Four-rope gantry crane remote control automation system
By introducing chaotic quantum coupling module and fractal neural network module into the remote control automation system of four-rope gantry crane, the system's slow response speed and insufficient dynamic adaptability in complex environments is solved, efficient, accurate and dynamic control effects are achieved, and hardware dependence is reduced.
Patent Information
- Application Number
- CN202411812328.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-10
- Publication Date
- 2025-02-28
AI Technical Summary
The existing four-rope gantry crane remote control automation system has problems such as slow response speed and insufficient dynamic adaptability in complex environments, and is highly dependent on hardware and high maintenance costs.
The chaotic quantum coupling module is used to convert complex environment data into high-dimensional dynamic states, and the fractal neural network module is used to realize multi-scale feature extraction and real-time optimization control instruction generation, which improves the adaptability and training efficiency of the network through the training algorithm of fractal dimension modulation.
It significantly improves the system's robustness and operation efficiency in complex environments, reduces hardware dependence, and improves multi-task scheduling capabilities and environmental adaptability.
Smart Images
Figure CN119536099A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of intelligent control and automation systems, and in particular to a remote control automation system for a four-rope gantry crane. Background Art
[0002] As the core equipment for large-scale industrial handling and loading and unloading operations, four-rope gantry cranes occupy an important position in port logistics, warehousing and industrial production scenarios. The traditional manual operation mode has the problems of low efficiency and poor safety. Especially in complex and harsh environments, manual operation has high risks. The remote control automation system can not only improve the operating accuracy and safety of the equipment, but also liberate manpower and reduce operating costs. It is an important direction for the intelligent development of gantry cranes.
[0003] Existing technologies usually rely on high-precision sensors and rule-based control algorithms to achieve data transmission and task scheduling, which has the following defects: Millimeter wave communications are severely affected by physical obstacles and environmental interference, and signal loss or interference may occur in complex scenarios. Existing technologies mainly rely on fixed rules and path planning algorithms, which are difficult to adapt to complex tasks and changing environments in real time, resulting in reduced operational efficiency and safety. Relying on a variety of high-precision hardware sensors, hardware integration is complex and maintenance costs are high, especially in multi-task collaboration and dynamic environments. The efficiency is limited. Summary of the invention
[0004] In view of the many problems existing in the above-mentioned prior art, the present invention provides a remote control automation system for a four-rope gantry crane. The present invention converts complex environmental data into a high-dimensional dynamic state through a chaotic quantum coupling module, and uses a fractal neural network module to realize multi-scale feature extraction and real-time optimization control instruction generation. The training algorithm of fractal dimension modulation improves the adaptability and training efficiency of the network, and finally realizes efficient, accurate and dynamic control of the four-rope gantry crane. The present invention shows remarkable robustness and operating efficiency in complex environments.
[0005] A four-rope gantry crane remote control automation system, comprising: The data acquisition module is used to collect and pre-process the sensor data of the four-rope gantry crane to generate a system status feature data set; A chaotic quantum coupling module is used to input the system state characteristic data set into a chaotic map to generate chaotic sequence data, perform quantum state encoding on the chaotic sequence data, and couple and optimize the chaotic sequence data with the quantum state through a chaotic quantum coupling dynamics equation to generate a chaotic quantum state data set; The fractal neural network module is used to determine the fractal dimension according to the chaotic quantum state data set, construct a fractal neural network structure based on the fractal dimension, use the chaotic quantum state data set to modulate the fractal neural network activation function, and generate control instruction data through the fractal neural network.
[0006] Preferably, the data acquisition module uses a strain type tension sensor to collect the tension values of the four steel ropes, uses an angle encoder to collect the vertical angle and horizontal angle of the four steel ropes, uses a nine-axis inertial measurement unit to collect the acceleration, angular velocity, and magnetic field strength of the sling, and uses a laser radar to scan the working environment to obtain point cloud data.
[0007] Preferably, the data acquisition module processes the tension value through a median filtering algorithm, processes the vertical angle and the horizontal angle through a Kalman filter, fuses the acceleration, angular velocity, and magnetic field strength through a quaternion algorithm, processes the point cloud data through noise reduction and downsampling, synchronizes the tension value, vertical angle, horizontal angle, acceleration, angular velocity, magnetic field strength, and point cloud data according to a unified timestamp, establishes a coordinate transformation matrix for spatial alignment, and uses a local anomaly factor algorithm to detect and correct outliers.
[0008] Preferably, the chaotic quantum coupling module combines the system state characteristic data set to form an initial state vector, performs chaotic mapping by adopting a region-by-region segmented mapping method, generates chaotic sequence data by iterative calculation, and the region-by-region segmented mapping adopts different mapping rules in different value intervals.
[0009] Preferably, the chaotic quantum coupling module converts the chaotic sequence data into a two-level quantum state representation, constructs a Hamiltonian operator containing single-bit and double-bit interactions based on the system dynamics characteristics, and uses the fourth-order Runge-Kutta method to numerically solve the Schrödinger equation to obtain evolution state data.
[0010] Preferably, the chaotic quantum coupling module couples and optimizes the chaotic sequence data with the two-level quantum state representation through the following chaotic quantum coupling dynamics equation:
[0011]
[0012] in, Represents chaotic sequence data; represents the time variable; represents the chaotic mapping function; represents the first weight coefficient; represents the second weight coefficient; represents the quantum state wave function; represents the first operator; represents the second operator; represents an imaginary unit; represents Planck's constant; represents the Hamiltonian operator.
