Power supply adaptive regulation and control system and method based on multi-modal data fusion

Abstract

The invention relates to the technical field of power supply automation control, in particular to a power supply self-adaptive regulation and control system and method based on multi-modal data fusion. The method comprises the following steps: acquiring internal mechanical structure data of a power supply, and recording running state data of the power supply; stress response characteristics of the internal mechanical structure data of the power supply under different power supply operation state data are analyzed, and a mechanical support structure response model is established; power supply electrical parameters in the power supply operation process are collected; performing parameter correlation detection on the power supply electrical parameters to obtain electrical parameter correlation data; and performing electrical vibration influence mapping on the mechanical support structure response model through the electrical parameter associated data to generate structure vibration mapping data. Through a multi-modal data fusion technology, comprehensive monitoring and accurate regulation and control of the power supply operation state are realized, so that the stability and the reliability of a power supply system are improved.

Description

Technical Field

[0001] This invention relates to the field of power supply automation control technology, and in particular to a power supply adaptive regulation system and method based on multimodal data fusion. Background Technology

[0002] The mechanical support structure of internal power supply components needs to have sufficient strength and stability to prevent damage under high-frequency vibration or impact. However, the mechanical support structure design of traditional power supplies is based only on the strength requirements under static conditions and does not fully consider the dynamic stress distribution under high-frequency vibration or impact environments. When faced with complex external vibration conditions, the mechanical support structure is prone to damage, which in turn affects the normal operation of the power supply. Furthermore, power supply regulation systems often rely on only a single type of data (such as voltage or current) for regulation and lack the ability to comprehensively process multi-source heterogeneous data (such as temperature, heat dissipation efficiency, GPU / CPU load, AI computing pressure, etc.). In addition, traditional power supply adaptive regulation technology mostly relies on manual settings and fixed rules, lacking intelligent adaptive capabilities. This technology is difficult to achieve automatic optimization and intelligent regulation when faced with complex operating environments and dynamic load changes. Summary of the Invention

[0003] Therefore, it is necessary to provide a power adaptive control system and method based on multimodal data fusion to solve at least one of the above-mentioned technical problems.

[0004] To achieve the above objectives, a power supply adaptive regulation method based on multimodal data fusion is provided, the method comprising the following steps: Step S1: Obtain data on the internal mechanical structure of the power supply and record the power supply operating status data; analyze the stress response characteristics of the internal mechanical structure data of the power supply under different power supply operating status data, and establish a mechanical support structure response model; Step S2: Collect the power supply electrical parameters during the power supply operation; perform parameter correlation detection on the power supply electrical parameters to obtain electrical parameter correlation data; use the electrical parameter correlation data to perform electrical vibration influence mapping on the mechanical support structure response model to generate structural vibration mapping data; Step S3: Collect power load data during power supply operation; mark the power load data as a time series and monitor fluctuating load to obtain fluctuating load sequence data; use the fluctuating load sequence data to perform power ripple response mapping on the mechanical support structure response model to generate ripple response mapping data. Step S4: Collect heat accumulation data during power supply operation, identify and record the concentrated heat accumulation area of ​​the power supply through the heat accumulation data; perform heat dissipation response mapping on the mechanical support structure response model based on the concentrated heat accumulation area of ​​the power supply, and generate heat dissipation response mapping data. Step S5: Perform multimodal data fusion based on structural vibration mapping data, ripple response mapping data, and heat dissipation response mapping data, and determine the power supply adaptive control strategy; dynamically adjust and control the power supply according to the power supply adaptive control strategy.

[0005] Preferably, step S2, which involves parametric correlation detection of the power supply electrical parameters, includes: Power supply electrical parameters are classified into voltage, current, and frequency according to parameter type; The voltage, current, and frequency are divided into multiple equal-length time segments in chronological order. Within each time period, the changing trends of voltage and current are analyzed synchronously to obtain voltage-current parameter correlation data; if both voltage and current rise or fall within the same time period, it is marked as a change in the same direction; if voltage rises while current falls, or voltage falls while current rises, it is marked as a change in opposite directions. Within each time period, the changing trends of frequency and current are analyzed synchronously to obtain frequency-current parameter correlation data; if both frequency and current rise or fall within the same time period, it is marked as a change in the same direction; if frequency rises while current falls, or frequency falls while current rises, it is marked as a change in opposite directions. Electrical parameter correlation data is determined based on pressure-current parameter correlation data and frequency-current parameter correlation data.

[0006] Preferably, step S2, which maps the electrical vibration influence on the mechanical support structure response model using electrical parameter correlation data, includes: Identify the input-end fixed support and the output-end insulating support in the mechanical support structure response model; The pressure-flow fluctuation in the pressure-flow parameter correlation data is evaluated to obtain the pressure-flow fluctuation degree value; Calculate the pressure flow vibration intensity factor for the fixed support at the input end based on the pressure flow fluctuation value; The pressure flow vibration intensity factor is mapped to the mechanical support structure response model, the degree of influence of the pressure flow factor is recorded, and the input vibration mapping data is generated. Frequency flow changes are detected in the frequency flow parameter correlation data to obtain the degree of frequency flow change value; The frequency current vibration intensity factor is calculated for the output end insulating bracket based on the frequency current variation value. The frequency-current vibration intensity factor is mapped to the mechanical support structure response model, the degree of influence of the frequency-current factor is recorded, and the output vibration mapping data is generated. The vibration mapping data at the input and output ends is mapped to the mechanical support structure response model, and the electrical vibration characteristics are detected to generate structural vibration mapping data.

[0007] Preferably, step S3, which involves time-series labeling of the power load data and monitoring of fluctuating loads, includes: Based on the start time of power supply operation, the power load data is divided into 1-millisecond time intervals, with each time interval corresponding to a power load data point. Each power load data point is assigned a time stamp, which is in milliseconds, to record the specific position of the data point in the time series. Analyze each power load data point in the time series one by one, and calculate the power difference between adjacent power load data points; When the power difference exceeds 5% of the rated power of the power supply, the power load data point is identified as a power load fluctuation point, and the power value and corresponding time identifier of the power load data point are recorded. All identified power load fluctuation points are arranged in order of time to form fluctuating load sequence data; in the fluctuating load sequence data, the power value, time mark and power difference of each fluctuation point are recorded.

