Cantilever crane anti-swing control method and device based on MLP neural network inversion

Abstract

The invention provides a cantilever crane anti-swing control method and device based on MLP neural network inversion, and relates to the technical field of electrical digital data processing.The method includes the steps that historical behavior data, displacement deviation value data and swing state data of an operator are collected, and a multi-layer sensor neural network model is established; an accurate control signal is generated in combination with the predicted value and the current state data, and the control performance is continuously optimized through an adjustment strategy and fine adjustment calibration; according to the system, the adaptive capacity to a complex dynamic environment is improved, and the control precision and the operation efficiency are remarkably improved.

Description

Technical Field

[0001] The invention relates to the technical field of electrical digital data, and in particular to a cantilever crane anti-sway control method and device based on MLP neural network inversion. Background Art

[0002] Cantilever cranes are widely used in ports, construction, warehousing and manufacturing industries due to their simple structure, flexible operation and low cost. Traditional cantilever crane operations mainly rely on manual experience, and the precise positioning of the load is achieved by manually adjusting the speed and direction. In the prior art, the publication number is CN118313241A, and the name is a cantilever crane anti-sway control method, which includes: establishing a dynamic model of a cantilever crane system; constructing a dynamic error of the cantilever crane system based on the dynamic model; estimating the uncertainty of the cantilever crane system using an RBF neural network; and designing a controller based on the dynamic error and the estimated uncertainty, so as to control the cantilever crane through the controller. Based on the nonlinear properties of the RBF neural network, it is used to approximate the unmodeled dynamics of the cantilever crane, improve the modeling accuracy of the cantilever crane, and reduce the influence of model parameter uncertainty on the anti-sway control performance.

[0003] The existing crane anti-sway control technology still has many shortcomings. The traditional anti-sway control is to intervene and adjust the anti-sway control signal under the manual adjustment of speed and direction buttons. However, this method often shows shortcomings such as insufficient control accuracy and cumbersome operation in the face of complex operating environments. In addition, the adaptability and accuracy of the single PID control strategy are also limited in nonlinear and large inertia dynamic environments. The above information disclosed in the above Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0004] The object of the present invention is to provide a cantilever crane anti-sway control method and device based on MLP neural network inversion to solve the problems raised in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions: The anti-sway control method of cantilever crane based on MLP neural network inversion includes the following specific steps: Step S1: Determine that the operation buttons for the anti-sway control of the cantilever crane are the speed control button and the direction control button in sequence; In each operation, the operator's historical behavior data for different combinations of speed control keys and direction control keys, as well as the historical displacement deviation data between the object to be hoisted and the target position and the historical swing state data of the object to be hoisted on the crane are collected; Based on the anti-sway mechanism of the crane anti-sway control system, a basic control signal corresponding to the swing state data is obtained; and the basic displacement change between the object to be hoisted and the target position is recorded by these basic control signals. The basic control signal is used to adjust the basic speed and basic direction of the object to be hoisted on the crane; Step S2: construct and train a multilayer perceptron neural network model including an input layer, a hidden layer and an output layer, wherein the input layer is used to receive preprocessed historical behavior data, historical displacement deviation data and historical swing state data under different speed control key and direction control key combinations; the hidden layer is used to perform feature extraction and pattern recognition; The output layer generates the corresponding basic control signal based on the features extracted by the hidden layer to achieve the response to the anti-sway control of the crane; Step S3: Under the combined control of the current speed control key and the direction control key, the closest displacement deviation data is matched from the historical displacement deviation data as the first prediction value; at the same time, the closest swing state data is matched from the historical swing state data as the second prediction value; Step S4: inputting the behavior data received in real time, and the matched first prediction value and second prediction value into the multi-layer perceptron neural network model constructed in step S2, integrating these data types in the hidden layer, performing feature extraction and pattern recognition, so that the output layer generates a corresponding basic control signal, and obtains the real-time basic displacement change between the object to be hoisted and the target position under the control of the basic control signal; Step S5: acquiring the basic control signal in step S4, comparing and analyzing the first predicted value with the real-time basic displacement change, generating a control signal adjustment index for providing an adjustment strategy for the basic control signal, so as to obtain an adjusted basic control signal; Step S6: Use the adjusted basic control signal as the anti-sway control signal of the crane, calculate the displacement deviation between the object to be hoisted and the target position after applying the anti-sway control signal, and set the judgment threshold of the deviation; if the adjusted displacement deviation is still greater than the judgment threshold, fine-tune and calibrate the control signal adjustment index, and use the fine-tune and calibrated control signal adjustment index for the next crane anti-sway control cycle.

