A multifunction integrated cable simulation working condition test platform

Abstract

The present application relates to the field of cable simulation working condition test, and more particularly to a multifunctional integrated cable simulation working condition test platform, comprising: a cable test module for collecting temperature parameters and load parameters of a test cable tested by a heater and a multi-degree-of-freedom loading head; a neural network module for determining control parameters by fitting the temperature parameters and the load parameters through a control parameter fitting model based on a BP neural network architecture; and a control module for generating an execution temperature of the heater and an execution force of the multi-degree-of-freedom loading head through a control algorithm of the control parameters. The present application enables the multifunctional test platform to autonomously learn control parameters corresponding to complex extreme working conditions without relying on artificial preset models, improves the adaptability of the test platform to complex working condition tests, and realizes stable and accurate parameter control of the multi-degree-of-freedom loading head and the heater of the platform.

Description

Technical Field

[0001] This invention relates to the field of cable simulation testing, and more particularly to a multifunctional integrated cable simulation testing platform. Background Technology

[0002] As a key component in power transmission, communication networks, and industrial sectors, the reliability of cables directly affects the safe and stable operation of infrastructure. Furthermore, with the rapid development of new energy, rail transportation, and deep-sea engineering, cables need to operate under complex conditions for extended periods, such as soil pressure in buried environments, dynamic tension during overhead installations, and heat radiation near high-temperature pipelines. This places stringent requirements on the testing of cables' mechanical properties, temperature resistance, and durability.

[0003] Currently, cable reliability testing mainly relies on multiple independent devices, such as tensile testing machines for tensile performance testing, compression testing machines to simulate soil compression, and high and low temperature chambers to control ambient temperature. However, this decentralized testing model has significant limitations. It is difficult to reproduce actual working conditions, such as the combined temperature and force conditions in buried compression, overhead tension, and high-temperature pipeline compression. Independent devices can only simulate a single physical field and cannot reflect the impact of multi-stress coupling on cable insulation aging, structural deformation, and failure mechanisms, leading to discrepancies between test results and actual service performance. Furthermore, the dispersion of multiple devices results in cumbersome testing procedures, requiring multiple sample transfers, low efficiency, and susceptibility to errors. The procurement and maintenance costs of multiple devices are high, and the utilization rate of laboratory space is low.

[0004] Therefore, the current trend is to use integrated testing platforms with multi-parameter collaborative control and intelligent dynamic adjustment capabilities to achieve cable simulation testing under complex actual working conditions.

[0005] Among these challenges, how to achieve the simulation of complex and extreme working conditions through the coordinated control of a multi-degree-of-freedom loading head and a heater in a multi-functional integrated cable simulation working condition test platform is a technical problem that needs to be solved. Summary of the Invention

[0006] To address this, the present invention provides a multifunctional integrated cable simulation test platform. Through a cable testing module, it achieves the coordinated application of temperature and multi-degree-of-freedom mechanical loads, simulating the complex and extreme temperature and mechanical conditions faced by cables in actual operation. By using a control parameter fitting model based on a BP neural network architecture, it can accurately capture the nonlinear relationship between temperature and load parameters. This allows the multifunctional test platform to learn control parameters corresponding to complex and extreme conditions autonomously, without relying on manually preset models, thus improving its adaptability to complex testing conditions and achieving stable and precise parameter control of the platform's multi-degree-of-freedom loading head and heater.

[0007] To achieve the above objectives, this invention proposes a multifunctional integrated cable simulation working condition test platform, comprising:

[0008] The cable testing module is used to collect temperature and load parameters of the test cable by the heater and the multi-degree-of-freedom loading head.

[0009] The neural network module is used to fit the temperature parameters and the load parameters to a control parameter fitting model based on a BP neural network architecture, determine the control parameters, and adjust the model parameters of the control parameter fitting model through backpropagation by testing a stationary loss function.

[0010] The control module is used to generate the operating temperature of the heater and the operating force of the multi-degree-of-freedom loading head by means of the control parameters through a control algorithm.

[0011] Furthermore, the neural network module includes:

[0012] The neural network forward propagation submodule is used to fit the temperature parameter and the load parameter to a control parameter fitting model based on a BP neural network architecture to generate the control parameter.

[0013] The neural network backpropagation submodule is used to adjust the model parameters by backpropagation of the load parameters to the control parameter fitting model through a test stationary loss function.

[0014] Furthermore, the load parameters include a force-displacement sequence and a force parameter sequence; the control parameter fitting model includes a thermodynamic compensation mechanism, a loading head resonance suppression sub-model, a gated fusion layer, a convolutional pooling layer, a fully connected layer, and a constraint layer; and the neural network forward propagation sub-module includes:

[0015] The input integration unit is used to integrate the temperature parameter, the force-displacement sequence, the force parameter sequence, the temperature setpoint, the displacement setpoint, and the force setpoint into channels of the basic input vector, and to generate a thermodynamic input vector based on the basic input vector and the thermal conductivity of the cable material, and to generate a loading head motion input vector based on the basic input vector and the stiffness of the cable material.