[0008] Preferably, step S3, which maps the power ripple response of the mechanical support structure response model using fluctuating load sequence data, includes: The switching contact points of the mechanical support structure response model are identified by using fluctuating load sequence data, and their response characteristics under power fluctuations are detected; when load fluctuations occur in the fluctuating load sequence data, the contact resistance value of the switching contact points is recorded. For each load fluctuation point in the fluctuating load sequence data, if the contact resistance value shows a linear increasing trend when the load fluctuation point appears, then the relationship between the load fluctuation point and the switch contact point is determined to be a response synchronization relationship. The load fluctuation points are mapped to the switch contact area of ​​the mechanical support structure response model, and the power change amplitude, change frequency and duration of the switch contact area are recorded. Based on the power change amplitude and duration of the fluctuation point, the ripple response intensity of the switch contact point is classified and ripple response mapping data is generated. When the power change exceeds 15% of the rated power and lasts for more than 10 milliseconds, it is marked as high-intensity ripple response data; When the power change is between 10% and 15% of the rated power and the duration is between 5 and 10 milliseconds, it is marked as medium intensity ripple response data. Data with a power change of less than 10% of the rated power and a duration of less than 5 milliseconds is classified as low-intensity ripple response data.

[0009] Preferably, in step S4, the heat accumulation data collected during the power supply operation, and the heat accumulation concentration areas of the power supply determined and recorded through the heat accumulation data, include: Temperature sensors are installed at the locations of the heat sink, power module, capacitors, and transformer inside the power supply; each temperature sensor collects temperature data at fixed time intervals of 1 second. For each sensor location, the difference between the current power supply operating temperature and the previous power supply operating temperature is used to obtain the power supply operating temperature difference. Identify areas of sustained temperature rise in the power supply operating temperature difference and mark them as potential areas of concentrated heat accumulation; If the temperature difference of the power supply operating temperature in the potential heat accumulation zone exceeds the preset temperature rise value, where the preset temperature rise value is 2 degrees Celsius per second, it is determined to be a heat accumulation zone of the power supply.

[0010] Preferably, step S4, which maps the heat dissipation response of the mechanical support structure response model based on the concentrated heat accumulation area of ​​the power source, includes: Measure the temperature change rate and peak temperature of each heat accumulation zone, and classify the heat accumulation zones into heat dissipation demand levels based on the temperature change rate and peak temperature of each zone: If the rate of temperature change exceeds 3 degrees Celsius per second or the peak temperature exceeds 80 degrees Celsius, it is classified as a high heat dissipation requirement area: If the temperature change rate is between 1 and 3 degrees Celsius per second or the peak temperature is between 60 and 80 degrees Celsius, it is classified as a medium heat dissipation requirement zone. If the rate of temperature change is less than 1 degree Celsius per second or the peak temperature is below 60 degrees Celsius, it is classified as a low heat dissipation requirement zone. Identify heat sinks, thermal pads, ventilation holes, and heat dissipation channels in the mechanical support structure response model to obtain the heat dissipation hardware structure characteristics; The heat dissipation demand regions of high heat dissipation demand region, medium heat dissipation demand region, and low heat dissipation demand region are respectively matched with the heat dissipation hardware structure features to generate heat dissipation response mapping data.

[0011] Preferably, step S5 involves multimodal data fusion based on structural vibration mapping data, ripple response mapping data, and heat dissipation response mapping data, and determining the power supply adaptive control strategy, including: Frequency domain analysis was performed on the structural vibration mapping data to extract vibration frequency components and identify the primary and secondary vibration frequencies. Based on the primary and secondary vibration frequencies, the structural layout vibration impact was assessed on the mechanical support structure response model to obtain structural layout vibration impact data. Time-domain analysis was performed on the ripple response mapping data to extract the ripple amplitude and ripple frequency; based on the ripple amplitude and ripple frequency, the component ripple influence was evaluated on the mechanical support structure response model to obtain component ripple influence data; Heat flow analysis is performed on the heat dissipation response mapping data to extract the temperature change rate and heat dissipation demand level; the heat dissipation structure impact assessment is performed on the mechanical support structure response model based on the temperature change rate and heat dissipation demand level to obtain heat dissipation structure impact data. The vibration impact data of structural layout, the ripple impact data of components, and the impact data of heat dissipation structure are multimodally weighted and fused according to a weight of 4:3:3 to obtain multimodal fused data; Calculate the fusion index of the multimodal fusion data. If the fusion index is less than 30, it indicates that the power supply is in good condition and a low-level power supply control strategy is determined. If the fusion index is between 30 and 70, it indicates that the power supply is in a moderate condition and a medium-level power supply control strategy is determined. If the fusion index is greater than 70, it indicates that the power supply is in a poor condition and a high-level power supply control strategy is determined.

[0012] Preferably, step S5, which involves dynamically adjusting and controlling the power supply according to the power supply adaptive control strategy, includes: The power supply is dynamically adjusted and controlled according to the low-level power regulation strategy, so that the output power of the power module is maintained at 90% of the rated power, the cooling fan is running at 40% of the rated speed, the heat sink heat dissipation area is maintained at 60% of the rated area, and the load distributor maintains the load distribution ratio as even. The power supply is dynamically adjusted and controlled according to the power supply level regulation strategy. The output power of the power module is gradually reduced to 70% of the rated power, with each adjustment step being 5% of the rated power and an adjustment interval of 2 seconds. The speed of the cooling fan is gradually increased to 60% of the rated speed, with each adjustment step being 10% of the rated speed and an adjustment interval of 3 seconds. The heat dissipation area of ​​the heat sink is gradually increased to 80% of the rated area, with each adjustment step being 10% of the rated area and an adjustment interval of 4 seconds. The load distributor dynamically adjusts the load distribution ratio based on real-time monitoring data, prioritizing the allocation to modules with lower loads, with each adjustment step being 5% of the total load and an adjustment interval of 5 seconds. According to the power supply low-level regulation strategy, the power supply is dynamically adjusted and controlled. The output power of the power module is immediately reduced to 50% of the rated power, with each adjustment step being 10% of the rated power and an adjustment interval of 1 second; the speed of the cooling fan is immediately increased to 100% of the rated speed, with each adjustment step being 20% ​​of the rated speed and an adjustment interval of 2 seconds; the heat dissipation area of ​​the heat sink is immediately increased to 100% of the rated area, with each adjustment step being 20% ​​of the rated area and an adjustment interval of 3 seconds; the load distributor urgently adjusts the load distribution ratio, giving priority to the backup module, with each adjustment step being 10% of the total load and an adjustment interval of 2 seconds.