[0006] A cantilever crane anti-sway control device based on MLP neural network inversion, the device is used to execute the cantilever crane anti-sway control method based on MLP neural network inversion, comprising: Data acquisition module: determine that the operation buttons for the cantilever crane anti-sway control are the speed control button and the direction control button in sequence; In each operation, the operator's historical behavior data for different combinations of speed control keys and direction control keys, as well as the historical displacement deviation data between the object to be hoisted and the target position and the historical swing state data of the object to be hoisted on the crane are collected; Based on the anti-sway mechanism of the crane anti-sway control system, a basic control signal corresponding to the swing state data is obtained; and the basic displacement change between the object to be hoisted and the target position is recorded by these basic control signals. The basic control signal is used to adjust the basic speed and basic direction of the object to be hoisted on the crane; Model building module: used to build and train a multi-layer perceptron neural network model including an input layer, a hidden layer and an output layer, wherein the input layer is used to receive pre-processed historical behavior data, historical displacement deviation data and historical swing state data under different speed control key and direction control key combinations; the hidden layer is used to perform feature extraction and pattern recognition; The output layer generates the corresponding basic control signal based on the features extracted by the hidden layer to achieve the response to the anti-sway control of the crane; Prediction module: used to match the closest displacement deviation data from the historical displacement deviation data as the first prediction value under the combined control of the current speed control key and the direction control key; at the same time, match the closest swing state data from the historical swing state data as the second prediction value; Basic control signal generation module: used to input the behavior data received in real time, and the matched first prediction value and second prediction value into the constructed multi-layer perceptron neural network model, integrate these data types in the hidden layer, perform feature extraction and pattern recognition, so that the output layer generates the corresponding basic control signal, and obtains the real-time basic displacement change between the object to be hoisted and the target position under the control of the basic control signal; Index generation module: used to obtain the basic control signal, compare and analyze the first predicted value with the real-time basic displacement change, and generate a control signal adjustment index for providing an adjustment strategy for the basic control signal to obtain an adjusted basic control signal; Fine-tuning calibration module: used to use the adjusted basic control signal as the anti-sway control signal of the crane, calculate the displacement deviation between the object to be hoisted and the target position after applying the anti-sway control signal, and set the judgment threshold of the deviation; if the adjusted displacement deviation is still greater than the judgment threshold, the control signal adjustment index is fine-tuned and calibrated, and the control signal adjustment index after fine-tuning and calibrating is used for the next crane anti-sway control cycle.

[0007] Compared with the prior art, the beneficial effects of the present invention are as follows: by collecting the operator's historical behavior data, displacement deviation data and swing state data, and establishing a multi-layer perceptron neural network model, the model receives and pre-processes historical data from different speed and direction control combinations through the input layer, and the hidden layer focuses on feature extraction and pattern recognition, and finally generates an optimized anti-sway control signal through the output layer; During the strategy implementation process, the trained neural network combines the predicted value with the current state data to generate accurate control signals, and continuously optimizes the control performance by adjusting the strategy and fine-tuning the calibration. The system not only improves its adaptability to complex dynamic environments, but also significantly improves control accuracy and operational efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 It is a schematic diagram of the overall method flow of the present invention; Figure 2 4 is a system module block diagram of the device of the present invention. DETAILED DESCRIPTION

[0009] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0010] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0011] Embodiment 1: See also Figure 1 , the present invention provides a technical solution: The anti-sway control method of cantilever crane based on MLP neural network inversion includes the following specific steps: Step S1: Determine that the operation buttons for the anti-sway control of the cantilever crane are the speed control button and the direction control button in sequence; In each operation, the operator's historical behavior data for different combinations of speed control keys and direction control keys, as well as the historical displacement deviation data between the object to be hoisted and the target position and the historical swing state data of the object to be hoisted on the crane are collected; Based on the anti-sway mechanism of the crane anti-sway control system, a basic control signal corresponding to the swing state data is obtained; and the basic displacement change between the object to be hoisted and the target position is recorded by these basic control signals. The basic control signal is used to adjust the basic speed and basic direction of the object to be hoisted on the crane; Step S2: construct and train a multilayer perceptron neural network model including an input layer, a hidden layer and an output layer, wherein the input layer is used to receive preprocessed historical behavior data, historical displacement deviation data and historical swing state data under different speed control key and direction control key combinations; the hidden layer is used to perform feature extraction and pattern recognition; The output layer generates the corresponding basic control signal based on the features extracted by the hidden layer to achieve the response to the anti-sway control of the crane; Step S3: Under the combined control of the current speed control key and the direction control key, the closest displacement deviation data is matched from the historical displacement deviation data as the first prediction value; at the same time, the closest swing state data is matched from the historical swing state data as the second prediction value; Step S4: inputting the behavior data received in real time, and the matched first prediction value and second prediction value into the multi-layer perceptron neural network model constructed in step S2, integrating these data types in the hidden layer, performing feature extraction and pattern recognition, so that the output layer generates a corresponding basic control signal, and obtains the real-time basic displacement change between the object to be hoisted and the target position under the control of the basic control signal; Step S5: acquiring the basic control signal in step S4, comparing and analyzing the first predicted value with the real-time basic displacement change, generating a control signal adjustment index for providing an adjustment strategy for the basic control signal, so as to obtain an adjusted basic control signal; Step S6: Use the adjusted basic control signal as the anti-sway control signal of the crane, calculate the displacement deviation between the object to be hoisted and the target position after applying the anti-sway control signal, and set the judgment threshold of the deviation; if the adjusted displacement deviation is still greater than the judgment threshold, fine-tune and calibrate the control signal adjustment index, and use the fine-tune and calibrated control signal adjustment index for the next crane anti-sway control cycle.