[0016] A thermodynamic processing unit is used to generate thermodynamic features by passing the thermodynamic input vector through a thermodynamic compensation mechanism;

[0017] The resonance suppression unit is used to pass the loading head motion input vector through the loading head resonance suppression sub-model to generate loading head features;

[0018] A gated fusion unit is used to combine the thermodynamic features and the loading head features through a gated fusion layer to generate a comprehensive physical feature.

[0019] A convolutional unit is used to pass the basic input vector through a convolutional pooling layer to generate comprehensive temporal features;

[0020] The mapping unit is used to perform feature mapping between the integrated temporal features and the integrated physical features through a fully connected layer to generate the original control parameters;

[0021] A constraint unit is used to output constraints on the original control parameters based on the thermal conductivity and stiffness of the cable material through a constraint layer, so as to generate the control parameters.

[0022] Furthermore, the thermodynamic compensation mechanism includes an expanded convolutional layer and a thermal delay compensation layer, and the thermodynamic processing unit includes:

[0023] The dilated convolutional subunit is used to extract the thermodynamic flow input vector temporally through the dilated convolutional layer to generate thermodynamic temporal features;

[0024] A thermal delay compensation subunit is used to generate the thermodynamic features by passing the current thermodynamic time series features and historical thermodynamic time series features through a thermal delay compensation layer.

[0025] Furthermore, the loaded head resonance suppression sub-model includes a differential feature extraction layer and a resonance suppression operation layer, and the resonance suppression unit includes:

[0026] The differential enhancement extraction subunit is used to reduce the fixed fluctuations of the loading head motion input vector through differential operation to generate loading head temporal features;

[0027] The resonance suppression subunit is used to pass the mechanical timing characteristics through the resonance suppression operation layer to generate the loading head characteristics.

[0028] Further, the convolutional pooling layer includes a first convolutional pooling layer, a second convolutional pooling layer, and a flattening layer, and the kernel size of the first convolutional pooling layer is smaller than the kernel size of the second convolutional pooling layer. The convolutional unit includes:

[0029] The first convolutional pooling subunit is used to pass the basic input vector through the first convolutional pooling layer to generate local temporal features;

[0030] The second convolutional pooling subunit is used to pass the local temporal features through the second convolutional pooling layer to generate global temporal features;

[0031] The flattening layer subunit is used to pass the global features through the flattening layer to generate the comprehensive temporal features.

[0032] Furthermore, the control parameters include temperature control parameters and force control parameters, and the constraint unit includes:

[0033] Thermodynamic constraint subunit, used to thermodynamically constrain temperature control parameters based on the temperature parameters and the maximum thermal energy of the heater;

[0034] The loading head constraint subunit is used to constrain the force control parameters based on the cable material stiffness and the maximum stress of the loading head.

[0035] In particular, the thermodynamic compensation mechanism and the loading head resonance suppression sub-model are specifically designed to learn the temperature field hysteresis effect and the mechanical system resonance interference, respectively. The gated fusion layer achieves adaptive coupling of physical field characteristics through dynamic weight allocation. This hierarchical learning architecture enables the BP neural network to accurately fit the strong nonlinear relationship under extreme scenarios, reducing the adjustment error of control parameters, so as to achieve stable and accurate parameter control of the multi-degree-of-freedom loading head and heater of the test platform under complex extreme conditions.

[0036] Furthermore, the neural network backpropagation submodule includes:

[0037] The loss function construction unit is used to construct the test stationary loss function based on the control parameter error term and the control parameter change rate.

[0038] Furthermore, the neural network backpropagation submodule also includes:

[0039] The backpropagation unit is used to perform backpropagation adjustments on the control parameter fitting model in the directions of control parameters, convolution, thermodynamics, and resonance suppression based on the test stationary loss function.

[0040] Furthermore, the heater and multi-degree-of-freedom loading head are integrated and installed inside the environmental simulation chamber;

[0041] The multi-degree-of-freedom loading head includes a hydraulic cylinder array and a universal joint for controlling the hydraulic cylinder array;

[0042] The heater includes at least a semiconductor cooling chip, an infrared heating plate, and a convection fan.

[0043] In particular, by providing real-time feedback on the load parameter sequence, the network weights are dynamically adjusted along gradient paths in the control parameter direction, convolution direction, thermodynamic direction, and resonance suppression direction. This improves the parameter convergence speed under extreme conditions compared to traditional unidirectional gradient descent.

