outputs the final prediction modulus value.

[0093] It should be noted that the input layer serves as the entry point of the prediction model. The input layer receives data from the training set. The multi-scale feature extraction layer uses one-dimensional depth-separable convolution kernels of different sizes to extract the features of the input data. Depth-separable convolution kernels can reduce the amount of computation and the number of parameters of the model while maintaining good feature extraction capabilities. In order to ensure that the output feature map size of all convolution operations is consistent, the multi-scale feature extraction layer uses zero padding technology. Zero padding is to add zero values ​​around the boundaries of the input data to expand the size of the data, thereby ensuring that the output size after the convolution operation is consistent with the expected size. The extracted features are then superimposed together to form a richer feature representation. In order to improve the training efficiency and performance of the model, the multi-scale feature extraction layer introduces residual connections. The residual connection directly adds the superimposed features to the data of the input layer to retain part of the information of the input data and help the model better learn the mapping relationship between input and output. The fused data is convoluted using a 1×1 convolution kernel. The 1×1 convolution kernel is mainly used to adjust the number of channels of the feature map and realize the fusion and interaction between different features. The output of the convolutional fusion layer is then input into two encoders respectively. The two encoders are used to capture the long-term dependencies of the input data and learn the associations and dependencies between different positions in the input data. The data output by the two encoders is input into the fully connected layer for feature concatenation. The fully connected layer further processes these features and prepares to pass them to the decoder. The decoder receives the data output by the fully connected layer and generates the final prediction modulus value prediction result corresponding to the input data.

[0094] S34, using the ADAM optimization algorithm to iteratively train the prediction model, optimize the hyperparameters of the prediction model, and obtain the final prediction model.

[0095] Among them, the ADAM optimization algorithm is used to iteratively train the prediction model, optimize the hyperparameters of the prediction model, and obtain the final prediction model.

[0096] The improved Bayesian optimization algorithm is used to optimize the hyperparameters of the prediction model. The calculation expression is:

[0097]

[0098] Let γ=P(y<y * ) is used to divide g(x) and h(x) and construct the following function:

[0099] P(x)=∫P(x|y)P(y)=γg(x)+(1-γ)h(x);

[0100] The formula is simplified to get the expression:

[0101]

[0102] The simplified expression is:

[0103]

[0104] In the formula, is the expected loss when the hyperparameter is x; γ is a weight parameter used to weight between g(x) and h(x); y* is a threshold used to divide the value range of y; P(x|y) is the probability of hyperparameter x when the loss is y; g(x) represents the loss function when the loss of f(x) is less than the threshold y*; h(x) represents the loss function when the loss of f(x) is greater than or equal to the threshold y*; P(x) represents the probability distribution of random variable x, and P(y) represents the probability distribution of random variable y;

[0105] The learning rate is 1e- 5 The Adam optimizer is used to optimize the weight of the prediction model. The training set is input into the prediction model for iterative training. During the training process, the Adam optimizer will update the weight of the model according to the gradient information until the loss function converges.

[0106] The prediction model was tested using the test set, and the MSE and RMSE evaluation indicators were used to evaluate the prediction model. The model was evaluated and screened based on the calculation results to determine the optimal prediction model.

[0107] It should be noted that, first, two loss functions g(x) and h(x) are defined, where g(x) represents the loss function when the loss of the prediction model is less than a certain threshold y, and h(x) represents the loss function when the loss of the prediction model is greater than or equal to the threshold y. These two loss functions together constitute part of the optimization objective; then, an integral expression is constructed to calculate the expected loss under a given hyperparameter x. This integral expression involves the loss functions g(x) and h(x), as well as their relationship with the threshold y*. At the same time, the probability distribution P(y) is introduced to represent the probability of the loss y, and the probability distribution P(x|y) is introduced to represent the probability of the hyperparameter x given the loss y. Afterwards, the improved Bayesian optimization algorithm is used to find the optimal hyperparameter x so that the expected loss Minimize. In this process, a weight parameter γ is introduced to weight between g(x) and h(x) to balance the optimization objectives under different loss conditions. Finally, the learning rate is 1e -5 The ADAM optimizer is used to optimize the weights of the prediction model, and the training set is input into the prediction model for iterative training. During the training process, the ADAM optimizer will update the weights of the model according to the gradient information until the loss function converges. Then, we use the test set to test the prediction model, and use the MSE and RMSE evaluation indicators to evaluate the prediction model to determine the optimal prediction model.