[0013] Preferably, the fractal neural network module uses a box counting method to perform multi-scale covering counting on the chaotic quantum state data set, calculates the fractal dimension through the logarithmic relationship between the covering number of the multi-scale covering counting and the scale parameter; and constructs an iterative function system containing multiple scaling factors and translation vectors based on the fractal dimension.
[0014] Preferably, the fractal neural network module constructs a fractal neural network structure with a four-dimensional input layer, two hidden layers and a four-dimensional output layer based on the fractal dimension, the four-dimensional input layer receives the system state feature data set, the two hidden layers construct network topologies with double and quadruple input dimensions respectively, and the four-dimensional output layer generates the control instruction data.
[0015] Preferably, the fractal neural network module adopts the following hybrid fractal neural network activation function:
[0016] in, represents the activation function; represents input variables; represents the hyperbolic tangent function; represents the quantum state wave function; represents a unitary operator; The norm square representing the inner product of the quantum state; The fractal neural network module uses a multi-objective loss function including a control error term, a chaotic coupling error term and a quantum state error term to train the fractal neural network structure, and uses an improved back propagation algorithm based on the fractal dimension modulation to update the network parameters of the fractal neural network structure.
[0017] Compared with the prior art, the advantages and beneficial effects of the present invention are: It effectively solves the problems of slow response speed and insufficient dynamic adaptability in existing technologies, and demonstrates higher robustness and adaptability in unstructured environments; The fractal dimension is used to optimize the network structure and parameters, which solves the limitations of rule-based control in existing technologies and greatly improves multi-task scheduling capabilities and environmental adaptability. Combining quantum state information to dynamically optimize activation functions and loss functions significantly improves the training efficiency and control accuracy of the network while reducing hardware dependence. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] Figure 1It is a structural block diagram of the system of the present invention; Figure 2 A schematic diagram of the relationship between the data acquisition module and the chaotic quantum coupling module in the present invention; Figure 3 A schematic diagram of the relationship between the chaotic quantum coupling module and the fractal neural network module in the present invention; Figure 4 It is a schematic diagram of the relationship between the fractal neural network module and the control instruction output in the present invention. DETAILED DESCRIPTION
[0019] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present disclosure. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is obvious that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present disclosure.
[0020] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "comprise", "include", etc. used herein indicate the existence of the features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.
[0021] All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.
[0022] like Figure 1 As shown, a four-rope gantry crane remote control automation system comprises: The data acquisition module is used to collect and pre-process the sensor data of the four-rope gantry crane to generate a system status feature data set; Preferably, the data acquisition module uses a strain type tension sensor to collect the tension values of the four steel ropes, uses an angle encoder to collect the vertical angle and horizontal angle of the four steel ropes, uses a nine-axis inertial measurement unit to collect the acceleration, angular velocity, and magnetic field strength of the sling, and uses a laser radar to scan the working environment to obtain point cloud data.
[0023] In the system of the present invention, the data acquisition module serves as the basic perception unit of the system, and collects the system operation status information through multi-source sensors. Specifically, the module uses a strain tension sensor to monitor the tension changes of four steel wire ropes in real time. The sensor is installed at the fixed end of each steel wire rope, and the tension value is obtained by measuring the strain signal generated when the steel wire rope is deformed. In order to accurately perceive the spatial posture of the spreader, the system installs high-precision angle encoders at the connection between the four steel wire ropes and the spreader, and collects the angle (vertical angle) of the steel wire rope relative to the vertical direction and the azimuth (horizontal angle) of the projection relative to the horizontal plane in real time. At the same time, a nine-axis inertial measurement unit is integrated on the spreader body. The unit includes a three-axis accelerometer, a three-axis gyroscope and a three-axis magnetometer, which respectively collect the linear acceleration, angular velocity and geomagnetic field strength data of the spreader for accurate reconstruction of the three-dimensional posture of the spreader. In addition, the system also installs a laser radar on the gantry crane frame, which scans the working environment in all directions by emitting laser beams and receiving reflected signals, and obtains high-precision three-dimensional point cloud data for sensing the distribution of obstacles in the working space.
[0024] Preferably, the data acquisition module processes the tension value through a median filtering algorithm, processes the vertical angle and the horizontal angle through a Kalman filter, fuses the acceleration, angular velocity, and magnetic field strength through a quaternion algorithm, processes the point cloud data through noise reduction and downsampling, synchronizes the tension value, vertical angle, horizontal angle, acceleration, angular velocity, magnetic field strength, and point cloud data according to a unified timestamp, establishes a coordinate transformation matrix for spatial alignment, and uses a local anomaly factor algorithm to detect and correct outliers.
[0025] In order to improve the quality and availability of collected data, the data acquisition module adopts a multi-level data processing strategy. First, for tension data, the median filter algorithm is used to remove sudden noise and outliers in the measurement process to improve the stability of the tension signal. For angle data, considering the influence of mechanical vibration and measurement error, the Kalman filter is used to filter the vertical angle and horizontal angle in real time. The filter effectively suppresses the random fluctuation of the angle signal through the prediction-correction recursive method. In terms of attitude data processing, the system uses the quaternion algorithm to perform multi-sensor fusion of acceleration, angular velocity and magnetic field intensity data, and eliminates the inherent defects of various sensors through complementary filtering to achieve stable and reliable attitude estimation. For the point cloud data obtained by the lidar, the noise reduction algorithm based on statistical analysis and the voxel grid method are used for downsampling processing, which significantly reduces the amount of data while retaining environmental characteristics.