[0044] Compared with the prior art, the beneficial effects of the present invention are that it achieves the coordinated application of temperature and multi-degree-of-freedom mechanical load through the cable testing module, which can simulate the complex extreme working conditions of temperature mechanics faced by the test cable in actual operation. Through the control parameter fitting model based on the BP neural network architecture, it can accurately capture the nonlinear relationship between temperature and load parameters, so that the multi-functional test platform does not need to rely on manually preset models, but can autonomously learn the control parameters corresponding to complex extreme working conditions, improve the adaptability of the test platform to complex working condition testing, and realize stable and accurate parameter control of the platform's multi-degree-of-freedom loading head and heater.

[0045] In particular, this invention employs a thermodynamic compensation mechanism and a loading head resonance suppression sub-model to specifically learn the temperature field hysteresis effect and the mechanical system resonance interference, respectively. The gated fusion layer achieves adaptive coupling of physical field characteristics through dynamic weight allocation. This hierarchical learning architecture enables the BP neural network to accurately fit the strongly nonlinear relationship under extreme scenarios, reducing the adjustment error of control parameters, thereby achieving stable and accurate parameter control of the multi-degree-of-freedom loading head and heater of the test platform under complex and extreme conditions.

[0046] In particular, this invention dynamically corrects network weights along gradient paths—the control parameter direction, convolution direction, thermodynamic direction, and resonance suppression direction—through real-time feedback of the load parameter sequence. This improves the parameter convergence speed under extreme conditions compared to traditional unidirectional gradient descent. Attached Figure Description

[0047] Figure 1 This is a schematic diagram of the structure of the multifunctional integrated cable simulation working condition test platform according to an embodiment of the present invention;

[0048] Figure 2 This is a flowchart illustrating the multifunctional integrated cable simulation working condition testing platform according to an embodiment of the present invention.

[0049] Figure 3 This is a schematic diagram of the forward propagation process of the control parameter fitting model of the multifunctional integrated cable simulation working condition test platform according to an embodiment of the present invention;

[0050] Figure 4 This is a schematic diagram of the backpropagation process of the control parameter fitting model of the multifunctional integrated cable simulation working condition test platform according to an embodiment of the present invention. Detailed Implementation

[0051] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0052] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0053] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate the direction or positional relationship, are based on the direction or positional relationship shown in the drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0054] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0055] like Figures 1 to 4 As shown, this invention provides a multifunctional integrated cable simulation test platform. Through a cable testing module, it achieves the coordinated application of temperature and multi-degree-of-freedom mechanical loads, simulating the complex and extreme temperature and mechanical conditions faced by cables in actual operation. By using a control parameter fitting model based on a BP neural network architecture, it can accurately capture the nonlinear relationship between temperature and load parameters. This allows the multifunctional test platform to learn control parameters corresponding to complex and extreme conditions autonomously, without relying on manually preset models, thus improving its adaptability to complex testing conditions and achieving stable and precise parameter control of the platform's multi-degree-of-freedom loading head and heater.

[0056] like Figure 1 and 2 As shown, this embodiment proposes a multifunctional integrated cable simulation working condition test platform, including:

[0057] The cable testing module is used to collect temperature and load parameters of the test cable by the heater and the multi-degree-of-freedom loading head.

[0058] The neural network module is used to fit the temperature parameters and the load parameters to a control parameter fitting model based on a BP neural network architecture, determine the control parameters, and adjust the model parameters of the control parameter fitting model through backpropagation by testing a stationary loss function.

[0059] The control module is used to generate the operating temperature of the heater and the operating force of the multi-degree-of-freedom loading head by means of the control parameters through a control algorithm.

[0060] In particular, it has achieved stable and accurate adjustment of test parameters for the multi-functional integrated cable simulation test platform under extreme temperature and force conditions, thereby improving the adaptability of the test platform to complex test conditions.

[0061] Furthermore, the heater and multi-degree-of-freedom loading head are integrated and installed inside the environmental simulation chamber;

[0062] The multi-degree-of-freedom loading head includes a hydraulic cylinder array and a universal joint for controlling the hydraulic cylinder array;

[0063] The heater includes at least a semiconductor cooling chip, an infrared heating plate, and a convection fan.

[0064] In particular, by combining the hydraulic cylinder array of the multi-degree-of-freedom loading head with the universal joint, it is possible to simulate the multi-directional mechanical loads that cables bear in actual use, such as tensile, bending, and torsional composite stresses. Combined with the temperature control function in the environmental simulation chamber, it is possible to achieve coordinated testing of mechanical and environmental factors, which is closer to the actual complex working conditions.