[0012] Further explanation: behavior data, displacement deviation and swing state data specifically include: Behavioral data included hand movement speed, hand movement frequency, and body posture indicators; The speed partitions of the speed control key are divided into {1, 2, ..., a, ..., A} in an ascending sequence, where a represents the index of the a-th speed partition, and A represents the total number of speed partitions. The larger the value of a is, the higher the speed of the corresponding speed partition is. set up represents the average speed of the ath speed partition, Respectively represent the average speeds corresponding to the minimum speed partition 1 and the maximum speed partition A of the speed control key; Divide the direction control buttons into a direction sequence set {1, 2, ..., b, ..., 6} according to the control direction, where b represents the index of the b-th control direction, and 1, 2, 3, 4, 5, and 6 represent forward, backward, left, right, up, and down, respectively; The number of combinations between {1,2,…,a,…,A} and {1,2,…,b,…,6} is A×6; The displacement deviation includes: the vector displacement between the object to be hoisted and the target position under the ath speed partition and the bth type of control direction, and the vector displacement is recorded as ; The swing state data includes: the swing state data of the object to be hoisted on the crane under the ath speed partition and the bth type control direction of the operator, the swing state data includes the swing amplitude and the swing frequency; and the swing state data under the ath speed partition and the bth type control direction is recorded as ;in, They represent the swing amplitude and swing frequency in the ath speed partition and the bth type of control direction respectively; Behavioral data is collected via HD cameras or wearable devices and used to analyze operator operating patterns and their impact on crane stability: Speed ​​control key: a control device used to adjust the crane's moving or lifting speed; changes in speed will affect the stability of the load to be hoisted; Direction control button: a button that controls the direction of load movement (such as front and back, left and right, up and down); rapid or unstable direction changes will cause the load to swing; Hand movement speed: measures the operator's hand movement control speed on each operating button; Hand movement frequency: record the frequency of the operator's hand movements on each operating button; Body posture indicators: The operator's overall body posture, including leaning angle and sitting position, is captured through a camera.