[0026] In order to achieve unified management of multi-source data, the system adopts a data synchronization mechanism based on high-precision timestamps to associate data collected by different sensors through time alignment. In terms of spatial information processing, by establishing a unified world coordinate system, the coordinate transformation matrix is used to convert data in different coordinate systems into the same reference system to achieve spatial alignment of data. Finally, the system uses a local anomaly factor algorithm to detect anomalies in all processed data, identifies and corrects outliers by analyzing the local density distribution of data points, and ensures the validity and reliability of the data. The comprehensive application of these data processing strategies provides high-quality system state characteristic data support for subsequent chaotic quantum coupling optimization and fractal neural network control.
[0027] A chaotic quantum coupling module is used to input the system state characteristic data set into a chaotic map to generate chaotic sequence data, perform quantum state encoding on the chaotic sequence data, and couple and optimize the chaotic sequence data with the quantum state through a chaotic quantum coupling dynamics equation to generate a chaotic quantum state data set; like Figure 2 As shown, in the system of the present invention, the chaotic quantum coupling module first needs to initialize the system state feature data set. This module combines the multi-dimensional sensor data from the data acquisition module, including the tension values, vertical angles, horizontal angles of the four steel wire ropes, and the posture information of the sling, into a unified initial state vector through feature extraction and dimensional mapping. This combination method takes into account the correlation and dimensional consistency between different physical quantities, ensuring that the state vector can fully reflect the dynamic characteristics of the system.
[0028] In order to enhance the nonlinear dynamic characteristics and optimization capabilities of the system, the chaotic quantum coupling module adopts a regional segmented mapping strategy for chaotic mapping. In the specific implementation, the state space is divided into several intervals, and different nonlinear mapping rules are adopted in each interval. When the state variable is in the interval [0, 0.5), a nonlinear mapping rule based on quadratic terms is adopted; when the state variable is in the interval [0.5, 1], a mapping rule based on linear terms is adopted. The design of this segmented mapping mechanism takes into account the differences in the dynamic characteristics of the four-rope gantry crane system under different working conditions, and can better adapt to the nonlinearity and uncertainty of the system. The system generates chaotic sequence data through iterative calculation. Each iteration is based on the calculation results of the previous time, thus forming a deterministic but unpredictable chaotic trajectory.
[0029] The advantages of this segmented mapping method are: first, by using different mapping rules in different intervals, the system's adaptability to complex dynamic environments is enhanced; second, the segmented mapping mechanism can effectively prevent the system from falling into a local optimal solution and improve the system's global search capability; finally, the chaotic sequence generated by this mapping method has good ergodicity and uniform distribution, providing high-quality initial data for subsequent quantum state encoding. In this way, the system can better handle various complex situations encountered by the four-rope gantry crane during actual operation, such as load mutations, external interference, etc., and improve the control accuracy and stability of the system.
[0030] Preferably, the chaotic quantum coupling module combines the system state characteristic data set to form an initial state vector, performs chaotic mapping by adopting a region-by-region segmented mapping method, generates chaotic sequence data by iterative calculation, and the region-by-region segmented mapping adopts different mapping rules in different value intervals.
[0031] In the system of the present invention, the chaotic quantum coupling module first needs to initialize the system state feature data set. This module combines the multi-dimensional sensor data from the data acquisition module, including the tension values of the four steel wire ropes, the vertical angle, the horizontal angle, and the posture information of the sling, into a unified initial state vector through feature extraction and dimensional mapping. This combination method takes into account the correlation and dimensional consistency between different physical quantities, ensuring that the state vector can fully reflect the dynamic characteristics of the system.
[0032] In order to enhance the nonlinear dynamic characteristics and optimization capabilities of the system, the chaotic quantum coupling module adopts a regional segmented mapping strategy for chaotic mapping. In specific implementation, the state space is divided into several intervals, and different nonlinear mapping rules are used in each interval. When the state variable is in the interval [0, 0.5), a nonlinear mapping rule based on quadratic terms is used, as shown in the following formula:
[0033] in, Indicates The state variable value of the iteration, Indicates The state variable value of the iteration, is the bifurcation parameter, usually taken as 3.99.
[0034] When the state variable is in the interval [0.5, 1], the mapping rule based on linear terms is adopted as follows:
[0035] in, and The meaning and The same is recorded in is the bifurcation parameter and takes the value of 3.99.
[0036] The design of this segmented mapping mechanism takes into account the differences in the dynamic characteristics of the four-rope gantry crane system under different working conditions, and can better adapt to the nonlinearity and uncertainty of the system. The system generates chaotic sequence data through iterative calculation, and each iteration is based on the calculation result of the previous one, thus forming a chaotic trajectory that is deterministic but difficult to predict.
[0037] The advantages of this segmented mapping method are: By adopting different mapping rules in different intervals, the system's adaptability to complex dynamic environments is enhanced. The segmented mapping mechanism can effectively prevent the system from falling into a local optimal solution and improve the system's global search capability. The chaotic sequence generated by this mapping method has good ergodicity and distribution uniformity, providing high-quality initial data for subsequent quantum state encoding.
[0038] For example, suppose the initial state vector of the system is [0.3, 0.7, 0.2, 0.8], where each component corresponds to the normalized tension value of the four steel ropes. Applying the segmented mapping, the first two components 0.3 and 0.2 will be The mapping rule of the last two components 0.7 and 0.8 will be The mapping rule. Taking the first component 0.3 as an example, substitute into the formula Calculate the next iteration value:
[0039] Similarly, the iterative values of other components can also be calculated through the corresponding mapping rules. After multiple iterations, the system will generate a set of chaotic sequences, such as [0.8379, 0.5083, 0.9923, 0.0616]. This set of chaotic sequences reflects the nonlinear evolution trajectory of the system state in the phase space and reflects the complex dynamic characteristics inherent in the system. Subsequent quantum state encoding and coupling optimization will be carried out on this basis.
