Synchronous control device for double winches of crane

Abstract

The invention relates to the technical field of crane control, and discloses a crane double-winch synchronous control method which comprises the following steps: acquiring working condition data of double winches of a crane in real time, calculating a rotating speed synchronous error and a displacement error of the double winches of the crane, and inputting the working condition data of the double winches of the crane into a pre-trained decision tree working condition classification model; outputting a working condition type, and calling a corresponding synchronous control model based on an LSTM neural network model trained for various working conditions according to the working condition type; and based on the monitored synchronous rotating speed error and synchronous displacement error, if the synchronous rotating speed error and the synchronous displacement error are smaller than the expected control precision, further finely adjusting the control parameters, re-examining the accuracy of the decision tree working condition classification model and the LSTM neural network synchronous control model, and optimizing and improving the models.

Description

Technical Field

[0001] The present invention relates to the technical field of crane control, and particularly relates to a synchronous control method for a double hoist of a crane. Background Art

[0002] In the prior art, the working conditions of the double hoist of the crane are not analyzed in detail, and a unified control method is adopted. There is a lack of analysis of the particularity of the synchronous control of the double hoist under different working conditions. For the different synchronous requirements of the double hoist, no classification control is carried out, resulting in poor synchronous control effect, affecting the stability and operation efficiency of the crane. It is impossible to monitor and adjust the synchronous speed error and synchronous displacement error in real time and accurately, resulting in the error exceeding the expected control accuracy. The pre-trained decision tree working condition classification model and the LSTM neural network model trained for multiple working conditions are not adopted, and the model is not optimized and improved according to the real-time data, resulting in the control model being difficult to adapt to the complex and changeable working conditions. With the change of the working conditions and the operation of the equipment, the control effect gradually deteriorates, and it is impossible to ensure long-term stable synchronous control. The control is carried out according to the preset parameters and cannot be adjusted in time according to the actual synchronous error, resulting in poor control effect. It is impossible to timely discover and correct the problems existing in the model, affecting the reliability and stability of the system.

[0003] Therefore, the present invention provides a synchronous control method for a double hoist of a crane. Summary of the Invention

[0004] The purpose of the present invention is to provide a synchronous control method for a double hoist of a crane to solve at least one of the above-mentioned prior art problems.

[0005] A synchronous control method for a double hoist of a crane includes the following steps:

[0006] By collecting the working condition data of the double hoist of the crane in real time, calculating the rotational speed synchronous error and displacement error of the double hoist of the crane, and inputting the working condition data of the double hoist of the crane into the pre-trained decision tree working condition classification model, the working condition type is output;

[0007] According to the working condition category, the corresponding synchronous control model is called based on the LSTM model trained for multiple working conditions;

[0008] Based on the monitored synchronous rotational speed error and synchronous displacement error, an error curve is drawn, and the control effect value is calculated to analyze and judge the magnitude of the synchronous rotational speed error and the expected control accuracy, and analyze and judge the magnitude of the synchronous displacement error and the expected control accuracy;

[0009] Calculate and analyze the control parameter gradient and fine-tune the control parameters, and re-examine the accuracy of the decision tree working condition classification model and the LSTM model, and optimize and improve the model.

[0010] Advantages of the present invention:

[0011] 1. By collecting the working condition data of the double hoists of the crane in real time, the operating state of the system is obtained, and the real-time and accurate control of the double hoist system is realized; the working condition data and the real-time error are spliced into a feature vector to improve the decision-making accuracy of the model; the working condition data and the real-time error are spliced into a feature vector to improve the decision-making accuracy of the model. The corresponding LSTM model handle is obtained by querying the model library through the hash table, and the appropriate model is quickly located and called; based on the time series input matrix of the real-time synchronization error, the dynamic change trend of the error is analyzed, the prediction error change rate and the system dynamic gain G are output, and accordingly the parameters in the fuzzy control algorithm are dynamically corrected, so that the system can automatically adjust the control parameters according to the change trend of the error; the proportional coefficient is adjusted according to the prediction error change rate, and the integral coefficient is adjusted in combination with the system dynamic gain G to avoid over-adjustment or overshoot of the system;