[0013] To further explain, the construction of the multi-layer perceptron neural network model specifically includes: Multilayer perceptron neural network model, hereinafter referred to as MLP neural network model; Input data preprocessing: remove outliers, standardize data, and generate model input format; Input layer settings: configure nodes according to the number of features; Hidden layer settings: Set two hidden layers and use the ReLU activation function; Output layer settings: two nodes, output load speed and basic direction adjustment signals; Model settings and optimization settings: use MSE loss and Adam optimization, and set the learning rate to 0.001; Description of input data preprocessing: Before the construction of the multi-layer perceptron (MLP) neural network model, the collected historical behavior data, historical displacement deviation data and historical swing state data are obtained, and these data are transmitted to the central control system in real time through the wireless network for data cleaning and preprocessing, including removing outliers and noise interference, and standardizing the data to meet the format required by the multi-layer perceptron (MLP) neural network input; and the preprocessed data is formed into a data set; Description of the input layer: according to the number of features of the input data, the number of nodes of the corresponding input layer is set; the input features of this embodiment are hand movement speed, hand movement frequency, displacement deviation, body posture inclination angle and sitting posture, speed control key and direction control key, and there are also A×6 combinations between {1,2,…,a,…,A} and {1,2,…,b,…,6}. According to these input features, the input layer of this embodiment has A×6+7 nodes; Description of hidden layers: Use at least two hidden layers for feature extraction and pattern recognition; The first hidden layer is used for feature extraction to identify the operation mode of hand movement speed and hand movement frequency; and uses the ReLU activation function; The second hidden layer is used for pattern recognition, identifying operational stability patterns associated with body posture; it is used to reduce the number of nodes to prevent overfitting and continues to use the ReLU activation function; Description of the output layer: The types of control signals are speed adjustment and direction adjustment of the load, and the number of nodes in the output layer is set to 2; Instructions for model compilation: Select mean square error (MSE) as the loss function; select Adam optimizer as the optimizer; set the learning rate to 0.001; The data set is divided into a training set and a validation set, 80% for training and 20% for validation; The batch size is 32 and the number of training iterations is 100 to ensure that the model is fully learned; Use callback functions represented by EarlyStopping to prevent model overfitting; Evaluate the model performance on the validation set and judge the model fitting effect based on the loss function value; Deploy the trained multi-layer perceptron neural network model into the control system of the jib crane; Real-time input of operator behavior data, displacement deviation data, and swing state data prediction values ​​allow the multi-layer perceptron neural network model to generate optimized control signals for anti-sway control in crane operations; These quantifiable parameters are fed into the MLP neural network to: Feature extraction: The hidden layer extracts features of hand movement speed and hand movement frequency to identify the operation mode; Pattern recognition: Identify behavioral patterns related to operational stability by analyzing body posture; Optimization control strategy: Based on the input behavior data, displacement deviation data and swing state data, the output layer generates optimized crane control signals to enhance the effect of anti-sway control; These analysis and processing steps ensure that the operator’s behavior data is effectively integrated into the control strategy, optimizing crane stability and operational safety.

[0014] Further explanation: setting the basic displacement change data set specifically includes: Under the combined control of speed partition a and control direction b, the original swing amplitude, swing frequency and displacement deviation data generated in the combined control operation are collected; Use the Min-Max normalization formula to process each data point: All data are normalized to the range of [0,1] to eliminate the dimension effect and improve the stability and accuracy of the model; For the combined control of speed partition a and control direction b, extract the corresponding normalized data: swing amplitude , Swing frequency And displacement deviation ; Set up a pre-simulated anti-sway mechanism, including: under the same group of speed control keys and direction control keys, determine the minimum speed and maximum speed of the speed control key corresponding to the speed partition a, and use this to simulate the swing state data of the object to be hoisted Corresponding numerical fluctuation range, any numerical combination is performed according to the numerical fluctuation range of the swing amplitude and the swing frequency to obtain different swing states of the object to be hoisted in the ath speed partition and the bth type of control direction, and these different swing states are represented by a sequence set {1,2,…,i,…,n}, where i represents the index of the i-th swing state data and n represents the total number of swing state data; the different swing state data in the ath speed partition and the bth type of control direction are represented as ; Where i represents the index of the i-th swing state data, and n represents the total number of swing state data; They respectively represent the swing amplitude and the swing frequency corresponding to the i-th swing state data under the speed partition a and the control direction b; the value of n is determined according to the total number of combinations of the swing amplitude and the swing frequency; Under speed partition a and control direction b, collect the basic control signal corresponding to the i-th swing state data to determine the basic displacement change between the object to be hoisted and the target position , and record in each simulation process, the basic control signals including the basic speed adjustment signal and the basic direction adjustment signal; in It represents the basic displacement change corresponding to the i-th swing state data under speed partition a and control direction b; They respectively represent the basic speed adjustment signal and the basic direction adjustment signal corresponding to the i-th swing state data under speed partition a and control direction b.

[0015] Further, obtaining the first prediction value and the second prediction value specifically includes: setting the speed partition and control direction corresponding to the current speed control key and direction control key to be a1 and b1 respectively, and a1∈{1,2,…,A}, b1∈{1,2,…,6}; Under the combined control of the current speed partition a1 and the control direction b1, access the historical displacement deviation data and the historical swing state data, use the Euclidean distance to calculate the similarity between the current data and the historical data; select the data with the highest similarity as the displacement deviation closest to the current combination and swing status data ; and the index of the current swing state data is recorded as i1, and i1∈{1,2,…,i,…,n}; where is the i1th swing state data in {1,2,…,n}; First prediction value: the displacement deviation to be matched As the first prediction value, it is then divided into M equal distance segments; Second prediction value: the swing state data that will be matched As the second prediction value; at the same time, the swing amplitude and frequency characteristics are extracted to input into the MLP neural network; MLP model input: input the feature data of the current speed partition a1, control direction b1 and the second predicted value into the pre-trained MLP neural network model; the MLP neural network outputs the basic control signal and obtains the real-time basic displacement change between the object to be hoisted and the target position ; in, It indicates the real-time basic displacement change corresponding to the i1th swing state data between the object to be hoisted and the target position in the current speed partition a1 and control direction b1.