[0040] Preferably, the chaotic quantum coupling module converts the chaotic sequence data into a two-level quantum state representation, constructs a Hamiltonian operator containing single-bit and double-bit interactions based on the system dynamics characteristics, and uses the fourth-order Runge-Kutta method to numerically solve the Schrödinger equation to obtain evolution state data.
[0041] The chaotic quantum coupling module first converts the chaotic sequence data into a two-level quantum state representation. The chaotic sequence data is generated by the previous chaotic mapping and has good ergodicity and distribution uniformity. In this system, the chaotic sequence data is mapped to the quantum state representation, specifically:
[0042] in, , is the normalized value of the chaotic sequence data, ranging from 0 to 1. This representation method ensures that the chaotic sequence data can be efficiently encoded into quantum states, expanding the expressive power of the state space.
[0043] Based on the quantum state representation, the module further constructs the Hamiltonian operator, which describes the interaction between single bits and double bits in the system, specifically:
[0044] in, and For the The Pauli matrix of qubits; It represents the coupling strength and reflects the interaction between bits. This Hamiltonian operator physically reflects the dynamic characteristics of the system, such as the tension change and system stability of the four-rope gantry crane under different working conditions.
[0045] To describe the time evolution of the quantum state, the system uses the Schrödinger equation:
[0046] Here, is Planck's constant, is the system Hamiltonian operator, is the quantum state that changes with time. To solve this equation, the module introduces the fourth-order Runge-Kutta method, which obtains an accurate representation of the quantum state at each time point by iterative calculation of the discretized time step.
[0047] In the specific implementation, the initial quantum state is assumed to be , the time step is . Each iteration is calculated as follows:
[0048] in,
[0049]
[0050]
[0051]
[0052] This method achieves a good balance between accuracy and efficiency and can stably track the time evolution trajectory of quantum states.
[0053] Through the above process, the chaotic quantum coupling module generates evolutionary state data with dynamic characteristics and high-dimensional information. The role of these data in system optimization is mainly reflected in the following aspects: The quantum state representation of chaotic sequences enables the system to explore a larger state space and enhance its adaptability to complex dynamic environments. The Hamiltonian operator reflects the dynamic characteristics of the four-rope gantry crane, enabling the system to capture and optimize nonlinear characteristics under different working conditions. By solving the Schrödinger equation, the module obtains an accurate description of the change of quantum states over time, providing more comprehensive dynamic information for the generation of subsequent control strategies.
[0054] Example, assuming that the four-rope gantry crane is under a specific working condition, the chaotic sequence data is [0.3, 0.7]. Through quantum state representation, the initial state is:
[0055]
[0056] Construct the Hamiltonian operator:
[0057] The Schrödinger equation is solved by the fourth-order Runge-Kutta method to obtain the evolution state:
[0058] in, and is the time-dependent quantum state coefficient.
[0059] Ultimately, the evolutionary data output by the module is used to train and optimize the fractal neural network, enabling it to generate high-precision control instructions and significantly improve the stability and operating efficiency of the four-rope gantry crane.
[0060] Preferably, the chaotic quantum coupling module couples and optimizes the chaotic sequence data with the two-level quantum state representation through the following chaotic quantum coupling dynamics equation:
[0061]
[0062] in, Represents chaotic sequence data; represents the time variable; represents the chaotic mapping function; represents the first weight coefficient; represents the second weight coefficient; represents the quantum state wave function; represents the first operator; represents the second operator; represents an imaginary unit; represents Planck's constant; represents the Hamiltonian operator.
[0063] The chaotic quantum coupling dynamics equation consists of two parts, which respectively describe the evolution of chaotic sequence data and the time variation of quantum state: Dynamic equation of chaotic sequence data: This equation combines the self-evolution of chaotic sequence data with the feedback of quantum states, enabling chaotic data to dynamically respond to the optimization information of quantum states, thus forming a collaborative optimization mechanism.
[0064] The time evolution equation of the quantum state: This equation realizes the dynamic adjustment of quantum state over time by introducing the coupling of Hamiltonian operator and chaotic sequence data, ensuring the adaptability of quantum state to complex environment.
[0065] In the actual application of the four-rope gantry crane, the implementation process of the chaotic quantum coupling module is as follows: Chaotic map generation: Based on the characteristic data such as the tension, angle and posture of the spreader, the initial chaotic state vector is constructed. Chaotic map function Adopting the regional mapping method, nonlinear and linear mapping rules are designed according to different dynamic characteristics, for example:
[0066] in, is the bifurcation parameter, which is used to adjust the chaotic characteristics of the mapping.
[0067] Quantum state construction: chaotic sequence data is mapped to angles , construct the quantum state wave function:
[0068] Use the above initial quantum state to represent the global state of the spreader.
[0069] Coupling optimization: Calculate the expected value based on the interaction between the chaotic sequence and the quantum state wave function , which is specifically expressed as:
[0070] This expected value adjusts the dynamic characteristics of the chaotic sequence data.
[0071] At the same time, the evolution equation of the quantum state is described by the Hamiltonian operator HHH, specifically:
[0072] in, and is the Pauli matrix, is the coupling strength.
[0073] The coupling operator B adjusts the influence of the chaotic sequence to make the quantum state respond dynamically to changes in the external environment.