[0065] Specifically, the multi-degree-of-freedom loading head of the multi-functional integrated cable simulation working condition test platform adopts a mechanism of hydraulic cylinder array and universal joint coupling, including: 3 sets of symmetrically arranged servo hydraulic cylinders (axial ±50kN, lateral ±10kN); the output end of the hydraulic cylinder array is connected to an adaptive universal joint, and the vector synthesis of tension / compression / lateral thrust is realized through ball joints; a six-dimensional force sensor is set at the end of the loading head to provide real-time feedback of composite load values.

[0066] Specifically, the multi-functional integrated cable simulation test platform adopts a compact environmental simulation chamber, which integrates a multi-degree-of-freedom loading head and a heater. The environmental simulation chamber uses a double-layer stainless steel body with ceramic fiber insulation material filling the interlayer. The heater is preferably a semiconductor cooling chip array arranged at the top of the environmental simulation chamber and an infrared heating plate installed at the bottom. The temperature uniformity inside the chamber is achieved through a convection fan.

[0067] like Figure 3 As shown, the neural network module further includes:

[0068] The neural network forward propagation submodule is used to fit the temperature parameter and the load parameter to a control parameter fitting model based on a BP neural network architecture to generate the control parameter.

[0069] The neural network backpropagation submodule is used to adjust the model parameters by backpropagation of the load parameters to the control parameter fitting model through a test stationary loss function.

[0070] In particular, the forward propagation of the BP neural network architecture enables nonlinear fitting of temperature and load parameters, dynamically generating control parameters adapted to complex operating conditions. By combining the backpropagation submodule with a test stationary loss function, model parameters can be dynamically adjusted based on real-time load parameter feedback, achieving online iterative optimization of the control parameter fitting model and improving the system's adaptability to operating condition fluctuations and long-term operational stability. Therefore, compared to traditional linear models, it is better able to capture the coupling relationships between parameters, improve the accuracy of the control strategy, and obtain accurate performance parameters of the test cable under complex operating conditions, thus laying the foundation for more precise performance improvement of cables under complex operating conditions.

[0071] Furthermore, the load parameters include a force-displacement sequence and a force parameter sequence; the control parameter fitting model includes a thermodynamic compensation mechanism, a loading head resonance suppression sub-model, a gated fusion layer, a convolutional pooling layer, a fully connected layer, and a constraint layer; and the neural network forward propagation sub-module includes:

[0072] The input integration unit is used to integrate the temperature parameter, the force-displacement sequence, the force parameter sequence, the temperature setpoint, the displacement setpoint, and the force setpoint into channels of the basic input vector, and to generate a thermodynamic input vector based on the basic input vector and the thermal conductivity of the cable material, and to generate a loading head motion input vector based on the basic input vector and the stiffness of the cable material.

[0073] A thermodynamic processing unit is used to generate thermodynamic features by passing the thermodynamic input vector through a thermodynamic compensation mechanism;

[0074] The resonance suppression unit is used to pass the loading head motion input vector through the loading head resonance suppression sub-model to generate loading head features;

[0075] A gated fusion unit is used to combine the thermodynamic features and the loading head features through a gated fusion layer to generate a comprehensive physical feature.

[0076] A convolutional unit is used to pass the basic input vector through a convolutional pooling layer to generate comprehensive temporal features;

[0077] The mapping unit is used to perform feature mapping between the integrated temporal features and the integrated physical features through a fully connected layer to generate the original control parameters;

[0078] A constraint unit is used to output constraints on the original control parameters based on the thermal conductivity and stiffness of the cable material through a constraint layer, so as to generate the control parameters.

[0079] In particular, by using the input integration unit, multi-dimensional parameters such as temperature, load, and material properties (thermal conductivity and stiffness of cable materials) are constructed into dual-channel input vectors of thermodynamics and kinematics, achieving deep integration of mechanical and thermal parameters. This avoids the limitations of single-physics field analysis in traditional models and improves the correlation and accuracy of control parameters under complex working conditions.

[0080] Specifically, dimension 1 of the basic input vector includes the temperature parameter, the force-displacement sequence, the force parameter sequence, the temperature setpoint, the displacement setpoint, and the force setpoint, which are six data channels. Dimension 2 of the basic input vector is the time step, which is obtained by taking the most recent 60 sampling points (approximately 3 seconds of data) through a sliding window.

[0081] It should be noted that the force displacement parameter sequence and displacement setting value mentioned in this embodiment are the force position of the multi-degree-of-freedom loading head on the test cable collected by the six-dimensional force sensor of the multi-degree-of-freedom loading head, and the force parameter sequence and force setting value mentioned in this embodiment are the magnitude of the force exerted by the multi-degree-of-freedom loading head on the test cable.

[0082] Specifically, the control parameters include nine PID control parameters corresponding to the heater temperature controller, the force position controller of the multi-degree-of-freedom loading head, and the force magnitude controller of the multi-degree-of-freedom loading head.