[0013] This invention acquires data on the internal mechanical structure of a power supply and records its operating status, enabling precise understanding of the power supply's mechanical structural characteristics under different operating conditions. Analyzing the stress response characteristics of the internal mechanical structure data under different operating conditions allows for the establishment of a mechanical support structure response model, providing a crucial foundation for subsequent comprehensive evaluation of power supply performance. This model accurately reflects the stress variation patterns of the mechanical structure during operation, thereby achieving a quantitative assessment of the power supply's mechanical stability. This provides a reliable basis for formulating subsequent control strategies, ensuring the stability and reliability of the mechanical structure during long-term operation and preventing power supply failures due to mechanical structure malfunctions. Collecting power supply electrical parameters during operation and performing parameter correlation detection yields electrical parameter correlation data, providing a comprehensive understanding of the intrinsic relationships within the power supply's electrical performance. Mapping the electrical vibration influence of the mechanical support structure response model using the electrical parameter correlation data generates structural vibration mapping data, quantifying the coupling relationship between electrical performance and mechanical structure vibration. This process accurately identifies the impact of electrical parameter changes on mechanical structure vibration, enabling the power supply to promptly detect potential vibration problems caused by fluctuations in electrical parameters during operation. This provides precise vibration control direction for subsequent regulation strategies, effectively reducing the negative impact of vibration on power supply performance and lifespan, and improving the stability and reliability of power supply operation. By collecting power load data during power supply operation and performing time-series labeling and fluctuating load monitoring, fluctuating load sequence data is obtained, allowing real-time monitoring of dynamic changes in power supply load. Power ripple response mapping is performed on the mechanical support structure response model using fluctuating load sequence data, generating ripple response mapping data and realizing a direct correlation between power load fluctuations and mechanical structure response. This allows the power supply system to accurately identify the ripple impact of power load fluctuations on the mechanical structure, thereby taking preventative control measures to optimize the power supply's operation under different load conditions, effectively suppressing the adverse effects of power ripple on the mechanical structure, improving the power supply's adaptability and stability under complex load conditions, and extending its lifespan. Collecting heat accumulation data during power supply operation and identifying and recording concentrated heat accumulation areas allows for accurate identification of key areas of internal heat distribution within the power supply. By mapping the heat dissipation response of the mechanical support structure to the heat accumulation concentration area of ​​the power supply, heat dissipation response mapping data is generated, achieving an organic combination of the power supply's heat dissipation characteristics and the mechanical structure response. This process can accurately assess the impact of heat accumulation on the stability of the mechanical structure, providing a scientific basis for heat dissipation design and control. By optimizing the heat dissipation strategy, the temperature stability of the mechanical structure of the power supply is ensured during high-load operation, avoiding mechanical performance degradation and failure risks caused by local overheating, ensuring efficient and stable operation of the power supply during long-term operation, and improving the overall performance and reliability of the power supply.Multimodal data fusion based on structural vibration mapping data, ripple response mapping data, and heat dissipation response mapping data enables a comprehensive assessment of power supply performance by considering multiple key factors during operation. An adaptive control strategy is determined, and the power supply is dynamically adjusted and controlled according to this strategy, allowing it to automatically optimize its performance under different operating conditions. This combination of multimodal data fusion and adaptive control effectively enhances the power supply's intelligence level, ensuring it maintains optimal operating conditions under complex circumstances, improving operating efficiency and reliability, reducing maintenance costs, and enhancing its adaptability in practical applications.

[0014] This specification also provides a power adaptive control system based on multimodal data fusion, used to execute the power adaptive control method based on multimodal data fusion as described above. The power adaptive control system based on multimodal data fusion includes: The mechanical support structure response model construction module is used to acquire internal mechanical structure data of the power supply, record power supply operating status data, analyze the stress response characteristics of the internal mechanical structure data of the power supply under different power supply operating status data, and establish a mechanical support structure response model. The electrical vibration impact mapping module is used to collect the power supply electrical parameters during the power supply operation process; perform parameter correlation detection on the power supply electrical parameters to obtain electrical parameter correlation data; and perform electrical vibration impact mapping on the mechanical support structure response model through the electrical parameter correlation data to generate structural vibration mapping data. The power ripple response mapping module is used to collect power load data during power supply operation; to perform time series labeling on the power load data and to monitor fluctuating loads to obtain fluctuating load sequence data; and to perform power ripple response mapping on the mechanical support structure response model using the fluctuating load sequence data to generate ripple response mapping data. The heat dissipation response mapping module is used to collect heat accumulation data during power supply operation, identify and record the concentrated heat accumulation area of ​​the power supply through the heat accumulation data, and perform heat dissipation response mapping on the mechanical support structure response model based on the concentrated heat accumulation area of ​​the power supply to generate heat dissipation response mapping data. The power adaptive control module is used to perform multimodal data fusion based on structural vibration mapping data, ripple response mapping data, and heat dissipation response mapping data, and to determine the power adaptive control strategy; and to dynamically adjust and control the power supply according to the power adaptive control strategy.

[0015] This invention achieves comprehensive monitoring and precise control of power supply operation status through the collaborative work of multiple modules. The mechanical support structure response model construction module accurately acquires internal mechanical structure data and operating status data of the power supply, analyzes stress response characteristics, and establishes a model, providing a foundation for subsequent control. The electrical vibration impact mapping module collects power supply electrical parameters and performs correlation detection to generate structural vibration mapping data, realizing a direct correlation between electrical performance and mechanical vibration. The power ripple response mapping module generates ripple response mapping data by time-series labeling and fluctuation monitoring of power load data, accurately identifying the impact of load fluctuations on the mechanical structure. The heat dissipation response mapping module collects heat accumulation data, identifies heat accumulation concentration areas, generates heat dissipation response mapping data, and optimizes heat dissipation strategies. The power supply adaptive control module, based on the fusion of the above multimodal data, determines the control strategy and dynamically adjusts the power supply operation status. This system, through multimodal data fusion and adaptive control, can monitor multiple key factors in power supply operation in real time, accurately identify potential problems and dynamically optimize power supply performance, significantly improve the stability, reliability and operating efficiency of the power supply under complex operating conditions, extend the service life of the power supply, reduce maintenance costs, and enhance the adaptability of the power supply in practical application scenarios, providing a strong guarantee for the intelligent and efficient operation of power supply systems. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the steps of a power supply adaptive control method based on multimodal data fusion. Figure 2 for Figure 1 A detailed flowchart illustrating the implementation steps of step S3. The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0017] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0018] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0019] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0020] To achieve the above objectives, please refer to Figures 1 to 2 A power supply adaptive regulation method based on multimodal data fusion, the method comprising the following steps: Step S1: Obtain data on the internal mechanical structure of the power supply and record the power supply operating status data; analyze the stress response characteristics of the internal mechanical structure data of the power supply under different power supply operating status data, and establish a mechanical support structure response model; Step S2: Collect the power supply electrical parameters during the power supply operation; perform parameter correlation detection on the power supply electrical parameters to obtain electrical parameter correlation data; use the electrical parameter correlation data to perform electrical vibration influence mapping on the mechanical support structure response model to generate structural vibration mapping data; Step S3: Collect power load data during power supply operation; mark the power load data as a time series and monitor fluctuating load to obtain fluctuating load sequence data; use the fluctuating load sequence data to perform power ripple response mapping on the mechanical support structure response model to generate ripple response mapping data. Step S4: Collect heat accumulation data during power supply operation, identify and record the concentrated heat accumulation area of ​​the power supply through the heat accumulation data; perform heat dissipation response mapping on the mechanical support structure response model based on the concentrated heat accumulation area of ​​the power supply, and generate heat dissipation response mapping data. Step S5: Perform multimodal data fusion based on structural vibration mapping data, ripple response mapping data, and heat dissipation response mapping data, and determine the power supply adaptive control strategy; dynamically adjust and control the power supply according to the power supply adaptive control strategy.