[0074] Numerical solution: Use the fourth-order Runge-Kutta method to solve the above two equations separately and obtain the time step The chaotic sequence data and quantum state update value in the . The update rule for each iteration is as follows:
[0075]
[0076] The bidirectional coupling optimization between chaotic sequence data and quantum states enables the system to dynamically adapt to load changes and environmental disturbances in the operation of the four-rope gantry crane. The nonlinear characteristics of chaotic mapping combined with the high-dimensional search capability of quantum states significantly improve the global search efficiency of the system and avoid falling into local optimality. The efficient numerical solution of dynamic equations enables the system to generate optimized state data and control instructions in real time, improving the accuracy and stability of the hoist operation.
[0077] In the embodiment, it is assumed that the initial chaotic sequence data of the system is , the quantum state wave function is:
[0078] The mapping function takes parameters , the Hamiltonian operator is set as:
[0079] Through iterative updates of the dynamic equations, new chaotic sequence data and quantum state wave functions are obtained to provide input data for the subsequent fractal neural network module. This mechanism ensures efficient operation and precise control of the four-rope gantry crane system in complex dynamic environments.
[0080] The fractal neural network module is used to determine the fractal dimension according to the chaotic quantum state data set, construct a fractal neural network structure based on the fractal dimension, use the chaotic quantum state data set to modulate the fractal neural network activation function, and generate control instruction data through the fractal neural network.
[0081] like Figure 3 As shown in the figure, the core idea of the fractal neural network module is to optimize the network structure by using the multi-scale characteristics of the fractal dimension, and to design a unique network activation function by combining the dynamic characteristics of chaotic quantum state data. The implementation of the entire module mainly includes the following key steps: Fractal dimension calculation: Receive the chaotic quantum state data set and use the box counting method to calculate the fractal dimension of the system. The box counting method obtains the fractal dimension value by counting the multi-scale coverage of the data. This value reflects the complexity of the data in the feature space. For the control system of the four-rope gantry crane, the calculation result of the fractal dimension can intuitively describe the complexity of the system operation state, such as the dynamic response characteristics of the hoist in a non-stationary state.
[0082] Fractal neural network structure construction: Based on the calculated fractal dimension, the topological structure of the neural network is designed. Specifically, the number of nodes and the connection relationship of the input layer, hidden layer and output layer are dynamically adjusted with the fractal dimension. In this system, the dimension of the input layer is determined by the number of features of the chaotic quantum state data, such as the tension, angle and posture of the hoist. The hidden layer structure adopts a multi-scale configuration according to the fractal dimension, such as a combination of double input dimension and quadruple input dimension. The output layer directly generates the control instructions of the four-rope gantry crane.
[0083] Fractal modulation of activation function: The activation function is modulated in combination with chaotic quantum state data. This design enables the activation function to not only capture nonlinear characteristics, but also dynamically respond to changes in the quantum state of the system, thereby improving the network's expressive power.
[0084] Multi-objective loss function: A multi-objective loss function including control error, chaotic coupling error and quantum state error is constructed to ensure that the neural network can simultaneously optimize the multi-dimensional performance indicators of the system. The design of the loss function fully considers the multi-objective control requirements of the four-rope gantry crane in complex environments, such as maintaining the stability of the spreader, optimizing tension distribution and avoiding obstacles.
[0085] Network parameter optimization: An improved back-propagation algorithm is used to update network parameters in combination with the dynamic modulation mechanism of fractal dimension. The weight update step is adjusted through the modulation factor of fractal dimension, so that the network can adapt to the multi-scale characteristics of the data and avoid overfitting or underfitting.
[0086] like Figure 4 As shown in the figure, the multi-layer structure of the fractal neural network is able to hierarchically process the complex features in the chaotic quantum state data. For example, the input layer captures the basic features of the tension and angle of the spreader, the hidden layer further extracts the nonlinear dynamic characteristics, and the output layer directly generates optimized control instructions.
[0087] The fractal modulation mechanism of the activation function enables the network to respond to quantum state changes in real time, dynamically adjust the modeling capabilities of the system state, and significantly improve the system's adaptability to non-stationary states.
[0088] The introduction of multi-objective loss function ensures that the system achieves a balance between control accuracy, stability and efficiency. Combined with the fractal modulated back propagation algorithm, the network can converge quickly, reduce training time, and improve control performance.
[0089] The fractal structure design of the network enhances the ability to handle multi-scale dynamic characteristics, enabling the control system to respond quickly to external disturbances in complex environments while maintaining the smoothness of the spreader operation.
[0090] Embodiment, in practical application, assuming that the chaotic quantum state data set contains the following features: hanger tension data (4-dimensional), angle data (2-dimensional), posture data (3-dimensional), and the fractal dimension calculation result is 1.7. According to the fractal dimension, the structure of the fractal neural network is designed as follows: The input layer dimension is 9 (the total number of features); The first hidden layer has a dimension of 18 (double the input layer dimension); The second hidden layer dimension is 36 (four times the input layer dimension); The output layer dimension is 4 (spreader control instructions).
[0091] The activation function is modulated by the quantum state wave function in the chaotic quantum state data.
[0092] The trained network inputs the spreader operation status data, and the output control instructions accurately adjust the tension distribution of the four steel wire ropes and the angle adjustment of the spreader to ensure the safety and efficiency of the operation. Under complex working conditions (such as strong winds or load changes), the system shows excellent adaptability and control performance through real-time dynamic adjustment.
[0093] Preferably, the fractal neural network module uses a box counting method to perform multi-scale covering counting on the chaotic quantum state data set, calculates the fractal dimension through the logarithmic relationship between the covering number of the multi-scale covering counting and the scale parameter; and constructs an iterative function system containing multiple scaling factors and translation vectors based on the fractal dimension.