[0012] 2. Based on gradient calculation, the proportional coefficient and integral coefficient in the fuzzy control algorithm are adjusted; the control parameters are optimized, the error data stages before and after adjustment are divided and detailed statistical analysis is carried out, and the error curve is drawn to show the error change trend; by calculating the sample coverage rate, the working condition category data with insufficient samples is supplemented in time to ensure the adaptability of the model to various working conditions; the threshold and decision tree branch conditions are adjusted according to the classification accuracy to improve the classification accuracy of the model. During the growth of the decision tree, it is decided whether to prune according to the gain ratio to prevent overfitting of the model. The weight variance of the model is calculated, and the L2 regularization method is used to optimize the weight distribution to avoid overfitting. Description of the Drawings

[0013] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0014] Figure 1 is a step flow chart of a method for synchronously controlling double hoists of a crane provided in Embodiment 1 of the present invention;

[0015] Figure 2 is a step flow chart of a method for synchronously controlling double hoists of a crane provided in Embodiment 2 of the present invention. Detailed Embodiments

[0016] To enable those skilled in the art to better understand the solution of the present invention, the following will clearly and completely describe the technical solution in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0017] Embodiment 1

[0018] As Figure 1 shown, a synchronous control method for a double hoist of a crane provided by an embodiment of the present invention specifically includes the following steps:

[0019] Step 1: By collecting the working condition data of the double hoist of the crane in real time, calculating the rotational speed synchronization error and displacement error of the double hoist of the crane, inputting the working condition data of the double hoist of the crane into a pre-trained decision tree working condition classification model, outputting the working condition type, and based on the working condition category, calling the corresponding synchronous control model according to the LSTM neural network model trained for multiple working conditions;

[0020] It should be noted that the working condition data represents the working state data collected in real time when the double hoist of the crane is working, including but not limited to: hoist rotational speed n, hoist displacement s, cargo weight W, ambient wind speed v, and ambient temperature T;

[0021] In some specific embodiments, the working condition data of the double hoist of the crane is obtained, the hoist rotational speeds of the double hoist of the crane obtained are marked as n1 and n2, the hoist displacements of the double hoist of the crane obtained are marked as s1 and s2, and a difference calculation is performed based on the obtained hoist rotational speeds n1 and n2 and the absolute value is taken to obtain the rotational speed synchronization error Δn;

[0022] A difference calculation is performed based on the obtained hoist rotational speeds s1 and s and the absolute value is taken to obtain the rotational speed synchronization error Δs;

[0023] The obtained cargo weight W, ambient wind speed v, and ambient temperature T are input into the pre-trained decision tree working condition classification model to output the working condition category;

[0024] Specifically, the cargo weight W is divided into three types: light load, normal load, and heavy load according to the weight threshold, the ambient wind speed v is divided into three types: weak wind, normal wind, and strong wind according to the wind speed threshold, and the ambient temperature T is divided into three types: low temperature, normal temperature, and high temperature according to the temperature threshold;

[0025] It should be noted that the weight threshold, wind speed threshold, and temperature threshold are classification thresholds preset by those skilled in the art of the present invention according to historical experience and can be adjusted according to the actual situation;

[0026] Output discrete working condition labels through a decision tree classification model, and each working condition label corresponds to a unique working condition code;

[0027] Exemplarily, working condition - 01 corresponds to normal load - normal wind, and working condition - 10 corresponds to heavy load - normal temperature;

[0028] Adopt a key - value pair database, where the key is the working condition code and the value is the binary file path of the pre - trained LSTM model;

[0029] Concatenate the working condition data with the real - time error rotational speed synchronization error Δn and rotational speed synchronization error Δs to form a feature vector;