[0021] In this embodiment of the invention, reference is made to Figure 1 The diagram shown is a flowchart illustrating the steps of a power adaptive regulation method based on multimodal data fusion according to the present invention. In this example, the power adaptive regulation method based on multimodal data fusion includes the following steps: Step S1: Obtain data on the internal mechanical structure of the power supply and record the power supply operating status data; analyze the stress response characteristics of the internal mechanical structure data of the power supply under different power supply operating status data, and establish a mechanical support structure response model; In this embodiment of the invention, during step S1, a high-precision 3D laser scanner is first used to perform a comprehensive scan of the internal mechanical structure of the power supply, acquiring its 3D geometric data. This includes the dimensions, shape, and spatial location information of key components such as the power supply casing, support frame, and cooling fan mounting bracket. The data is stored in point cloud format with a point cloud density of no less than 1000 points per square centimeter to ensure accurate capture of mechanical structural details. Simultaneously, multiple sensors embedded within the power supply collect operational status data. A temperature sensor collects the real-time temperature of key components of the power supply at a sampling frequency of 10 times per second; a voltage sensor monitors the power supply output voltage with a sampling accuracy of 0.01 volts; and a current sensor measures the power supply output current with a sampling accuracy of 0.01 amperes. This operational status data is transmitted to the data processing center via a wired communication interface with a millisecond-level latency. When analyzing the correlation between the internal mechanical structure data and operating status data of the power supply, the finite element analysis software ANSYS was used to model the mechanical structure. The point cloud data acquired through scanning was imported into the software, and the internal mechanical structure of the power supply was divided into several micro-elements using mesh generation technology. The element type adopted was an eight-node hexahedral element, with an element size not exceeding 1 mm to ensure analysis accuracy. For different power supply operating states, such as normal operation, overload operation, and startup, the stress distribution of the mechanical structure under corresponding operating conditions was calculated based on the temperature, voltage, and current data collected by sensors. For example, under normal operating conditions, the internal temperature of the power supply is 45 degrees Celsius, the output voltage is 12 volts, and the current is 5 amps. At this time, ANSYS software calculations show that the stress concentration area of ​​the power supply cooling fan mounting bracket is located at the connection between the bracket and the fan, with a maximum stress value of 150 MPa. Under overload operation conditions, the output current reaches 10 amps, and the maximum stress value of the cooling fan mounting bracket rises to 250 MPa, while the stress concentration area remains unchanged. Based on the above analysis results, a mechanical support structure response model was established. The model parameters include the elastic modulus of the structural material (210 GPa, corresponding to aluminum alloy), Poisson's ratio (0.33), yield strength (300 MPa), and the geometric parameters of the structure (e.g., the cross-sectional area of ​​the support frame is 10 square centimeters, and its length is 20 centimeters). The model employs a nonlinear finite element analysis method, considering the elastoplastic properties of the material. Through iterative calculations, it simulates the stress response variation of the mechanical structure under different operating conditions. The solution accuracy of the model is set to a stress error of no more than 1% and a displacement error of no more than 0.1 mm. This model can accurately predict the mechanical behavior of the internal mechanical structure of the power supply under various complex operating conditions, providing a reliable structural safety assessment basis for subsequent adaptive power supply control.

[0022] Step S2: Collect the power supply electrical parameters during the power supply operation; perform parameter correlation detection on the power supply electrical parameters to obtain electrical parameter correlation data; use the electrical parameter correlation data to perform electrical vibration influence mapping on the mechanical support structure response model to generate structural vibration mapping data; In this embodiment of the invention, electrical parameters during power supply operation are first acquired using a high-precision electrical parameter acquisition module. This module employs a high-precision Hall current sensor and a high-precision voltage sensor to monitor the power supply's output current and output voltage in real time. The Hall current sensor has a measurement range of 0 to 30 amperes, an accuracy of ±0.1%, and a sampling frequency of 1000 times / second; the voltage sensor has a measurement range of 0 to 24 volts, an accuracy of ±0.05%, and also a sampling frequency of 1000 times / second. The acquired power supply electrical parameters include real-time output current (I_out) and real-time output voltage (V_out), and the data is stored in a local database in time-series format. Each data point includes a timestamp, current value, and voltage value. Subsequently, parameter correlation detection is performed on the acquired power supply electrical parameters. A correlation coefficient analysis-based method is used to calculate the correlation between the output current (I_out) and the output voltage (V_out). Specifically, the acquired current and voltage data are time-aligned to ensure that their timestamps are completely consistent. Then, the linear correlation between the two is calculated using the Pearson correlation coefficient formula, which is: ; in, and Let be the current and voltage values ​​at the i-th time point, respectively. and Here, n represents the average values ​​of current and voltage, respectively, and n is the total number of data points. The calculated correlation coefficient (r) is stored as electrical parameter correlation data, with a value range of [-1, 1], representing the strength of the linear correlation between current and voltage. Finally, the electrical vibration influence is mapped onto the mechanical support structure response model using the electrical parameter correlation data. Modal analysis technology is used, combined with the electrical parameter correlation data, to generate structural vibration mapping data. Specifically, modal analysis is first performed on the internal mechanical structure of the power supply. The structure is excited to vibration using a modal analyzer, and its vibration response signal is collected to obtain the natural frequencies and mode shapes of the structure. The excitation frequency range of the modal analyzer is 0 to 500 Hz, and the sampling frequency is 2000 times / second. Then, based on the magnitude of the electrical parameter correlation data (correlation coefficient r), the degree of influence of changes in electrical parameters during power supply operation on the vibration of the mechanical structure is evaluated. When the absolute value of the correlation coefficient r is large, it indicates that the changes between current and voltage have a strong influence on the vibration of the mechanical structure. At this point, based on the modal analysis results, the vibration effects caused by changes in electrical parameters are mapped onto the corresponding mode shapes of the mechanical structure, generating structural vibration mapping data, including vibration frequency, amplitude, and phase information. For example, when the correlation coefficient r is 0.8, it indicates that changes in current and voltage have a significant impact on the vibration of the mechanical structure. Modal analysis reveals that the mode shape at a certain natural frequency (e.g., 150 Hz) is the main affected mode shape, with a vibration amplitude of 0.05 mm and a phase of 30 degrees. This information is stored as structural vibration mapping data and used for subsequent analysis.