[0083] Furthermore, the thermodynamic compensation mechanism includes an expanded convolutional layer and a thermal delay compensation layer, and the thermodynamic processing unit includes:

[0084] The dilated convolutional subunit is used to extract the thermodynamic flow input vector temporally through the dilated convolutional layer to generate thermodynamic temporal features;

[0085] A thermal delay compensation subunit is used to generate the thermodynamic features by passing the current thermodynamic time series features and historical thermodynamic time series features through a thermal delay compensation layer.

[0086] In particular, the thermal delay compensation layer integrates current and historical thermal time series characteristics to construct a time correlation model of temperature field evolution. This model can accurately quantify and offset the thermal conduction hysteresis effect, such as the temperature response delay caused by the thermal inertia of cable materials, and reduce test errors caused by temperature control hysteresis, such as deviations in the timing of thermal stress loading. It is especially suitable for cyclic testing scenarios under complex working conditions.

[0087] Specifically, the thermodynamic flow input vector, in addition to the basic input vector, also includes at least the following: Where T represents the current temperature, i.e., the temperature parameter. Indicates the temperature setpoint. Indicates heater power. Indicates the thermal conductivity of the cable material. Indicates the cooling rate of the cable material. This represents the rate of temperature change.

[0088] Specifically, a dilated convolutional layer can be represented as:

[0089] ;

[0090] In the formula, Indicates the characteristics of thermal time series, Represents the ReLU activation function. These represent the kernel matrix and bias vector of the dilated convolutional layer, respectively. This represents the thermodynamic flow input vector.

[0091] Understandably, dilated convolution layers can better capture long-range heat conduction.

[0092] Specifically, the thermal delay compensation layer can be represented as:

[0093] ;

[0094] In the formula, These represent the thermodynamic time series characteristics of the current time t and the historical thermodynamic time series characteristics of t-1, respectively. Indicates thermodynamic characteristics, This represents the thermal inertia compensation factor, which is preferably 0.5.

[0095] Therefore, the thermal delay compensation layer can compensate for the delay in heat propagation.

[0096] Furthermore, the loaded head resonance suppression sub-model includes a differential feature extraction layer and a resonance suppression operation layer, and the resonance suppression unit includes:

[0097] The differential enhancement extraction subunit is used to reduce the fixed fluctuations of the loading head motion input vector through differential operation to generate loading head temporal features;

[0098] The resonance suppression subunit is used to pass the mechanical timing characteristics through the resonance suppression operation layer to generate the loading head characteristics.

[0099] In particular, the resonance suppression operation layer can specifically identify the mechanical resonance frequency generated by the multi-degree-of-freedom loading head in the high-frequency motion of multi-directional loading and unloading, and actively cancel the resonance energy through the amplitude attenuation algorithm to avoid load output distortion caused by resonance, and ensure that the actual applied value of the mechanical load parameter is consistent with the set value.

[0100] Specifically, the loading head motion input vector, in addition to the basic input vector, also includes at least the following: Where D represents the location of the loading header, This represents the loading head position setting value, V represents the loading head velocity, and F represents the force sequence. The value of k represents the cooling rate of the cable material, and k represents the stiffness of the cable material. This represents the force-displacement sequence of the loading head. This indicates the acceleration of the loading head.

[0101] Specifically, the difference operation can be expressed as:

[0102] ;

[0103] In the formula, Indicates primary features, Represents the ReLU activation function. This represents the input vector of the loaded head motion. These represent the convolution weight matrix and the bias, respectively. Represents the first-order difference. This indicates that the difference between adjacent feature points is calculated along the time dimension for the primary feature. This represents the calculation of the difference between adjacent feature points along the time dimension using the first-order difference, as described above in the form of the loading head motion input vector. This indicates that for the time dimension t2-t1, t3-t2, ..., This represents a second-order difference.

[0104] Specifically, the resonance suppression operational layer can be represented as:

[0105] ;

[0106] In the formula, This indicates that the Fast Fourier Transform (RFT) transforms the second-order difference... The time-domain signal is converted into a frequency-domain representation. This represents the magnitude M of the frequency domain signal. This indicates the maximum value point in the amplitude spectrum, corresponding to the main resonant frequency of the signal. , This represents the dot product operation. Represents the Dirac function, This represents the attenuation coefficient, preferably 0.85. This represents the suppressed spectrum. This represents the Inverse Fourier Transform (IRFT), which restores the suppressed spectrum to the loaded header features of the time-domain signal. .

[0107] Specifically, the gated fusion layer can be represented as:

[0108] ;

[0109] In the formula, Indicates the gating weight, This represents a one-dimensional convolution operation. Indicates thermodynamic characteristics, Indicates the characteristics of the loading header. This represents the Sigmoid activation function. It represents the comprehensive physical characteristics.