[0108] In this embodiment, the ADAM optimization algorithm and the improved Bayesian optimization algorithm are used to iteratively train the prediction model and optimize the hyperparameters. The ADAM optimization algorithm designs independent adaptive learning rates for different parameters by calculating the first-order moment estimate and the second-order moment estimate of the gradient, thereby accelerating the convergence speed of the model and reducing the training time. The improved Bayesian optimization algorithm avoids a large number of invalid calculations in traditional grid search or random search methods by intelligently selecting hyperparameter combinations for evaluation, thereby improving the efficiency of hyperparameter optimization. The improved Bayesian optimization algorithm can find the hyperparameter combination that optimizes the performance of the prediction model, thereby improving the prediction accuracy and generalization ability of the model. The improved Bayesian optimization algorithm can find a more stable and reliable hyperparameter combination by considering the mutual influence and constraints between the hyperparameters, thereby enhancing the stability of the model.

[0109] S4, input the counterweight weight to be input and the relative distance of the moving mechanism 4 to the end of the main beam 1 into the optimal prediction model to obtain the compensation modulus value of the current position.

[0110] S5, calculating and obtaining a modulus value coefficient according to the compensation modulus value at the current position and the corresponding counterweight weight.

[0111] Wherein, step S5 includes the following steps:

[0112] According to the compensation modulus value obtained by the prediction model and the corresponding counterweight weight and the absolute distance between the mobile mechanism 4 and the end of the main beam 1, a functional relationship is constructed, and the expression is:

[0113] M i,z =f(Ti ,b i )

[0114] Where M i,z is the compensation modulus value corresponding to the counterweight weight at the position of the i-th detection piece 2, and f() is the function equation;

[0115] According to the compensation modulus value and the corresponding counterweight weight, the modulus value coefficient of the counterweight at the corresponding position is calculated, and the expression is:

[0116]

[0117] In the formula, K i is the relationship coefficient between the i-th compensation modulus value and the corresponding counterweight weight.

[0118] S6, calculating and obtaining the disturbance of the main beam 1 according to the length of the main beam 1, the measured modulus value at the current position and the modulus value coefficient.

[0119] The expression in step S6 is:

[0120]

[0121] Among them, y max Indicates the maximum deflection of the counterweight corresponding to the current position; M i Indicates the modulus value of the counterweight corresponding to the current position detected by the detection part, K i is the modulus coefficient of the weight corresponding to the current position, b i is the absolute distance between the current position and the end of the main beam 1, l represents the length of the main beam 1; a represents a constant.

[0122] It should be noted that, first, a corresponding number of monitoring components 2 are installed according to the length of the main beam 1; these monitoring components 2 are evenly arranged along the length direction of the main beam, and a monitoring component 2 is specially arranged at the center to ensure that the deformation of the main beam 1 is fully captured; then, the weight of the counterweight mechanism 3 is changed, and it is moved horizontally on the crane main beam 1 through the moving mechanism 4; during the movement, the distance measuring component 5 is used to measure the relative distance value of the moving mechanism relative to the end of the main beam 1, and the monitoring component 2 is used to measure the modulus value corresponding to each monitoring point. These measured values ​​are constructed together with the corresponding counterweight weight to form a data sample set, which provides rich data support for subsequent model training; then, a prediction model is constructed based on the improved Informer network model. The model has a strong time series prediction ability and is suitable for dealing with dynamic problems such as crane main beam disturbance that change over time. The data sample set is input into the prediction model for training. Through continuous iteration and optimization, the optimal prediction model is obtained. After the model training is completed, these parameters can be input into the optimal prediction model according to the counterweight weight to be input and the relative distance of the moving mechanism relative to the end of the crane main beam. The model will output the compensation modulus value of the current position, which reflects the additional deformation of the main beam under a given counterweight and position. Then, the modulus coefficient is calculated based on the compensation modulus value of the current position and the corresponding counterweight weight. This coefficient reflects the degree of deformation of the crane main beam under unit counterweight. Finally, the disturbance of the crane main beam is calculated based on the length of the crane main beam, the measured modulus value at the current position and the modulus coefficient. The disturbance reflects the degree of bending deformation of the main beam and can evaluate the overall stability of the crane. By combining the monitoring device and the prediction model, efficient and accurate monitoring of the disturbance changes of the crane main beam is achieved. This not only improves the accuracy and reliability of monitoring, but also provides strong technical support for the safety assessment of cranes.