[0023] Step S3: Collect power load data during power supply operation; mark the power load data as a time series and monitor fluctuating load to obtain fluctuating load sequence data; use the fluctuating load sequence data to perform power ripple response mapping on the mechanical support structure response model to generate ripple response mapping data. In this embodiment of the invention, in step S3, power load data during power supply operation is first acquired using a high-precision power sensor. This power sensor employs high-precision Hall effect power measurement technology, with a measurement range of 0 to 1000 watts, an accuracy of ±0.2%, and a sampling frequency of 2000 times / second. The acquired power load data includes real-time power values ​​(P_load), stored in a local database in time-series format, with each data point containing a timestamp and power value. Next, the acquired power load data is time-series labeled. The power load data is sorted using timestamp information to ensure that the data is arranged in chronological order. Then, fluctuating load monitoring is performed on the power load data. A sliding window algorithm is used to analyze the power load data, with a window size of 100 milliseconds and a step size of 10 milliseconds. Within each sliding window, the standard deviation (σ_power) of the power load data is calculated to assess the degree of power load fluctuation. When the standard deviation σ_power exceeds a preset threshold (e.g., 5 watts), it is determined to be a fluctuating load event, and the start time, end time, and maximum power fluctuation amplitude (ΔP_max) of the event are recorded. This information constitutes fluctuating load sequence data, which is stored in a database. Finally, power ripple response mapping is performed on the mechanical support structure response model using the fluctuating load sequence data, generating ripple response mapping data. The vibration response of the mechanical support structure is monitored using a dynamic signal analyzer with a sampling frequency of 5000 times / second and a frequency analysis range of 0 to 1000 Hz. During fluctuating load events, the vibration acceleration signal of the mechanical support structure is recorded. The vibration acceleration signal is converted from the time domain to the frequency domain using Fourier transform to analyze the vibration frequency components and amplitude caused by power ripple. For example, in a certain fluctuating load event, the standard deviation of the power load σ_power is 8 watts, and the maximum power fluctuation amplitude ΔP_max is 15 watts. Fourier transform analysis of the vibration acceleration signal reveals a significant vibration frequency component at 100 Hz, with an amplitude of 0.1g (10% of gravitational acceleration). This vibration frequency and amplitude information is stored as ripple response mapping data for subsequent power adaptive control analysis.

[0024] Step S4: Collect heat accumulation data during power supply operation, identify and record the concentrated heat accumulation area of ​​the power supply through the heat accumulation data; perform heat dissipation response mapping on the mechanical support structure response model based on the concentrated heat accumulation area of ​​the power supply, and generate heat dissipation response mapping data. In this embodiment of the invention, in step S4, heat accumulation data during the operation of the power supply is first collected using a thermal imager. The thermal imager employs non-contact infrared thermal imaging technology, with a temperature measurement range of 0℃ to 100℃, a spatial resolution of 0.1℃, and an imaging frequency of 1 frame / second. The thermal imager performs real-time thermal imaging of key components inside the power supply, acquiring temperature distribution images of the power supply casing, heat sink, capacitors, transformers, and other parts. The collected heat accumulation data is stored in the form of a temperature image sequence, with each frame containing temperature information for various parts inside the power supply. Next, the collected heat accumulation data is analyzed to determine the concentrated heat accumulation areas of the power supply. The temperature images are processed using image processing technology to extract temperature information. First, the temperature images are converted to grayscale, enhancing the contrast of temperature differences. Then, a threshold segmentation algorithm is used to segment the grayscale images, identifying areas with temperatures higher than a preset threshold (e.g., 60℃) as concentrated heat accumulation areas. The preset threshold is determined based on the normal operating temperature range of the internal components of the power supply. In the segmented image, areas of concentrated heat accumulation are highlighted, and their location coordinates and area size are recorded. For example, if the temperature of a capacitor region inside the power supply exceeds 60℃, this region is identified as a concentrated heat accumulation area, with location coordinates (X1, Y1) and area A1 square millimeters. This information is recorded and stored. Finally, a heat dissipation response mapping is performed on the mechanical support structure response model based on the concentrated heat accumulation areas of the power supply, generating heat dissipation response mapping data. Thermal analysis of the mechanical support structure is performed using finite element analysis software, with the location coordinates and area size of the concentrated heat accumulation areas used as heat source input parameters. In the finite element analysis, the thermal conductivity coefficient is set to 0.2 W / (m·K), the convective heat transfer coefficient to 10 W / (m²·K), and the radiative heat transfer coefficient to 5 W / (m²·K). These parameters are determined based on the thermophysical properties of the mechanical support structure material and the operating environment. The temperature distribution and heat dissipation path of the mechanical support structure in the concentrated heat accumulation areas are calculated through thermal analysis. For example, the analysis revealed that the temperature of the mechanical support structure near the heat accumulation concentration area rose to 50°C. The heat dissipation path mainly involved heat conduction through contact with the heat sink, with a heat dissipation power of P_heat W. This temperature distribution and heat dissipation path information was stored as heat dissipation response mapping data for subsequent power supply adaptive regulation analysis.

[0025] Step S5: Perform multimodal data fusion based on structural vibration mapping data, ripple response mapping data, and heat dissipation response mapping data, and determine the power supply adaptive control strategy; dynamically adjust and control the power supply according to the power supply adaptive control strategy.