[0094] The box counting method is a classic fractal dimension calculation method. Its core idea is to cover the target data set on grids of different scales, count the number of grids required to cover the data set (covering number), and analyze the relationship between the covering number and the grid scale.
[0095] Specifically for the chaotic quantum state data set, the data is first normalized in the feature space and mapped to The high-dimensional space (where is the feature dimension). Then, choose different scale parameters Divide the grid and calculate the distribution of statistical points in each grid. Indicated in scale The number of grid cells required to cover the dataset.
[0096] Fractal Dimension Calculated by the following formula:
[0097] Through limited The fractal dimension is calculated by linear fitting the logarithmic relationship between the coverage number and the scale parameter. , reflecting the distribution complexity of the data set in the feature space.
[0098] Based on the fractal dimension, an iterated function system (IFS) is constructed. IFS is a fractal geometry generation tool, the core of which is to use multiple scaling factors and translation vectors to generate complex fractal structures in an iterative way.
[0099] In this system, IFS is defined by the following rules:
[0100]
[0101] in, Indicates Iteration function; Represents the scaling factor, which controls the shrinkage ratio of the feature space; Represents the translation vector, which is used to adjust the offset of the iterative function; Represents the number of iterated functions, determined by the fractal dimension.
[0102] Through the calculation results of the fractal dimension, the values of the scaling factor and the translation vector are dynamically adjusted to ensure that the generated network structure can adapt to the multi-scale distribution characteristics of the data.
[0103] In practical application: Data preprocessing and box counting: Receive chaotic quantum state data sets (e.g., feature data such as spreader tension, angle, and posture).
[0104] Normalize the data to high-dimensional space.
[0105] Choose a range of scale parameters ,For example , and gradually divide the feature space.
[0106] At each scale, count the number of grids required to cover the data points , generating a logarithmic plot of the cover number versus the scale parameter.
[0107] Fractal Dimension Calculation: Using Linear Regression and Fitting is performed to obtain the fractal dimension For example, if the linear fit slope is 1.8, then , indicating that the distribution of the dataset in the feature space is close to a two-dimensional structure.
[0108] Iterated function system generation: Determine the number of iterated functions based on the fractal dimension and the scaling factor and translation vector For example, when When , four iterative functions are selected, with scaling factors of 0.5, 0.3, 0.1, and 0.1 respectively, and the translation vector is optimized according to the data distribution.
[0109] The fractal structure in the feature space is generated in an iterative manner to ensure the multi-scale expression capability of the fractal network.
[0110] Fractal dimension guides the construction of network structure by quantifying the complexity of data sets, so that the network can extract global and local features at different scales. For example, the multi-frequency characteristics of tension changes during the operation of the spreader can be captured and effectively modeled through fractal dimension calculation.
[0111] The iterative function system generated based on fractal dimension can dynamically adjust the network structure parameters (such as the number of hidden layer nodes and connection mode), so that the network can adapt to complex environmental changes and improve the control accuracy of spreader operation.
[0112] The fractal neural network generated by IFS maps the complex information in the chaotic quantum state data set to specific control instructions through dynamic modulation and multi-scale feature extraction, achieving precise control of the spreader's operating status. For example, in the case of wind changes or load swings, the system can quickly adjust the wire rope tension and spreader angle.
[0113] In this example, it is assumed that the chaotic quantum state data set contains the tension data (4D), angle data (2D) and posture data (3D) of the spreader operation. The scale parameter is selected by the box counting method. , statistical coverage By linear fitting, the fractal dimension is calculated as =1.8.
[0114] According to the fractal dimension, four iterative functions are constructed, and the scaling factors and translation vectors are:
[0115]
[0116]
[0117]
[0118] The fractal neural network generated by the above function is used to model the operating status of the spreader, and finally outputs control instructions for adjusting the wire rope tension and the spreader angle. In actual operation, the network can quickly respond to load changes and optimize the spreader's operating parameters in real time to ensure the efficiency and safety of the operation.
[0119] Preferably, the fractal neural network module constructs a fractal neural network structure with a four-dimensional input layer, two hidden layers and a four-dimensional output layer based on the fractal dimension, the four-dimensional input layer receives the system state feature data set, the two hidden layers construct network topologies with double and quadruple input dimensions respectively, and the four-dimensional output layer generates the control instruction data.
[0120] The design of the fractal neural network module is inspired by the multi-scale characteristics of fractal theory. It quantifies the complexity of chaotic quantum state data through fractal dimensions and uses it for the dynamic configuration of neural network topology. Specifically, the module includes the following core links: Feature reception of the input layer: The four-dimensional structure of the input layer directly corresponds to the core system state feature data of the four-rope gantry crane, such as the tension, angle, and posture of the spreader. After being processed by the chaotic quantum coupling module, these features provide high-quality data containing nonlinear dynamic characteristics.
[0121] Multi-scale design of hidden layers: The number of nodes in the hidden layer is determined by the input layer dimension and the fractal dimension, and the configuration of double and quadruple input dimensions is adopted. For example, when the input layer is 4-dimensional, the first hidden layer is 8-dimensional and the second hidden layer is 16-dimensional.
[0122] This hierarchical design captures the global and local dynamic characteristics of the system state by gradually increasing the complexity of feature expression.
[0123] Control command generation at the output layer: The four-dimensional structure of the output layer corresponds to the control command of the four-rope gantry crane, such as the tension adjustment of each wire rope or the real-time adjustment of the angle of the hoist. The network directly generates the precise data required for control through multi-layer nonlinear mapping.