[0110] Further, the convolutional pooling layer includes a first convolutional pooling layer, a second convolutional pooling layer, and a flattening layer, and the kernel size of the first convolutional pooling layer is smaller than the kernel size of the second convolutional pooling layer. The convolutional unit includes:

[0111] The first convolutional pooling subunit is used to pass the basic input vector through the first convolutional pooling layer to generate local temporal features;

[0112] The second convolutional pooling subunit is used to pass the local temporal features through the second convolutional pooling layer to generate global temporal features;

[0113] The flattening layer subunit is used to pass the global features through the flattening layer to generate the comprehensive temporal features.

[0114] In particular, considering the multi-timescale characteristics of load and temperature parameters in cable testing, such as millisecond-level load impact and minute-level temperature conduction, small kernel convolution provides fast response to high-frequency dynamics, while large kernel convolution captures slow changing trends. The synergy of the two enables the model to maintain high-precision feature extraction capabilities under both transient and steady-state conditions, providing comprehensive temporal basis for the generation of control parameters.

[0115] Specifically, the first convolutional pooling layer uses 32 convolutional kernels, a convolutional size of 5, a stride of 1, a ReLU activation function, and a MaxPooling pooling size of 2. The second convolutional layer uses 64 convolutional kernels, a convolutional size of 3, a stride of 1, a ReLU activation function, and a MaxPooling pooling size of 2. The flattening layer is used to convert the aforementioned multi-dimensional global temporal features into a one-dimensional vector to adapt to subsequent fully connected layer processing. The fully connected layer has 128 nodes and performs feature mapping on the concatenated features of the integrated temporal features and the integrated physical features.

[0116] Furthermore, the control parameters include temperature control parameters and force control parameters, and the constraint unit includes:

[0117] Thermodynamic constraint subunit, used to thermodynamically constrain temperature control parameters based on the temperature parameters and the maximum thermal energy of the heater;

[0118] The loading head constraint subunit is used to constrain the force control parameters based on the cable material stiffness and the maximum stress of the loading head.

[0119] Specifically, the thermodynamic constraint can be expressed as:

[0120] ;

[0121] In the formula, Let i and d represent the temperature T control parameters of the temperature PID controller, respectively. Let i and d represent the temperature control parameters T and d respectively in the original control parameters. This represents the maximum thermal energy of the heater, where Indicates the maximum allowable temperature of the heater. Indicates temperature parameter, This indicates the maximum available heat dissipation power of the test platform. Indicates the time step of the control cycle. Minimum effective threshold.

[0122] Specifically, force constraints can be expressed as:

[0123] ;

[0124] In the formula, Let p represent the displacement D control parameter of the loading head displacement controller and d represent the force F control parameter of the loading head force controller, respectively. This represents the p control parameter for force-displacement D control and the d control parameter for force-force F magnitude control in the original control parameters. This indicates the maximum force exerted by the loading head. This indicates the maximum displacement of the loading head, and k represents the stiffness of the cable material. The minimum speed threshold required by the current operating conditions.

[0125] More specifically, the control module determines the execution temperature as follows: It controls the infrared heating plate of the heater using `heater_power=min(0.6*P+0.4*I,MAX_HEATER_POWER)`, controls the thermoelectric cooler of the heater using `tec_power=min(0.4*abs(P)+0.6*abs(I),MAX_TEC_POWER)`, and controls the convection fan of the heater using `fan_speed=int(MAX_FAN_SPEED*(0.3+0.7*abs(D) / MAX_DERIVATIVE))`, where `MAX_HEATER_POWER` represents the maximum power of the infrared heating plate, `MAX_TEC_POWER` represents the maximum power of the thermoelectric cooler, `MAX_FAN_SPEED` represents the maximum speed of the convection fan, and `MAX_DERIVATIVE` represents the maximum acceleration of the convection fan. ,in Here, e_T represents the set value, and e_T is the error between the set operating temperature and the actual temperature. The i-th and d-th control parameters of the temperature PID controller are represented by , respectively, and integral and derivative are the integral and derivative of the error, respectively. The control of the multi-degree-of-freedom loading head is determined similarly or based on existing PID universal joint attitude calculation algorithms and PID hydraulic cylinder force control algorithms, and will not be elaborated further here.

[0126] In particular, the thermodynamic compensation mechanism and the loading head resonance suppression sub-model are specifically designed to learn the temperature field hysteresis effect and the mechanical system resonance interference, respectively. The gated fusion layer achieves adaptive coupling of physical field characteristics through dynamic weight allocation. This hierarchical learning architecture enables the BP neural network to accurately fit the strong nonlinear relationship under extreme scenarios, reducing the adjustment error of control parameters, so as to achieve stable and accurate parameter control of the multi-degree-of-freedom loading head and heater of the test platform under complex extreme conditions.