[0026] In this embodiment of the invention, in step S5, multimodal data fusion is first performed on the structural vibration mapping data, ripple response mapping data, and heat dissipation response mapping data. A data fusion algorithm-based processing technique is used to integrate the three data types. Specifically, a weighted average method is used to fuse the vibration frequency (f_vibration), amplitude (A_vibration), and phase (φ_vibration) in the structural vibration mapping data; the ripple frequency (f_ripple) and ripple amplitude (A_ripple) in the ripple response mapping data; and the temperature distribution (T_distribution) and heat dissipation path (P_heat) in the heat dissipation response mapping data. Different weighting coefficients are assigned to each data type based on their importance to power supply performance. For example, the weighting coefficient for the structural vibration mapping data is 0.4, the weighting coefficient for the ripple response mapping data is 0.3, and the weighting coefficient for the heat dissipation response mapping data is 0.3. A comprehensive evaluation index (I_comprehensive) is calculated through weighted averaging to reflect the overall impact on the power supply's operating status. Based on the comprehensive evaluation index (I_comprehensive), an adaptive power supply control strategy is determined. A threshold judgment method is adopted, and the comprehensive evaluation index is divided into different control levels according to a preset threshold range. For example, when I_comprehensive is less than threshold 1 (e.g., 0.5), it indicates that the power supply is operating well and no control is needed; when I_comprehensive is between threshold 1 and threshold 2 (e.g., 1.0), it indicates that the power supply is operating slightly abnormal and requires light control; when I_comprehensive is greater than threshold 2, it indicates that the power supply is operating severely abnormal and requires heavy control. Based on the control level, corresponding adaptive control strategies are formulated, including adjusting the power supply output power, changing the cooling fan speed, and optimizing the power supply operating mode. The power supply is dynamically adjusted and controlled according to the adaptive control strategy. Closed-loop control technology is adopted, where the controller receives the control strategy instructions, converts them into specific control signals, and sends them to the power supply's execution components. For example, if the control strategy is light control, the controller adjusts the output power from the current value P_current to the target value P_target, with an adjustment step of ΔP_step (e.g., 10 watts) and an adjustment time of T_adjust (e.g., 1 second). Simultaneously, based on the heat dissipation response mapping data, the controller adjusts the cooling fan speed from the current speed RPM_current to the target speed RPM_target, with an adjustment step of ΔRPM_step (e.g., 100 rpm) and an adjustment time of T_fan_adjust (e.g., 0.5 seconds). Through dynamic adjustment control, the power supply's operating state is optimized to the target state, ensuring stable and efficient operation under various working conditions.

[0027] Of particular importance is that step S1 involves acquiring data on the internal mechanical structure of the power supply and recording the power supply's operating status data, including: Data on the meshing clearance of gears inside the power supply is collected, and the axial displacement during gear meshing is measured using a high-precision displacement sensor with a measurement accuracy of 0.01 mm and a measurement frequency of 10 times per second. Record the compression state data of the spring inside the power supply, and use a pressure sensor to detect the compression force of the spring under different loads. Record the curve of the relationship between compression force and time with a time resolution of 1 millisecond. The swing angle of the internal linkage of the power supply is measured, and the change of the swing angle of the linkage during operation is monitored in real time using an angle encoder. The angle measurement accuracy is 0.1 degrees, and the data update frequency is 50 times per second.

[0028] In this embodiment of the invention, a high-precision laser displacement sensor is used to measure the meshing clearance data of the gears inside the power supply. This sensor is installed near the gear meshing area and can measure the axial displacement of the gears during meshing in a non-contact manner. The sensor has a measurement accuracy of 0.01 mm and a measurement frequency of 10 times per second. During the measurement process, the laser displacement sensor emits a laser beam that illuminates the gear surface, and the axial displacement of the gear is calculated by detecting the displacement change of the reflected light. The sensor outputs the measured displacement data in digital signal form and stores it in a local database via a data acquisition card. Each data point includes a timestamp and the corresponding axial displacement (X_gear). For recording the compression state data of the springs inside the power supply, a high-precision pressure sensor is used to detect the compressive force of the springs under different loads. The pressure sensor is installed at the force-bearing end of the spring and can measure the compressive force on the spring in real time. The sensor's measurement range is 0 to 100 Newtons, with an accuracy of ±0.1 Newtons. The data acquisition system records the relationship curve between compressive force and time with a time resolution of 1 millisecond. During measurement, the pressure sensor converts the compressive force into an electrical signal. After amplification and filtering, the signal is sampled by a data acquisition card at a frequency of 1000 times per second, and the sampled data is stored as a time series. Each data point includes a timestamp and the corresponding compressive force value (F_spring). When measuring the swing angle of the internal connecting rod, a high-precision angle encoder is used for real-time monitoring. The angle encoder is mounted on the rotating shaft of the connecting rod and can accurately measure the change in the swing angle of the connecting rod during operation. The angle encoder has a measurement accuracy of 0.1 degrees and a data update frequency of 50 times per second. During measurement, the angle encoder converts the rotation angle of the connecting rod into a digital signal through an internal photoelectric or magnetoelectric conversion element. This signal is transmitted to the data processing system through a communication interface and stored in the form of a time series. Each data point includes a timestamp and the corresponding swing angle (θ_link).

[0029] Of particular importance, step S1 involves analyzing the stress response characteristics of the internal mechanical structure data of the power supply under different power supply operating states and establishing a mechanical support structure response model, including: Based on the current waveform characteristics in the power supply operating status data, the internal mechanical structure data of the power supply is classified and processed, and the mechanical structure data is divided into different data subsets according to the frequency and amplitude range of the current waveform. For each independent component of the internal mechanical structure of the power supply, the stress response characteristic parameters under different current waveform conditions are extracted, specifically the stress amplitude, stress phase, and stress frequency components. By combining the geometric shape and size information of the internal mechanical structure of the power supply with the voltage fluctuations in the power supply's operating status data, a response model of the mechanical support structure is established through a multiphysics coupling analysis method.