[0123] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for monitoring main beam disturbance, characterized in that: The following steps are involved: S1, installing a corresponding number of monitoring components (2) according to the length of the main beam (1); S2, the weight of the counterweight mechanism (3) is changed, and the moving mechanism (4) is moved horizontally on the main beam (1), and the modulus value corresponding to each monitoring point and the relative distance value of the moving mechanism (4) relative to the end of the main beam (1) are respectively measured according to the distance measuring component (5) and the monitoring component (2), and a data sample set is established according to the corresponding counterweight weight; S3, build a prediction model based on the improved informer network model, input the data sample set into the prediction model for training, and obtain the optimal prediction model; S4, inputting the weight of the counterweight to be input and the relative distance of the moving mechanism (4) relative to the end of the main beam (1) into the optimal prediction model to obtain the compensation modulus value of the current position; S5, calculating and obtaining a modulus value coefficient according to the compensation modulus value at the current position and the corresponding counterweight weight; S6, calculating and obtaining the disturbance of the main beam (1) according to the length of the main beam (1), the measured modulus value at the current position and the modulus value coefficient.

2. The main beam disturbance monitoring method according to claim 1, characterized in that: In step S1, a corresponding number of monitoring components (2) are installed according to the length of the main beam (1), wherein the number of the monitoring components (2) arranged is a plurality, a monitoring component (2) is arranged at the center of the main beam (1), and the plurality of monitoring components (2) are equally divided and arranged on the main beam (1) according to the length of the main beam (1).

3. The main beam disturbance monitoring method according to claim 2, characterized in that: The monitoring element (2) is one of a strain sensor, a resistive strain sensor, a fiber Bragg grating strain sensor or an ultra-weak fiber Bragg grating strain sensor.

4. The main beam disturbance monitoring method according to claim 3, characterized in that: Step S2 includes the following sub-steps: S21, placing the moving mechanism (4) at a position of the monitoring member (2) at any end of the main beam (1), and placing the counterweight mechanism (3) in an unloaded state, and obtaining a modulus value of the current position as M0; S22, moving the moving mechanism (4) to the monitoring component (2) at the center of the main beam (1) under no-load conditions, and obtaining a modulus value M1 at the center of the main beam (1) under no-load conditions; S23, keeping the position unchanged, adding the weight of the counterweight block to the moving mechanism (4), collecting the weight of the added counterweight block, and obtaining the modulus value M2 of the center position of the main beam (1) under this weight; S24, keeping the weight unchanged, causing the moving mechanism (4) to move from the center position of the main beam (1) to the end of the main beam (1) on the other side away from the initial end, and when moving to the end of the other side, while keeping the position unchanged, adding the weight of the counterweight block to the moving mechanism (4), collecting the weight of the added counterweight block, and keeping the weight unchanged, causing the moving mechanism (4) to move from the end of the other side away from the initial end to the initial end side; S25, making the moving mechanism (4) move back and forth, and increasing the weight of the counterweight block in sequence, and obtaining the modulus value M when it moves to each position where the detection member (2) is installed during the process i , the relative distance b between the moving mechanism (4) and the end of the main beam (1) i and the corresponding counterweight weight T i ; S26, according to the modulus value M at each position during the movement of the moving mechanism (4) i , weight of counterweight T i The relative distance b between the moving mechanism (4) and the end of the main beam (1) i Perform association and establish a data sample set.

5. The main beam disturbance monitoring method according to claim 4, characterized in that: Step S3 includes the following sub-steps: S31, preprocessing the data sample set, and calculating the relative distance b between each subset in the data sample set. i , weight of counterweight T i and modulus value M i Perform data cleaning, fill in missing values, process outliers, and amplify data through data enhancement to obtain a standard data set for model value prediction; S32, divide the standard data set into a training set, a validation set and a test set, and the ratio of the training set, the validation set and the test set is 7:2:1; S33, constructing a prediction model based on the network architecture of informer, wherein the prediction model includes an input layer, a multi-scale feature extraction layer, a convolutional fusion layer, two encoders, a fully connected layer and a decoder, wherein the output end of the input layer is connected to the input end of the multi-scale feature extraction layer, the output layer of the multi-scale feature extraction layer is residually connected to the input layer, and the output layer of the multi-scale feature extraction layer is connected to the input end of the convolutional fusion layer, the output end of the convolutional fusion layer is respectively connected to the input ends of the two encoders, the output ends of the two encoders are connected to the input end of the fully connected layer, the output end of the fully connected layer is connected to the input end of the decoder, and the decoder outputs the final prediction modulus value; S34, using the ADAM optimization algorithm to iteratively train the prediction model, optimize the hyperparameters of the prediction model, and obtain the final prediction model.