[0016] Further explanation, the control signal adjustment index is defined as , the calculation formula is as follows: ; in is the displacement deviation in the first prediction value; is the real-time base displacement change; It is the adjustment sensitivity coefficient, which reflects the sensitivity of the system to changes in the external environment, and its value range is (0, 1); It is a trade-off factor used to balance the influence of the numerator and denominator to avoid exponential imbalance under extreme conditions; is the system response time constant, reflecting the system's response speed to input changes; is the change in the basic speed adjustment, indicating the difference between the current speed and the speed of the previous cycle; It is the angle change of the basic direction adjustment, which indicates the difference between the current direction and the direction of the previous cycle; Adjust sensitivity factor The calculation formula is as follows: ; in is the maximum prediction error change, for the maximum environmental change; System response time constant Determined through the system’s step response experiment, specifically the time required for the control system to reach 63.2% response; Base speed adjustment change The calculation formula is as follows: ; They are the current control speed and the control speed of the previous cycle respectively; The amount of change in the base direction adjustment The calculation formula is as follows: ; They are the current direction and the direction of the previous cycle respectively; set up The valid value range of is (0,1), through , The settings are used to ensure this range; when When it is closer to 0, it means that the difference between the first predicted value and the real-time basic displacement change is smaller, and the adjustment amount of the basic control signal is smaller; when The closer it is to 1, the greater the difference between the first predicted value and the real-time basic displacement change, and the greater the adjustment amount of the basic control signal; The adjustment strategy of the basic control signal is as follows: definition The larger the interval value, the more the signal is adjusted for the basic speed. and fundamental direction adjustment signals The larger the adjustment range, the greater the value of p1<p2<p3; and further define p1<p2<p3; in this embodiment, p1, p2, p3 are respectively set to 3, 13, 25, and the values ​​of p1, p2, p3 are adjusted according to the experimental data obtained by the expert group; when The value is (0,0.25), and When the base speed is adjusted and fundamental direction adjustment signals On this basis, the p1% value is increased; when The value is (0,0.25), and When the base speed is adjusted and fundamental direction adjustment signals On this basis, the p1% value is reduced; when The value is (0,0.25), and When the speed adjustment signal and fundamental direction adjustment signals No adjustments are required; when The value is in [0.25,0.53], and When the base speed is adjusted and fundamental direction adjustment signals On this basis, the p2% value is increased; when The value is in [0.25,0.53], and When the base speed is adjusted and fundamental direction adjustment signals On this basis, the p2% value is reduced; when The value is (0.53,1), and When the base speed is adjusted and fundamental direction adjustment signals On this basis, the p3% value is increased; when The value is (0.53,1), and When the base speed is adjusted and fundamental direction adjustment signals On this basis, the p3% value is reduced.

[0017] The numerical adjustments of the above-mentioned value ranges are 100% adjusted according to experimental data or the expert group system and will not be elaborated on here.

[0018] Further explanation: the displacement deviation between the object to be hoisted and the target position after applying the adjustment strategy is set as Q1, and the judgment threshold of the displacement deviation Q1 is set as Th. If Q1>Th, the control signal adjustment index after fine-tuning calibration is expressed as: ; in, To fine-tune the control signal adjustment index after calibration, adjust the original control signal index at the same time Update; Th is set as the deviation of the target position, and the deviation range is set to ±3%.