[0030] Call the decision tree classification interface to obtain the working condition code, query the model library through a hash table, and obtain the handle of the corresponding LSTM model;

[0031] Normalize the real - time collected Δn, Δs, W, v, and T according to the mean and variance corresponding to the working condition category; retain the data of the last k = 50 time steps to form a time - series input matrix;

[0032] Based on the formed time - series input matrix of the real - time synchronous rotational speed error Δn and synchronous displacement error Δs, analyze the dynamic change trend of the error and output the predicted error change rate and the system dynamic gain G;

[0033] After obtaining the predicted error change rate and the system dynamic gain G, use them to dynamically correct the parameter adjustment amounts ΔKp and ΔKi in the fuzzy control algorithm;

[0034] According to the predicted error change rate, analyze and judge the change trend of the predicted error change rate. If the predicted error change rate rises, it indicates that there is a risk of further increase in the error. Increase the proportional coefficient ΔKp to accelerate the system's response speed to the error and enhance the control effect; if the predicted error change rate drops, ΔKp can be reduced to prevent the system from over - adjusting;

[0035] For the integral coefficient ΔKi, adjust it in combination with the system dynamic gain G;

[0036] Compare the system dynamic gain G with a preset gain threshold;

[0037] If the system dynamic gain G is greater than the preset gain threshold G1, it indicates that the system is sensitive to the input signal. Reduce ΔKi to avoid the integral term accumulating too fast and causing system overshoot;

[0038] If the system dynamic gain G is less than the preset gain threshold G0, it indicates that the system is sensitive to the input signal. Increase ΔKi to enhance the integral effect and eliminate the steady - state error;

[0039] After the control parameters are adjusted, the adjusted control parameters are applied to the double-winch synchronous control system;

[0040] The technical solution of the embodiment of the present invention is as follows: by collecting the working condition data of the double winches of the crane in real time, obtaining the operating state of the system, and realizing the real-time and accurate control of the double-winch system; calculating the rotational speed synchronization error and displacement error, discovering the deviation in the operation process of the double-winch system, and providing a quantitative basis for the adjustment of the control strategy; splicing the working condition data and the real-time error into a feature vector to improve the decision-making accuracy of the model; splicing the working condition data and the real-time error into a feature vector to improve the decision-making accuracy of the model, querying the model library through a hash table to obtain the corresponding LSTM model handle, quickly positioning and calling the appropriate model; analyzing the dynamic change trend of the error based on the time series input matrix of the real-time synchronization error, outputting the predicted error change rate and the system dynamic gain G, and dynamically correcting the parameters in the fuzzy control algorithm accordingly, so that the system can automatically adjust the control parameters according to the change trend of the error; adjusting the proportional coefficient according to the predicted error change rate, and combining the system dynamic gain G to adjust the integral coefficient to avoid over-adjustment or overshoot of the system.

[0041] Embodiment 2

[0042] As Figure 1 shown, the embodiment of the present invention provides a method for synchronously controlling the double winches of a crane

[0043] Step 2: Based on the monitored synchronous rotational speed error and synchronous displacement error, if the synchronous rotational speed error and synchronous displacement error are less than the expected control accuracy, further fine-tune the control parameters, and re-examine the accuracy of the decision tree working condition classification model and the LSTM neural network synchronous control model, and optimize and improve the model;

[0044] It should be noted that the control parameters include the proportional coefficient and the integral coefficient;

[0045] In some specific embodiments, continuously monitor the real-time synchronous rotational speed error and synchronous displacement error, compare the error data before and after adjustment, and evaluate the control effect;

[0046] After completing the optimization of the model weights by the adaptive particle swarm algorithm or triggering parameter retraining and applying new control parameters, divide the monitored error data into two stages: before adjustment and after adjustment according to the time sequence;

[0047] Based on the obtained error data before and after adjustment, calculate the mean, variance, maximum value, and minimum value of the error data before adjustment and the error data after adjustment respectively;