[0127] Specifically, a fully connected layer can be represented as:

[0128] ;

[0129] In the formula, y represents the original control parameters, and ReLU represents the ReLU activation function. represents the learnable convolutional matrix and bias vector of the fully connected convolutional layer, respectively, and x represents the concatenated vector combining the temporal features and the combined physical features.

[0130] like Figure 4 As shown, the neural network backpropagation submodule further includes:

[0131] The loss function construction unit is used to construct the test stationary loss function based on the control parameter error term and the control parameter change rate.

[0132] The backpropagation unit is used to perform backpropagation adjustments on the control parameter fitting model in the directions of control parameters, convolution, thermodynamics, and resonance suppression based on the test stationary loss function.

[0133] In particular, an ordered backpropagation path consisting of control parameters, convolution, thermodynamics, and resonance suppression is adopted to achieve directional gradient transfer from the output layer to the physical field submodule: the control parameter direction prioritizes the correction of the mapping weights of the fully connected layer to ensure that the deviation between the output parameters and the set values ​​converges quickly; the convolution direction specifically optimizes the convolution kernel parameters for extracting temporal features to improve the modeling accuracy of dynamic load temperature sequences; the thermodynamic and resonance suppression directions respectively adjust the internal parameters of the compensation mechanism and the suppression model to avoid gradient interference across physical fields, making their optimization more targeted.

[0134] Specifically, the test stationary loss function can be expressed as:

[0135] ;

[0136] In the formula, This indicates the test stationary loss function. These represent the two weighting coefficients. This represents the time-weighted absolute error of the control parameters. This represents the control parameter error term for the first three cycles. The control parameter change rate term represents the comprehensive fluctuation amplitude of the control quantity that minimizes the L2 norm for the heater power change, loading head displacement rate, and loading head force change rate.

[0137] Specifically, the direction of the control parameter is as follows: The convolution direction is .

[0138] Specifically, the process of adjusting the thermodynamic compensation mechanism through backpropagation can be expressed as follows: The process of adjusting the loading head resonance suppressor sub-model through backpropagation can be expressed as follows: , This represents mechanical flow gradient filtering. This represents the weight matrix of the loaded head resonance suppression sub-model. This represents the cutoff frequency (typically 100Hz), used to suppress high-frequency noise. The above formula represents the process of solving for the gradient value of the loss function.

[0139] In particular, by providing real-time feedback on the load parameter sequence, the network weights are dynamically adjusted along gradient paths in the control parameter direction, convolution direction, thermodynamic direction, and resonance suppression direction. This improves the parameter convergence speed under extreme conditions compared to traditional unidirectional gradient descent.

[0140] In this embodiment, the cable testing module achieves the coordinated application of temperature and multi-degree-of-freedom mechanical loads, simulating the complex and extreme thermodynamic conditions faced by the test cable in actual operation. Through a control parameter fitting model based on a BP neural network architecture, the nonlinear relationship between temperature and load parameters can be accurately captured. This allows the multi-functional testing platform to learn control parameters corresponding to complex extreme conditions autonomously, without relying on manually preset models, thus improving the platform's adaptability to complex testing conditions and achieving stable and precise parameter control of the platform's multi-degree-of-freedom loading head and heater. Thermodynamic compensation mechanisms and a loading head resonance suppression sub-model are used to specifically learn the temperature field hysteresis effect and mechanical system resonance interference, respectively. The gated fusion layer achieves adaptive coupling of physical field characteristics through dynamic weight allocation. This hierarchical learning architecture enables the BP neural network to accurately fit the strong nonlinear relationship under extreme scenarios, reducing the adjustment error of control parameters and achieving stable and precise parameter control of the test platform's multi-degree-of-freedom loading head and heater for complex extreme conditions. Through real-time feedback of the load parameter sequence, the network weights are dynamically corrected along the gradient paths of the control parameter direction, convolution direction, thermodynamic direction, and resonance suppression direction. Compared to traditional unidirectional gradient descent, it improves the convergence speed of parameters under extreme conditions.