[0030] In this embodiment of the invention, during implementation, the internal mechanical structure data of the power supply is first classified based on the current waveform characteristics in the power supply operating status data. Fast Fourier Transform (FFT) technology is used to perform frequency domain analysis on the acquired current waveform data to extract the frequency and amplitude information of the current waveform. Specifically, the current waveform data is divided into time windows with a window length of 100 milliseconds and a step size of 10 milliseconds. Within each window, the spectrum of the current waveform is calculated to determine its main frequency components and amplitude range. Based on preset frequency and amplitude thresholds, the mechanical structure data is divided into different data subsets. For example, current waveforms with frequencies below 50 Hz and amplitudes less than 10 amps are classified as low-frequency, low-amplitude subsets; current waveforms with frequencies between 50 and 100 Hz and amplitudes between 10 and 20 amps are classified as mid-frequency, mid-amplitude subsets; and current waveforms with frequencies above 100 Hz and amplitudes greater than 20 amps are classified as high-frequency, high-amplitude subsets. Each subset corresponds to a set of mechanical structure data, stored in the database, and the data subsets are identified as SubSet_Low, SubSet_Medium, and SubSet_High. Next, for each independent component of the power supply's internal mechanical structure, stress response characteristic parameters under different current waveform conditions are extracted. Finite element analysis (FEA) technology, combined with modal analysis and transient dynamic analysis methods, is used to perform stress analysis on the mechanical structure components. Taking the cooling fan bracket inside the power supply as an example, under low-frequency, low-amplitude current waveform conditions, the corresponding loads and boundary conditions are applied using the finite element software ANSYS, and the calculated stress amplitude (σ_Low) of the cooling fan bracket is 50 MPa, the stress phase (φ_Low) is 0 degrees, and the stress frequency component (f_Low) is 50 Hz. Under medium-frequency, medium-amplitude current waveform conditions, the calculated stress amplitude (σ_Medium) is 100 MPa, the stress phase (φ_Medium) is 30 degrees, and the stress frequency component (f_Medium) is 75 Hz. Under high-frequency, high-amplitude current waveform conditions, the calculated stress amplitude (σ_High) is 150 MPa, the stress phase (φ_High) is 60 degrees, and the stress frequency component (f_High) is 120 Hz. These stress response characteristic parameters are stored in the corresponding current waveform subsets. Finally, the geometric shape and dimensional information of the internal mechanical structure of the power supply are combined with the voltage fluctuation data in the power supply's operating status data to establish a response model of the mechanical support structure using a multiphysics coupling analysis method. The geometric model and dimensional parameters of the mechanical structure, including the cross-sectional area (A = 10 cm²), length (L = 20 cm), and material properties (elastic modulus E = 210 GPa, Poisson's ratio ν = 0.3) of the cooling fan bracket are imported into the multiphysics simulation software COMSOL Multiphysics. Simultaneously, voltage fluctuation data is used as the electric field excitation input, with a voltage fluctuation range of 11 to 13 volts and a frequency of 50 to 100 Hz.Through multiphysics coupling analysis, considering the interaction between electric and mechanical fields, the stress distribution and deformation of the mechanical support structure under different voltage fluctuation conditions are calculated. The solution accuracy of the model is set to a stress error of no more than 1% and a displacement error of no more than 0.1 mm. For example, when the voltage fluctuation frequency is 75 Hz and the amplitude is 12 volts, the maximum stress of the cooling fan bracket is 120 MPa and the maximum displacement is 0.05 mm.

[0031] Preferably, step S2, which involves parametric correlation detection of the power supply electrical parameters, includes: Power supply electrical parameters are classified into voltage, current, and frequency according to parameter type; The voltage, current, and frequency are divided into multiple equal-length time segments in chronological order. Within each time period, the changing trends of voltage and current are analyzed synchronously to obtain voltage-current parameter correlation data; if both voltage and current rise or fall within the same time period, it is marked as a change in the same direction; if voltage rises while current falls, or voltage falls while current rises, it is marked as a change in opposite directions. Within each time period, the changing trends of frequency and current are analyzed synchronously to obtain frequency-current parameter correlation data; if both frequency and current rise or fall within the same time period, it is marked as a change in the same direction; if frequency rises while current falls, or frequency falls while current rises, it is marked as a change in opposite directions. Electrical parameter correlation data is determined based on pressure-current parameter correlation data and frequency-current parameter correlation data.

[0032] In this embodiment of the invention, the power supply electrical parameters are divided into voltage (V), current (I), and frequency (f) according to parameter type. Time-series data of these parameters are acquired through a high-precision electrical parameter acquisition module, where the sampling frequency for voltage and current is 1000 times / second, and the sampling frequency for frequency is 100 times / second, ensuring high time resolution of the data. Next, the voltage, current, and frequency are divided into multiple equal-length time intervals according to time sequence. The length of each time interval is set to 100 milliseconds to ensure that transient changes in the electrical parameters can be captured. For example, from the acquired electrical parameter data, data from time point t0 to t0+100ms is extracted as a dataset for one time interval, including a voltage dataset (V_segment), a current dataset (I_segment), and a frequency dataset (f_segment). Within each time interval, the changing trends of voltage and current are analyzed synchronously. Linear fitting techniques are used to fit the voltage and current data separately, and their slopes (slope_V and slope_I) are calculated. If the slopes of voltage and current are both positive or both negative within the same time period, they are marked as changing in the same direction (labeled "same direction"). If the voltage slope is positive and the current slope is negative, or vice versa, they are marked as changing in opposite directions (labeled "reverse"). For example, if the voltage slope_V is 0.5V / s and the current slope_I is -0.3A / s within a certain time period, the voltage-current parameter correlation data for that time period is marked as "reverse". Similarly, the trends of frequency and current are analyzed synchronously within each time period. The same linear fitting technique is used to fit the frequency and current data separately, and their slopes (slope_f and slope_I) are calculated. If the slopes of frequency and current are both positive or both negative within the same time period, they are marked as changing in the same direction; if the frequency slope is positive and the current slope is negative, or vice versa, they are marked as changing in opposite directions. For example, if the frequency slope_f is 0.1 Hz / s and the current slope_I is -0.2 A / s within a certain time period, then the frequency-current parameter correlation data for that time period is marked as "reverse". Finally, electrical parameter correlation data is determined based on the voltage-current parameter correlation data and the frequency-current parameter correlation data. The voltage-current parameter correlation data and the frequency-current parameter correlation data for each time period are integrated to form a complete electrical parameter correlation dataset. For example, if the voltage-current parameter correlation data is "reverse" and the frequency-current parameter correlation data is "same direction" within a certain time period, then the electrical parameter correlation data for that time period is recorded as "voltage-current reverse, frequency-current same direction". This data is stored in a database for subsequent power supply adaptive control analysis.

[0033] Preferably, step S2, which maps the electrical vibration influence on the mechanical support structure response model using electrical parameter correlation data, includes: Identify the input-end fixed support and the output-end insulating support in the mechanical support structure response model; The pressure-flow fluctuation in the pressure-flow parameter correlation data is evaluated to obtain the pressure-flow fluctuation degree value; Calculate the pressure flow vibration intensity factor for the fixed support at the input end based on the pressure flow fluctuation value; The pressure flow vibration intensity factor is mapped to the mechanical support structure response model, the degree of influence of the pressure flow factor is recorded, and the input vibration mapping data is generated. Frequency flow changes are detected in the frequency flow parameter correlation data to obtain the degree of frequency flow change value; The frequency current vibration intensity factor is calculated for the output end insulating bracket based on the frequency current variation value. The frequency-current vibration intensity factor is mapped to the mechanical support structure response model, the degree of influence of the frequency-current factor is recorded, and the output vibration mapping data is generated. The vibration mapping data at the input and output ends is mapped to the mechanical support structure response model, and the electrical vibration characteristics are detected to generate structural vibration mapping data.