[0124] In specific applications: The construction of network topology includes: Input layer: Receives system state feature data sets, where each input node corresponds to a feature dimension. For example, tension, angle, and posture data each provide several features, which together form a 4-dimensional input.
[0125] First hidden layer: The number of nodes is 8 (double the dimension of the input layer), the network connection weights are randomly initialized, and the distribution range is modulated in combination with the fractal dimension.
[0126] The second hidden layer has 16 nodes (four times the dimension of the input layer). It is connected to the first hidden layer to extract complex features step by step.
[0127] Output layer: The number of nodes is 4, which directly generates control instructions for adjusting the wire rope tension and the angle of the sling.
[0128] The use of activation functions includes: Between each layer, a fractal modulation activation function is used. For example, the activation function is:
[0129] in, is the traditional nonlinear activation function, is the quantum state of chaotic quantum state data, is the fractal modulation factor.
[0130] The data flow and training process include: data flow is transmitted from the input layer to the hidden layer, and then to the output layer, and finally the control instructions are generated through layer-by-layer calculation. The loss function is combined with multi-objective optimization design, including control error, dynamic stability and computational efficiency. The network parameters are updated using an improved back propagation algorithm, combined with the dynamic modulation mechanism of fractal dimension, to optimize the convergence speed and accuracy.
[0131] The network structure combines the fractal dimension to dynamically adjust the number of nodes and connection relationships, enabling it to efficiently capture the complex nonlinear characteristics of the spreader operation. For example, under external load changes or wind interference, the network can adaptively adjust the modeling capabilities of tension and angle through step-by-step processing of hidden layers.
[0132] The output layer directly generates instructions for the control of the four-rope gantry crane, and combines the activation function of fractal modulation to ensure that the instructions have high accuracy and real-time performance. In practical applications, the system can maintain the stability and safety of the hoisting operation under complex working conditions.
[0133] The network topology design guided by fractal dimension avoids node redundancy and, combined with the improved back-propagation algorithm, significantly improves the network training and operation efficiency.
[0134] Example, assuming that a four-rope gantry crane is in a certain working condition, its chaotic quantum state data provides the following four-dimensional features: tension data (1 dimension), vertical angle (1 dimension), horizontal angle (1 dimension) and angular velocity of the hoisting device (1 dimension). The specific design of the network is as follows: Input layer: The number of nodes is 4, which receive the above four features respectively.
[0135] First hidden layer: The number of nodes is 8, combined with the double input dimension configuration, and dynamically adjusted through connection weights and fractal modulation factors.
[0136] The second hidden layer has 16 nodes and extracts high-order relationships between features through multi-layer nonlinear mapping.
[0137] Output layer: The number of nodes is 4, corresponding to the tension of each wire rope and the angle adjustment instructions of the sling.
[0138] During operation, assuming that the initial tension state of the spreader needs to be adjusted in real time, the network inputs real-time data, extracts features step by step through the hidden layer, and finally outputs four-dimensional control instructions, which are the tension change applied to each wire rope and the angle correction value of the spreader.
[0139] Preferably, the fractal neural network module adopts the following hybrid fractal neural network activation function:
[0140] in, represents the activation function; represents input variables; represents the hyperbolic tangent function; represents the quantum state wave function; represents a unitary operator; The norm square representing the inner product of the quantum state; The fractal neural network module uses a multi-objective loss function including a control error term, a chaotic coupling error term and a quantum state error term to train the fractal neural network structure, and uses an improved back propagation algorithm based on the fractal dimension modulation to update the network parameters of the fractal neural network structure.
[0141] The activation function is passed Capture nonlinear characteristics and use the dynamic characteristics of quantum state wave functions to modulate the output to adapt it to the complex dynamic characteristics of the system. For example, when the system load changes dramatically, the modulation factor of the quantum state can reflect the environmental changes in real time, dynamically adjust the output value of the activation function, and enhance the network's ability to model nonlinear environments.
[0142] The loss function design of the module combines the multi-objective optimization requirements of the system, including the following key items: The control error term is used to measure the deviation between the actual control command and the target control value; The chaotic coupling error term reflects the synergy between the output data of the chaotic quantum coupling module and the input data of the fractal neural network; The quantum state error term evaluates the degree of response of the network output to the quantum state modulation factor.
[0143] The multi-objective loss function is in the form of:
[0144] in, , and represent control error, chaotic coupling error and quantum state error respectively, , , is the weight coefficient, which is used to balance the relative importance of each objective.
[0145] The network parameter update adopts the back propagation algorithm, but combines the fractal dimension to modulate the update step size and direction. Specifically, the gradient update rule is:
[0146] in, is the learning rate; is the fractal dimension modulation factor, defined as ,in is the current fractal dimension, is the base fractal dimension, is the modulation coefficient.
[0147] The introduction of fractal dimension modulation ensures that the network can dynamically adjust the learning rate and update direction to adapt to the multi-scale distribution characteristics of the data, thereby improving training efficiency and optimization effects.
[0148] The hybrid activation function combines nonlinearity and quantum state modulation characteristics, which can effectively capture complex dynamic characteristics and adapt to environmental changes. For example, in the case of load fluctuations or external interference, the activation function can dynamically adjust the response amplitude of the network output to ensure the accuracy and real-time performance of the control instructions.
[0149] The introduction of multi-objective loss functions enables the network to simultaneously optimize control accuracy, dynamic stability, and energy efficiency. For example, while ensuring the balance of wire rope tension, the angle adjustment of the hoist and the operating stability are optimized, which improves the overall performance of the system.