6. The main beam disturbance monitoring method according to claim 5, characterized in that: Step S33 includes inputting the training set into the input layer, the input layer outputs the input data to the multi-scale feature extraction layer for feature extraction, the multi-scale feature extraction layer uses one-dimensional depth-separable convolution kernels of 1×3, 1×5 and 1×7 to extract features respectively, and uses zero padding to make the output feature map size of all convolution operations consistent; superimposing the features obtained by the multi-scale feature extraction layer, and introducing residual connection to fuse the superimposed features with the data of the input layer, the convolution fusion layer uses a 1×1 convolution kernel to perform a convolution operation on the fused data, and inputs them into each encoder respectively to capture the long-term dependency of the input data; the data output by the two encoders are input into the fully connected layer for feature splicing, and the spliced ​​data are input into the decoder, and the decoder outputs the final prediction modulus value.

7. The main beam disturbance monitoring method according to claim 6, characterized in that: In step S34, the prediction model is iteratively trained using the ADAM optimization algorithm to optimize the hyperparameters of the prediction model to obtain the final prediction model, wherein: The improved Bayesian optimization algorithm is used to optimize the hyperparameters of the prediction model. The calculation expression is: Let Y = P(y <y * ) is used to divide g(x) and h(x) and construct the following function: P(x)=∫P(x|y)P(y)=γg(x)+(1-γ)h(x); The formula is simplified to get the expression: The simplified expression is: In the formula, is the expected loss when the hyperparameter is x; γ is a weight parameter used to weight between g(x) and h(x); y* is a threshold used to divide the value range of y; P(x|y) is the probability of hyperparameter x when the loss is y; g(x) represents the loss function when the loss of f(x) is less than the threshold y*; h(x) represents the loss function when the loss of f(x) is greater than or equal to the threshold y*; P(x) represents the probability distribution of random variable x, and P(y) represents the probability distribution of random variable y; Adopt ADAM optimizer with learning rate of 1e-5, use Adam optimizer to optimize the weight of prediction model, input training set into prediction model for iterative training, during training, Adam optimizer will update the weight of model according to gradient information until loss function converges; The prediction model was tested using the test set, and the MSE and RMSE evaluation indicators were used to evaluate the prediction model. The model was evaluated and screened based on the calculation results to determine the optimal prediction model.

8. The main beam disturbance monitoring method according to claim 7, characterized in that: Step S5 includes the following steps: According to the compensation modulus value obtained by the prediction model and the corresponding counterweight weight and the absolute distance between the end of the moving mechanism (4) and the main beam (1), a functional relationship is constructed, and the expression is: M i，z =f(T i ,b i ) Where M i，z is the compensation modulus value corresponding to the counterweight weight at the position of the i-th detection part (2), and f() is the function equation; According to the compensation modulus value and the corresponding counterweight weight, the modulus value coefficient of the counterweight at the corresponding position is calculated, and the expression is: In the formula, K i is the relationship coefficient between the i-th compensation modulus value and the corresponding counterweight weight.

9. The main beam disturbance monitoring method according to claim 8, characterized in that: In step S6, the disturbance of the main beam (1) is calculated based on the length of the main beam (1), the measured modulus value at the current position and the modulus value coefficient, and the expression is: Among them, y max Indicates the maximum deflection of the counterweight corresponding to the current position; M i Indicates the modulus value of the counterweight corresponding to the current position detected by the detection part, K i is the modulus coefficient of the counterweight corresponding to the current position, b i is the absolute distance between the current position and the end of the main beam (1), l represents the length of the main beam (1); and a represents a constant.

10. A main beam disturbance monitoring device, implemented by the main beam disturbance monitoring method according to any one of claims 1 to 9, characterized in that: The invention comprises a main beam (1), a plurality of monitoring components (2), a counterweight mechanism (3), a moving mechanism (4) and a distance measuring component (5), wherein the main beam (1) is erected on the ground; the plurality of monitoring components (2) are arranged at different positions on the main beam (1) and are used to detect the modulus values ​​at different positions of the main beam (1); the moving mechanism (4) is arranged on the main beam (1) and can move linearly along the length extension direction of the main beam (1); the counterweight mechanism (3) is suspended on the moving mechanism (4) and is used to simulate the disturbance change caused to the main beam (1) under different loads; and the distance measuring component (5) is arranged on the moving mechanism (4) and is used to measure the relative distance between the moving mechanism (4) and the end of the main beam (1).

Images