[0019] Embodiment 2: See also Figure 2 , a cantilever crane anti-sway control device based on MLP neural network inversion, the device is used to execute the cantilever crane anti-sway control method based on MLP neural network inversion, comprising: Data acquisition module: determine that the operation buttons for the cantilever crane anti-sway control are the speed control button and the direction control button in sequence; In each operation, the operator's historical behavior data for different combinations of speed control keys and direction control keys, as well as the historical displacement deviation data between the object to be hoisted and the target position and the historical swing state data of the object to be hoisted on the crane are collected; Based on the anti-sway mechanism of the crane anti-sway control system, a basic control signal corresponding to the swing state data is obtained; and the basic displacement change between the object to be hoisted and the target position is recorded by these basic control signals. The basic control signal is used to adjust the basic speed and basic direction of the object to be hoisted on the crane; Model building module: used to build and train a multi-layer perceptron neural network model including an input layer, a hidden layer and an output layer, wherein the input layer is used to receive pre-processed historical behavior data, historical displacement deviation data and historical swing state data under different speed control key and direction control key combinations; the hidden layer is used to perform feature extraction and pattern recognition; The output layer generates the corresponding basic control signal based on the features extracted by the hidden layer to achieve the response to the anti-sway control of the crane; Prediction module: used to match the closest displacement deviation data from the historical displacement deviation data as the first prediction value under the combined control of the current speed control key and the direction control key; at the same time, match the closest swing state data from the historical swing state data as the second prediction value; Basic control signal generation module: used to input the behavior data received in real time, and the matched first prediction value and second prediction value into the constructed multi-layer perceptron neural network model, integrate these data types in the hidden layer, perform feature extraction and pattern recognition, so that the output layer generates the corresponding basic control signal, and obtains the real-time basic displacement change between the object to be hoisted and the target position under the control of the basic control signal; Index generation module: used to obtain the basic control signal, compare and analyze the first predicted value with the real-time basic displacement change, and generate a control signal adjustment index for providing an adjustment strategy for the basic control signal to obtain an adjusted basic control signal; Fine-tuning calibration module: used to use the adjusted basic control signal as the anti-sway control signal of the crane, calculate the displacement deviation between the object to be hoisted and the target position after applying the anti-sway control signal, and set the judgment threshold of the deviation; if the adjusted displacement deviation is still greater than the judgment threshold, the control signal adjustment index is fine-tuned and calibrated, and the control signal adjustment index after fine-tuning and calibrating is used for the next crane anti-sway control cycle.

[0020] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0021] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. Those skilled in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0022] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0023] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. The anti-sway control method of cantilever crane based on MLP neural network inversion is characterized by: The specific steps include: Step S1: Determine that the operation buttons for the anti-sway control of the cantilever crane are the speed control button and the direction control button in sequence; In each operation, the operator's historical behavior data for different combinations of speed control keys and direction control keys, as well as the historical displacement deviation data between the object to be hoisted and the target position and the historical swing state data of the object to be hoisted on the crane are collected; Based on the anti-sway mechanism of the crane anti-sway control system, a basic control signal corresponding to the swing state data is obtained; and the basic displacement change between the object to be hoisted and the target position is recorded by these basic control signals. The basic control signal is used to adjust the basic speed and basic direction of the object to be hoisted on the crane; Step S2: construct and train a multilayer perceptron neural network model including an input layer, a hidden layer and an output layer, wherein the input layer is used to receive preprocessed historical behavior data, historical displacement deviation data and historical swing state data under different speed control key and direction control key combinations; the hidden layer is used to perform feature extraction and pattern recognition; The output layer generates the corresponding basic control signal based on the features extracted by the hidden layer to achieve the response to the anti-sway control of the crane; Step S3: Under the combined control of the current speed control key and the direction control key, the closest displacement deviation data is matched from the historical displacement deviation data as the first prediction value; at the same time, the closest swing state data is matched from the historical swing state data as the second prediction value; Step S4: inputting the behavior data received in real time, and the matched first prediction value and second prediction value into the multi-layer perceptron neural network model constructed in step S2, integrating these data types in the hidden layer, performing feature extraction and pattern recognition, so that the output layer generates a corresponding basic control signal, and obtains the real-time basic displacement change between the object to be hoisted and the target position under the control of the basic control signal; Step S5: acquiring the basic control signal in step S4, comparing and analyzing the first predicted value with the real-time basic displacement change, generating a control signal adjustment index for providing an adjustment strategy for the basic control signal, so as to obtain an adjusted basic control signal; Step S6: Use the adjusted basic control signal as the anti-sway control signal of the crane, calculate the displacement deviation between the object to be hoisted and the target position after applying the anti-sway control signal, and set the judgment threshold of the deviation; if the adjusted displacement deviation is still greater than the judgment threshold, fine-tune and calibrate the control signal adjustment index, and use the fine-tune and calibrated control signal adjustment index for the next crane anti-sway control cycle.

2. The cantilever crane anti-sway control method based on MLP neural network inversion according to claim 1 is characterized in that: Behavior data, displacement deviation and swing state data, including: Behavioral data included hand movement speed and hand movement frequency; Determine the speed range that the speed control key can execute. The speed control key is used to adjust the moving speed of the object to be hoisted on the cantilever crane. Its speed range is divided into an increasing sequence {1,2,…,a,…,A}, where a represents the index of the ath speed partition, and A represents the total number of speed partitions. The larger the value of a is, the higher the speed of the corresponding speed partition is, and each speed partition has the same length; set up represents the average speed of the ath speed partition, Respectively represent the average speeds corresponding to the minimum speed partition 1 and the maximum speed partition A of the speed control key; Divide the direction control buttons into a direction sequence set {1, 2, ..., b, ..., 6} according to the control direction, where b represents the index of the b-th control direction, and 1, 2, 3, 4, 5, and 6 represent forward, backward, left, right, up, and down, respectively; Set the number of combinations between {1,2,…,a,…,A} and {1,2,…,b,…,6} to A×6; The displacement deviation is the vector displacement between the object to be hoisted and the target position under the ath speed partition and the bth type of control direction, and the displacement deviation is recorded as ; The swing state data includes the swing amplitude and swing frequency of the object to be hoisted on the crane under the a-th speed partition and the b-th control direction; and the swing state data under the a-th speed partition and the b-th control direction is recorded as ;in, They represent the swing amplitude and swing frequency in the ath speed partition and the bth type of control direction respectively.