[0150] The fractal dimension modulated back-propagation algorithm significantly improves the network training efficiency by dynamically adjusting the learning rate and update step size, and shows stronger robustness when facing complex data distribution.
[0151] In the embodiment, it is assumed that the system state feature data of the four-rope gantry crane includes 4-dimensional features such as tension, angle and posture. The activation function design and training process of the module are as follows: After the input tension and angle features are calculated by the previous layer weights, the neuron input variables are obtained. The output of the activation function is:
[0152] In which, it is assumed that the quantum state wave function , the unitary operator represents the quantum state rotation matrix, and the modulation factor is calculated as =0.64. The final activation function output is =0.462×0.64=0.296.
[0153] Assume that the training goal is to balance the wire rope tension and the angle adjustment of the hoist. The control error between the actual output and the target value is 0.02, the chaotic coupling error is 0.03, the quantum state error is 0.01, and the weight coefficients are The total loss is:
[0154] Based on loss gradient, combined with fractal dimension =1.7 and base dimension =2.0, the modulation factor is The parameter update step size is:
[0155] Finally, after multiple iterations of training, the network outputs precise control instructions for tension adjustment and angle correction, providing strong support for the efficient and stable operation of the four-rope gantry crane.
[0156] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware.
[0157] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.
Claims
1. A four-rope gantry crane remote control automation system, characterized in that: include: The data acquisition module is used to collect and pre-process the sensor data of the four-rope gantry crane to generate a system status feature data set; A chaotic quantum coupling module is used to input the system state characteristic data set into a chaotic map to generate chaotic sequence data, perform quantum state encoding on the chaotic sequence data, and couple and optimize the chaotic sequence data with the quantum state through a chaotic quantum coupling dynamics equation to generate a chaotic quantum state data set; The fractal neural network module is used to determine the fractal dimension according to the chaotic quantum state data set, construct a fractal neural network structure based on the fractal dimension, use the chaotic quantum state data set to modulate the fractal neural network activation function, and generate control instruction data through the fractal neural network.
2. The four-rope gantry crane remote control automation system according to claim 1 is characterized in that: The data acquisition module uses a strain-type tension sensor to collect the tension values of the four steel ropes, uses an angle encoder to collect the vertical angles and horizontal angles of the four steel ropes, uses a nine-axis inertial measurement unit to collect the acceleration, angular velocity, and magnetic field strength of the sling, and uses a laser radar to scan the working environment to obtain point cloud data.
3. The four-rope gantry crane remote control automation system according to claim 2 is characterized in that: The data acquisition module processes the tension value through a median filtering algorithm, processes the vertical angle and the horizontal angle through a Kalman filter, fuses the acceleration, angular velocity, and magnetic field intensity through a quaternion algorithm, processes the point cloud data through noise reduction and downsampling, synchronizes the tension value, vertical angle, horizontal angle, acceleration, angular velocity, magnetic field intensity, and point cloud data according to a unified timestamp, establishes a coordinate transformation matrix for spatial alignment, and uses a local anomaly factor algorithm to detect and correct outliers.
4. The four-rope gantry crane remote control automation system according to claim 1 is characterized in that: The chaotic quantum coupling module combines the system state characteristic data sets to form an initial state vector, performs chaotic mapping in a region-by-region segmented mapping manner, and generates chaotic sequence data through iterative calculation. The region-by-region segmented mapping adopts different mapping rules in different value intervals.
5. The four-rope gantry crane remote control automation system according to claim 4 is characterized in that: The chaotic quantum coupling module converts the chaotic sequence data into a two-level quantum state representation, constructs a Hamiltonian operator containing single-bit and double-bit interactions based on the system dynamics characteristics, and uses the fourth-order Runge-Kutta method to numerically solve the Schrödinger equation to obtain evolution state data.
6. The four-rope gantry crane remote control automation system according to claim 5 is characterized in that: The chaotic quantum coupling module couples and optimizes the chaotic sequence data with the two-level quantum state representation through the following chaotic quantum coupling dynamics equation: in, Represents chaotic sequence data; represents the time variable; represents the chaotic mapping function; represents the first weight coefficient; represents the second weight coefficient; represents the quantum state wave function; represents the first operator; represents the second operator; represents an imaginary unit; represents Planck's constant; represents the Hamiltonian operator.
7. The four-rope gantry crane remote control automation system according to claim 1 is characterized in that: The fractal neural network module uses a box counting method to perform multi-scale covering counting on the chaotic quantum state data set, calculates the fractal dimension through the logarithmic relationship between the covering number of the multi-scale covering counting and the scale parameter; and constructs an iterative function system including multiple scaling factors and translation vectors based on the fractal dimension.
8. The four-rope gantry crane remote control automation system according to claim 7 is characterized in that: The fractal neural network module constructs a fractal neural network structure with a four-dimensional input layer, two hidden layers and a four-dimensional output layer based on the fractal dimension. The four-dimensional input layer receives the system state feature data set, the two hidden layers construct network topologies with double and quadruple input dimensions respectively, and the four-dimensional output layer generates the control instruction data.
9. The four-rope gantry crane remote control automation system according to claim 1, characterized in that: The fractal neural network module adopts the following hybrid fractal neural network activation function: in, represents the activation function; represents input variables; represents the hyperbolic tangent function; represents the quantum state wave function; represents a unitary operator; represents the norm square of the inner product of the quantum state; The fractal neural network module uses a multi-objective loss function including a control error term, a chaotic coupling error term and a quantum state error term to train the fractal neural network structure, and uses an improved back propagation algorithm based on the fractal dimension modulation to update the network parameters of the fractal neural network structure.