[0048] With time as the horizontal axis and the synchronous speed error and synchronous displacement error as the vertical axes respectively, plot the error curves before and after adjustment; through intuitive curve comparison, observe the changing trends of the synchronous speed error and synchronous displacement error at different stages;

[0049] Based on the obtained synchronous speed error Δn and synchronous displacement error Δs, through the formula obtain the control effect value, where α and β are preset proportionality coefficients, and T is the monitoring time interval;

[0050] Based on the obtained control effect value, through the formula calculate the gradient of the proportionality coefficient ΔK p through the formula calculate the gradient of the integral coefficient ΔK i ;

[0051] Based on the obtained gradient of the proportionality coefficient ΔK p and the gradient of the integral coefficient ΔK i analyze and judge the adjustment directions of the proportionality coefficient ΔK p and the integral coefficient ΔK i ;

[0052] Specifically, if the gradient of the proportionality coefficient ΔK p is greater than 0, then decrease the proportionality coefficient ΔK p , if the gradient of the proportionality coefficient ΔK p is less than 0, then increase the proportionality coefficient ΔK p ;

[0053] If the gradient of the proportionality coefficient ΔK i is greater than 0, then decrease the proportionality coefficient ΔK i , if the gradient of the proportionality coefficient ΔK i is less than 0, then increase the proportionality coefficient ΔK i ;

[0054] Based on the gradient descent method, specifically adjust the parameters. Through the formula obtain the adjusted proportionality coefficient, and through the formula obtain the adjusted integral coefficient, where, is the proportionality coefficient before adjustment, is the integral coefficient before adjustment, and η is the preset learning rate;

[0055] Apply the adjusted control parameters to the double-winch synchronous control system of the crane, conduct real-time monitoring again, and recalculate the performance index X g , if X g decreases, it indicates that the adjustment direction is correct, and continue to make fine adjustments; if Xg If it increases, adjust the change direction and adjust the learning rate;

[0056] Based on cross-validation, verify the threshold boundaries of the cargo weight W, environmental wind speed v, and environmental temperature T, and calculate the classification accuracy Acc through the formula where TP is the true positive example, TN is the true negative example, FP is the false positive example, and FN is the false negative example;

[0057] Exemplarily, divide the data close to the weight threshold into two groups. If the cargo weight W is less than the weight threshold, mark it as the small data group. If the cargo weight W is greater than the weight threshold, mark it as the large data group. Input the two groups of data into the decision tree model, calculate the classification accuracy, and analyze the rationality of the judgment threshold. If the classification accuracy Acc is less than the standard accuracy, adjust the weight threshold and the decision tree branch conditions;

[0058] After the optimization based on the threshold, during the growth process of the decision tree, calculate the gain ratio GR of the node through the formula where G(A,T) is the information gain of attribute A for the sample set T, and SI(A,T) is the split information measure of attribute A for the sample set T;

[0059] Based on the obtained gain ratio GR of the node, if the gain ratio GR is less than the preset standard gain ratio, stop the splitting of the corresponding node and perform pruning;

[0060] After the optimization of the decision tree, calculate the variance σ(w) of the LSTM model weights through the formula where N w is the number of weights, w i is the i-th weight, is the mean of the weights;

[0061] If the variance σ(w) of the LSTM model weights is greater than the preset variance threshold, it indicates that the weight distribution is uneven, mark it as the model with uneven weights, and use the regularization method for adjustment;

[0062] Based on the obtained model with uneven weights, obtain the optimized loss function through the formula where λ is the regularization coefficient, and optimize the weight distribution by adjusting the value of the regularization coefficient λ;