[0141] Those skilled in the art will recognize that the modules and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0142] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0143] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A multifunctional integrated cable simulation working condition test platform, characterized in that, The method comprises the following steps: a cable test module is used to collect temperature parameters and load parameters of a test cable tested by a heater and a multi-degree-of-freedom loading head; a neural network module is used to determine control parameters by fitting the temperature parameters and the load parameters through a control parameter fitting model based on a BP neural network architecture, and to perform back propagation adjustment of model parameters of the control parameter fitting model through a test stationary loss function; a control module is used to generate an execution temperature of the heater and an execution force of the multi-degree-of-freedom loading head through the control parameters by a control algorithm; the neural network module comprises: a neural network forward propagation submodule is used to generate the control parameters by fitting the temperature parameters and the load parameters through a control parameter fitting model based on a BP neural network architecture; a neural network back propagation submodule is used to perform back propagation adjustment of model parameters of the control parameter fitting model through a test stationary loss function; the load parameters comprise a force displacement sequence and a force parameter sequence, the control parameter fitting model comprises a thermodynamic compensation mechanism, a loading head resonance suppression submodule, a gating fusion layer, a convolution pooling layer, a full connection layer and a constraint layer, and the neural network forward propagation submodule comprises: an input integration unit is used to integrate the temperature parameters, the force displacement sequence, the force parameter sequence, a temperature set value, a displacement set value and a force set value into channels of a basic input vector respectively, to generate a thermodynamic input vector based on the basic input vector and a cable material thermal conductivity coefficient, and to generate a loading head motion input vector based on the basic input vector and a cable material stiffness; a thermodynamic processing unit is used to generate thermodynamic features by passing the thermodynamic input vector through a thermodynamic compensation mechanism; a resonance suppression unit is used to generate loading head features by passing the loading head motion input vector through a loading head resonance suppression submodule; a gating fusion unit is used to generate comprehensive physical features by passing the thermodynamic features and the loading head features through a gating fusion layer; a convolution unit is used to generate comprehensive time sequence features by passing the basic input vector through a convolution pooling layer; a mapping unit is used to generate original control parameters by performing feature mapping of the comprehensive time sequence features and the comprehensive physical features through a full connection layer; a constraint unit is used to perform output constraint of the original control parameters based on the cable material thermal conductivity coefficient and the cable material stiffness through a constraint layer to generate the control parameters.

2. The multifunction integrated cable analog working condition test platform according to claim 1, characterized in that, the thermodynamic compensation mechanism comprises an expansion convolution layer and a thermal delay compensation layer, and the thermodynamic processing unit comprises: an expansion convolution subunit is used to perform time sequence extraction of the thermodynamic input vector through an expansion convolution layer to generate thermal time sequence features; a thermal delay compensation subunit is used to generate the thermodynamic features by passing current thermal time sequence features and historical thermal time sequence features through a thermal delay compensation layer.

3. The multifunction integrated cable analog working condition test platform according to claim 1, characterized in that, the loading head resonance suppression submodule comprises a differential feature extraction layer and a resonance suppression operation layer, and the resonance suppression unit comprises: a differential enhancement extraction subunit configured to reduce fixed fluctuations of the loading head motion input vector through a differential operation to generate a loading head timing feature; a resonance suppression subunit configured to generate the loading head feature through a resonance suppression operation layer of the loading head timing feature.

4. The multifunctional integrated cable analog working condition test platform according to claim 1, characterized in that, The convolution pooling layer includes a first convolution pooling layer, a second convolution pooling layer, and a flattening layer, and a convolution kernel size of the first convolution pooling layer is smaller than a convolution kernel size of the second convolution pooling layer, and the convolution unit includes: a first convolution pooling subunit configured to generate a local timing feature through a first convolution pooling layer of the basic input vector; a second convolution pooling subunit configured to generate a global timing feature through a second convolution pooling layer of the local timing feature; a flattening layer subunit configured to generate the comprehensive timing feature through a flattening layer of the global timing feature.

5. The multifunctional integrated cable analog working condition test platform according to claim 1, characterized in that, The control parameters include temperature control parameters and force control parameters, and the constraint unit includes: a thermodynamic constraint subunit configured to perform thermodynamic constraint on the temperature control parameters based on the temperature parameters and a maximum thermal energy of the heater; a loading head constraint subunit configured to perform force constraint on the force control parameters based on the cable material stiffness and a maximum stress of the loading head.

6. The multifunctional integrated cable analog working condition test platform according to claim 1, characterized in that, The neural network back propagation sub-module includes: a loss function construction unit configured to construct the test stationary loss function based on a control parameter error term and a control parameter change rate.

7. The multifunction integrated cable analog operating condition test platform according to claim 1, characterized in that, The neural network back propagation sub-module further includes: a back propagation unit configured to sequentially perform back propagation adjustment of the control parameter fitting model in a control parameter direction, a convolution direction, a thermodynamic direction, and a resonance suppression direction based on the test stationary loss function.

8. The multifunction integrated cable analog working condition test platform according to any one of claims 1 to 7, characterized in that, The heater and the multi-degree-of-freedom loading head are integrated in the environmental simulation cabin. The multi-degree-of-freedom loading head includes a hydraulic cylinder array and a gimbal controlling the hydraulic cylinder array. The heater at least includes a semiconductor refrigeration sheet, an infrared heating plate, and a convection fan.

Images