3. The cantilever crane anti-sway control method based on MLP neural network inversion according to claim 2 is characterized in that: The change of foundation displacement includes: Under the same group of speed control keys and direction control keys, determine the minimum speed and maximum speed of the speed control key corresponding to speed partition a, and use this to simulate the swing state data of the object to be hoisted Corresponding numerical fluctuation range, any numerical combination is performed according to the numerical fluctuation range of the swing amplitude and the swing frequency to obtain different swing states of the object to be hoisted in the ath speed partition and the bth type of control direction, and these different swing states are represented by a sequence set {1,2,…,i,…,n}, where i represents the index of the i-th swing state data and n represents the total number of swing state data; the different swing state data in the ath speed partition and the bth type of control direction are represented as ; They represent the swing amplitude and swing frequency corresponding to the i-th swing state data under speed partition a and control direction b respectively; Under speed partition a and control direction b, collect the basic control signal corresponding to the i-th swing state data to determine the basic displacement change between the object to be hoisted and the target position , and record in each simulation process, the basic control signals including the basic speed adjustment signal and the basic direction adjustment signal; in It represents the basic displacement change corresponding to the i-th swing state data under speed partition a and control direction b; They respectively represent the basic speed adjustment signal and the basic direction adjustment signal corresponding to the i-th swing state data under speed partition a and control direction b.

4. The cantilever crane anti-sway control method based on MLP neural network inversion according to claim 3 is characterized in that: The construction of the multi-layer perceptron neural network model includes: Multilayer perceptron neural network model, hereinafter referred to as MLP neural network model; Input data preprocessing: remove outliers, standardize data, and generate model input format; Input layer settings: configure nodes according to the number of features; Hidden layer settings: Set two hidden layers and use the ReLU activation function; Output layer settings: two nodes, output load speed and basic direction adjustment signals; Model settings and optimization settings: Use MSE loss and Adam optimization, and set the learning rate to 0.

001.

5. The cantilever crane anti-sway control method based on MLP neural network inversion according to claim 4 is characterized in that: The acquisition of the first prediction value and the second prediction value specifically includes: recording the speed partition and the control direction corresponding to the current speed control key and the direction control key as a1 and b1, respectively, and a1∈{1,2,…,A}, b1∈{1,2,…,6}; Under the combined control of the current speed partition a1 and the control direction b1, access the historical displacement deviation data and the historical swing state data, use the Euclidean distance to calculate the similarity between the current data and the historical data; select the data with the highest similarity as the displacement deviation closest to the current combination and swing status data ; The index of the current swing state data is recorded as i1, and i1∈{1,2,…,i,…,n}; First prediction value: the displacement deviation to be matched As the first prediction value; Second prediction value: the swing state data that will be matched as the second prediction value; at the same time, extracting the swing amplitude and frequency characteristics; MLP model input: The operator's current behavior data, as well as the characteristic data in the speed partition a1, control direction b1, first predicted value and second predicted value are input into the pre-trained MLP neural network model; the MLP neural network outputs the basic control signal and obtains the real-time basic displacement change between the object to be hoisted and the target position ; in, It indicates the real-time basic displacement change corresponding to the i1th swing state data between the object to be hoisted and the target position in the current speed partition a1 and control direction b1.

6. The cantilever crane anti-sway control method based on MLP neural network inversion according to claim 5 is characterized in that: Define the control signal adjustment index as , the calculation formula is as follows: ； in is the displacement deviation in the first prediction value; is the real-time base displacement change; It is the adjustment sensitivity coefficient, which reflects the sensitivity of the system to changes in the external environment, and its value range is (0, 1); It is a trade-off factor used to balance the influence of the numerator and denominator to avoid exponential imbalance under extreme conditions; is the system response time constant, reflecting the system's response speed to input changes; is the change in base speed adjustment; is the angular change of the base direction adjustment.