[0063] The technical solution of the embodiment of the present invention is as follows: By constructing performance indicators including synchronous speed error and synchronous displacement error, and adjusting the proportional coefficient and integral coefficient in the fuzzy control algorithm based on gradient calculation; optimizing control parameters to effectively reduce errors; continuously monitoring the real-time synchronous speed error and synchronous displacement error, dividing the error data stages before and after adjustment and conducting detailed statistical analysis, and drawing error curves to show the error change trend; examining and optimizing the decision tree working condition classification model from two aspects of sample coverage and boundary conditions, calculating the sample coverage rate, timely supplementing the data of the working condition categories with insufficient samples to ensure the adaptability of the model to various working conditions; using cross-validation to verify at the threshold boundary, adjusting the threshold and decision tree branch conditions according to the classification accuracy to improve the classification accuracy of the model. During the growth process of the decision tree, determine whether to prune according to the gain ratio to prevent overfitting of the model, calculate the weight variance of the model, and use the L2 regularization method to optimize the weight distribution to avoid overfitting.

[0064] The above has described an embodiment of the present invention in detail, but the content described is only a preferred embodiment of the present invention and cannot be considered as limiting the scope of implementation of the present invention; all equal changes and improvements made according to the scope of the application of the present invention shall still fall within the scope covered by the patent of the present invention.

Claims

1. A crane double winch synchronous control method, characterized in that: The following steps are involved: By collecting the working condition data of the crane double winches in real time, calculating the speed synchronization error and displacement error of the crane double winches, inputting the working condition data of the crane double winches into the pre-trained decision tree working condition classification model, and outputting the working condition type; According to the working condition category, the corresponding synchronous control model is called based on the LSTM model trained for various working conditions; Based on the synchronous speed error and synchronous displacement error obtained through monitoring, the error curve is drawn, and the control effect value is calculated to analyze and judge the size of the synchronous speed error and the expected control accuracy. The control parameter gradient is calculated and analyzed and the control parameters are fine-tuned. The accuracy of the decision tree working condition classification model and the LSTM model are reviewed to optimize and improve the model.

2. A crane double winch synchronous control method according to claim 1, characterized in that: The specific method for obtaining the decision tree working condition classification model is: The working condition data of the double winches of the crane are obtained, the winch speeds of the double winches of the crane are marked as n1 and n2, the winch displacements of the double winches of the crane are marked as s1 and s2, and the difference is calculated based on the obtained winch speeds n1 and n2 and the absolute value is taken to obtain the speed synchronization error Δn; Based on the obtained winch speeds s1 and s, the difference is calculated and the absolute value is taken to obtain the speed synchronization error Δs; The obtained cargo weight W, ambient wind speed v and ambient temperature T are input into the pre-trained decision tree working condition classification model to output the working condition category.

3. A crane double winch synchronous control method according to claim 2, characterized in that: The specific method for obtaining the LSTM model is: The decision tree classification model is used to output discrete working condition labels, and each working condition label corresponds to a unique working condition code; A key-value pair database is used, where the key is the working condition code and the value is the binary file path of the pre-trained LSTM model; The working condition data and the real-time error speed synchronization error Δn and the speed synchronization error Δs are spliced ​​into a feature vector; Call the decision tree classification interface, obtain the working condition code, query the model library through the hash table, and obtain the handle of the corresponding LSTM model.

4. A crane double winch synchronous control method according to claim 1, characterized in that: The specific process of obtaining the control parameters is as follows: Normalize the real-time collected Δn, Δs, W, v and T according to the mean and variance corresponding to the working condition category; retain the data of the latest k = 50 time steps to form a time series input matrix; Based on the time series input matrix of the real-time synchronous speed error Δn and synchronous displacement error Δs, the dynamic change trend of the error is analyzed and the predicted error change rate is output. and system dynamic gain G; Based on the obtained prediction error change rate and system dynamic gain G, they are used to dynamically correct the control parameter adjustment amounts ΔKp and ΔKi in the fuzzy control algorithm.

5. A crane double winch synchronous control method according to claim 1, characterized in that: The specific adjustment process of the control parameters is as follows: For the integral coefficient ΔKi, it is adjusted in combination with the system dynamic gain G; Compare the system dynamic gain G with a preset gain threshold; If the system dynamic gain G is greater than the preset gain threshold G1, it indicates that the system is sensitive to the input signal, and ΔKi is reduced to avoid the integral term accumulating too quickly and causing system overshoot; If the system dynamic gain G is less than the preset gain threshold G0, it indicates that the system is sensitive to the input signal, and ΔKi is increased to enhance the integral effect and eliminate the steady-state error; After the control parameter adjustment is completed, the adjusted control parameters are applied to the dual winch synchronous control system.

6. A crane double winch synchronous control method according to claim 1, characterized in that: The specific process of obtaining the error curve is as follows: Continuously monitor the real-time synchronous speed error and synchronous displacement error, compare the error data before and after adjustment, and evaluate the control effect; After completing the adaptive particle swarm algorithm optimization model weight or triggering parameter retraining and applying new control parameters, the monitored error data is divided into two stages, before and after adjustment, in chronological order; Based on the obtained error data before and after adjustment, respectively calculate the mean, variance, maximum value and minimum value of the error data before and after adjustment; With time as the horizontal axis and synchronous speed error and synchronous displacement error as the vertical axes, the error curves before and after adjustment are plotted; through intuitive curve comparison, the changing trends of synchronous speed error and synchronous displacement error at different stages are observed.

7. A crane double winch synchronous control method according to claim 1, characterized in that: The specific method for obtaining the control effect value is: Based on the obtained synchronous speed error Δn and synchronous displacement error Δs, the formula The control effect value is obtained, where α and β are preset proportional coefficients and T is the monitoring time interval.

8. A crane double winch synchronous control method according to claim 1, characterized in that: The specific method for obtaining the control parameter gradient is: Based on the obtained control effect value, the formula Calculate the proportionality coefficient ΔK p The gradient of Calculate the integral coefficient ΔK i The gradient of Based on the obtained proportionality coefficient ΔK p The gradient and integral coefficient ΔK i The gradient of the proportional coefficient ΔK is analyzed and determined p and the integral coefficient ΔK i adjustment direction.

9. A crane double winch synchronous control method according to claim 1, characterized in that: The specific optimization process of the control parameters is: The parameters are adjusted based on the gradient descent method, and the formula Get the adjusted proportional coefficient through the formula Get the adjusted integral coefficient, where is the proportional coefficient before adjustment, is the integral coefficient before adjustment, η is the preset learning rate; Apply the adjusted control parameters to the crane double winch synchronization control system, conduct real-time monitoring again, and recalculate the performance index X g , if X g If X decreases, it means the adjustment direction is correct and continue to make fine adjustments. g If it increases, the direction will be changed and the learning rate will be adjusted; Based on cross-validation, the threshold boundaries of cargo weight W, ambient wind speed v, and ambient temperature T are verified, and the formula The classification accuracy Acc is calculated, where TP is a true positive example, TN is a true negative example, FP is a false positive example, and FN is a false negative example.

10. A crane double winch synchronous control method according to claim 1, characterized in that: The specific optimization method of the model is: After threshold-based optimization, during the decision tree growth process, the formula Calculate the node gain rate GR, where G(A,T) is the information gain of attribute A to sample set T, and SI(A,T) is the split information measure of attribute A to sample set T; Based on the obtained node gain rate GR, if the gain rate GR is less than the preset standard gain rate, the corresponding node splitting is stopped and pruning is performed; After optimization based on decision tree, the formula The variance σ(w) of the LSTM model weight is calculated, where N w is the number of weights, w i is the i-th weight, is the mean of the weights; If the variance σ(w) of the LSTM model weight is greater than the preset variance threshold, it indicates that the weight distribution is uneven and is marked as an uneven weight model. Regularization methods are used to adjust it. Based on the obtained weight imbalance model, the formula The optimized loss function is obtained, where λ is the regularization coefficient, and the weight distribution is optimized by adjusting the value of the regularization coefficient λ.

Images