7. The cantilever crane anti-sway control method based on MLP neural network inversion according to claim 6 is characterized in that: set up The valid value range of is (0,1), through , settings to ensure this range; when When it is closer to 0, it means that the difference between the first predicted value and the real-time basic displacement change is smaller, and the adjustment amount of the basic control signal is smaller; when The closer it is to 1, the greater the difference between the first predicted value and the real-time basic displacement change, and the greater the adjustment amount of the basic control signal.

8. The cantilever crane anti-sway control method based on MLP neural network inversion according to claim 7 is characterized in that: The adjustment strategy of the basic control signal is as follows: definition The larger the interval value, the more the signal is adjusted for the basic speed. and fundamental direction adjustment signals The larger the adjustment range, the greater the definition of p1<p2<p3; when The value is (0,0.25), and When the base speed is adjusted and fundamental direction adjustment signals On this basis, the p1% value is increased; when The value is (0,0.25), and When the base speed is adjusted and fundamental direction adjustment signals On this basis, the p1% value is reduced; when The value is (0,0.25), and When the speed adjustment signal and fundamental direction adjustment signals No adjustments are required; when The value is in [0.25,0.53], and When the base speed is adjusted and fundamental direction adjustment signals On this basis, the p2% value is increased; when The value is in [0.25,0.53], and When the base speed is adjusted and fundamental direction adjustment signals On this basis, the p2% value is reduced; when The value is (0.53,1), and When the base speed is adjusted and fundamental direction adjustment signals On this basis, the p3% value is increased; when The value is (0.53,1), and When the base speed is adjusted and fundamental direction adjustment signals On this basis, the p3% value is reduced.

9. The cantilever crane anti-sway control method based on MLP neural network inversion according to claim 8 is characterized in that: The displacement deviation between the object to be hoisted and the target position after applying the adjustment strategy is set as Q1, and the judgment threshold of the displacement deviation Q1 is set as Th. If Q1>Th, the control signal adjustment index after fine-tuning calibration is expressed as: ； in, To fine-tune the control signal adjustment index after calibration, adjust the original control signal index at the same time Update is performed and Th is set as the deviation amount from the target position.

10. A cantilever crane anti-sway control device based on MLP neural network inversion, characterized in that: The device is used to execute the cantilever crane anti-sway control method based on MLP neural network inversion according to any one of claims 1 to 9, comprising: Data acquisition module: determine that the operation buttons for the cantilever crane anti-sway control are the speed control button and the direction control button in sequence; In each operation, the operator's historical behavior data for different combinations of speed control keys and direction control keys, as well as the historical displacement deviation data between the object to be hoisted and the target position and the historical swing state data of the object to be hoisted on the crane are collected; Based on the anti-sway mechanism of the crane anti-sway control system, a basic control signal corresponding to the swing state data is obtained; and the basic displacement change between the object to be hoisted and the target position is recorded by these basic control signals. The basic control signal is used to adjust the basic speed and basic direction of the object to be hoisted on the crane; Model building module: used to build and train a multi-layer perceptron neural network model including an input layer, a hidden layer and an output layer, wherein the input layer is used to receive pre-processed historical behavior data, historical displacement deviation data and historical swing state data under different speed control key and direction control key combinations; the hidden layer is used to perform feature extraction and pattern recognition; The output layer generates the corresponding basic control signal based on the features extracted by the hidden layer to achieve the response to the anti-sway control of the crane; Prediction module: used to match the closest displacement deviation data from the historical displacement deviation data as the first prediction value under the combined control of the current speed control key and the direction control key; at the same time, match the closest swing state data from the historical swing state data as the second prediction value; Basic control signal generation module: used to input the behavior data received in real time, and the matched first prediction value and second prediction value into the constructed multi-layer perceptron neural network model, integrate these data types in the hidden layer, perform feature extraction and pattern recognition, so that the output layer generates the corresponding basic control signal, and obtains the real-time basic displacement change between the object to be hoisted and the target position under the control of the basic control signal; Index generation module: used to obtain the basic control signal, compare and analyze the first predicted value with the real-time basic displacement change, and generate a control signal adjustment index for providing an adjustment strategy for the basic control signal to obtain an adjusted basic control signal; Fine-tuning calibration module: used to use the adjusted basic control signal as the anti-sway control signal of the crane, calculate the displacement deviation between the object to be hoisted and the target position after applying the anti-sway control signal, and set the judgment threshold of the deviation; if the adjusted displacement deviation is still greater than the judgment threshold, the control signal adjustment index is fine-tuned and calibrated, and the control signal adjustment index after fine-tuning and calibrating is used for the next crane anti-sway control cycle.

Images