[0034] In this embodiment of the invention, the input-end fixed bracket and the output-end insulating bracket in the mechanical support structure are identified. The mechanical support structure is modeled using 3D modeling software, and the positions of the input-end fixed bracket and the output-end insulating bracket are marked in the model. The input-end fixed bracket is used to fix the input components of the power supply, and the output-end insulating bracket is used to support the output components of the power supply and ensure electrical insulation. Next, the voltage and current fluctuations in the voltage-current parameter correlation data are evaluated. The standard deviation calculation method is used to quantify the trends of voltage and current changes. Specifically, within each time period, the standard deviations (σ_V and σ_I) of the voltage dataset (V_segment) and the current dataset (I_segment) are calculated to obtain the voltage-current fluctuation degree value (Δ_PQ). For example, if the voltage standard deviation σ_V is 0.5 volts and the current standard deviation σ_I is 0.3 amperes within a certain time period, then the voltage-current fluctuation degree value Δ_PQ is σ_V + σ_I = 0.8. Based on the voltage-current fluctuation degree value Δ_PQ, the voltage-current vibration intensity factor (F_PQ_input) of the input-end fixed bracket is calculated. A linear mapping method is used to map the pressure-flow fluctuation value Δ_PQ to the range of vibration intensity factors. The calculation formula for the pressure-flow vibration intensity factor is set as: F_PQ_input = K_PQ × Δ_PQ, where K_PQ is the mapping coefficient, with a value of 1.2. For example, if the pressure-flow fluctuation value Δ_PQ is 0.8, then the pressure-flow vibration intensity factor F_PQ_input of the input end fixed support is 1.2 × 0.8 = 0.96. The pressure-flow vibration intensity factor F_PQ_input is mapped to the mechanical support structure response model, and the degree of influence of the pressure-flow factor is recorded. Using finite element analysis software, the pressure-flow vibration intensity factor is used as the vibration excitation of the input end fixed support, and its influence on the mechanical structure is calculated. The vibration response of the input end fixed support is recorded, including vibration displacement (D_input), vibration velocity (V_input), and vibration acceleration (A_input). Input end vibration mapping data is generated, storing the values ​​of vibration displacement, velocity, and acceleration and their corresponding timestamps. Frequency and current changes in the associated frequency and current parameter data are detected. The rate of change of frequency and current (Δf and ΔI) is calculated using the differential method. Within each time period, the rate of change of frequency Δf = (f_end - f_start) / Δt and the rate of change of current ΔI = (I_end - I_start) / Δt are calculated, where Δt is the length of the time period. The frequency and current change magnitude value (Δ_fI) is obtained as Δf + ΔI. For example, if the rate of change of frequency Δf is 0.05 Hz / s and the rate of change of current ΔI is 0.1 amperes / s within a certain time period, then the frequency and current change magnitude value Δ_fI is 0.15. Based on the frequency and current change magnitude value Δ_fI, the frequency and current vibration intensity factor (F_fI_output) of the output-end insulating support is calculated.Using the same linear mapping method, the calculation formula for the frequency-current vibration intensity factor is set as: F_fI_output = K_fI × Δ_fI, where K_fI is the mapping coefficient, with a value of 1.5. For example, if the frequency-current variation value Δ_fI is 0.15, then the frequency-current vibration intensity factor F_fI_output of the output-end insulating support is 1.5 × 0.15 = 0.225. The frequency-current vibration intensity factor F_fI_output is mapped to the mechanical support structure response model, and the degree of influence of the frequency-current factor is recorded. Using finite element analysis software, the frequency-current vibration intensity factor is used as the vibration excitation of the output-end insulating support, and its influence on the mechanical structure is calculated. The vibration response of the output-end insulating support is recorded, including vibration displacement (D_output), vibration velocity (V_output), and vibration acceleration (A_output). Output-end vibration mapping data is generated, storing the values ​​of vibration displacement, velocity, and acceleration and their corresponding timestamps. Finally, based on the input-end vibration mapping data and the output-end vibration mapping data, it is mapped to the mechanical support structure response model, and electrical vibration characteristics are detected. Modal analysis techniques, combined with vibration mapping data from both the input and output ends, are used to analyze the overall vibration response of a mechanical structure. The natural frequencies, mode shapes, and vibration modes of the mechanical structure are extracted to generate structural vibration mapping data. This data includes the frequency distribution, amplitude, and phase relationship of the mechanical structure under different vibration excitations.

[0035] As an example of the present invention, reference is made to... Figure 2 As shown, step S3, which involves time-series labeling of the power load data and monitoring of fluctuating loads, includes: S31: Based on the start time of power supply operation, the power load data is divided into 1-millisecond time intervals, with each time interval corresponding to a power load data point. S32: Assign a time stamp to each power load data point. The time stamp is in milliseconds and records the specific position of the data point in the time series. S33: Analyze each power load data point in the time series one by one, and calculate the power difference between adjacent power load data points; S34: When the power difference exceeds 5% of the rated power of the power supply, identify the power load data point as a power load fluctuation point, and record the power value and corresponding time identifier of the power load data point; S35: Arrange all identified power load fluctuation points in the order of time identifiers to form fluctuating load sequence data; in the fluctuating load sequence data, record the power value, time identifier, and power difference of each fluctuation point.

[0036] In this embodiment of the invention, in step S31, power load data is first collected and segmented using a high-precision data acquisition system based on the start time of power supply operation. This system is equipped with a high-precision power sensor with a sampling frequency of 1000 times / second, capable of continuously collecting power load data at 1-millisecond intervals. The collected power load data is stored in time series format, with each time interval corresponding to a power load data point (P_load_point), containing a power value (P_value) and a corresponding timestamp. In step S32, a time identifier (Time_ID) is assigned to each power load data point. The time identifier is in milliseconds, starting from the start time (t0) of power supply operation, sequentially marking the specific position of each data point in the time series. For example, the time identifier for the first power load data point is 0 milliseconds, the second is 1 millisecond, and so on. The time identifier and the power value (P_value) are stored together in a database to form a complete power load data sequence, with the data format (Time_ID, P_value). Proceed to step S33, analyze each power load data point in the time series one by one, and calculate the power difference (ΔP) between adjacent power load data points. Specifically, for each data point (Time_ID_i, P_value_i), calculate the power difference between it and the previous data point (Time_ID_i-1, P_value_i-1), i.e., ΔP_i = P_value_i - P_value_i-1. For example, if the power value of a data point is 100 watts and the power value of the previous data point is 95 watts, then the power difference ΔP_i is 5 watts. In step S34, determine whether the power difference exceeds 5% of the power supply's rated power (P_nominal). The power supply's rated power is a known parameter; for example, if P_nominal is 500 watts, then the 5% threshold is 25 watts. When the power difference ΔP_i exceeds 25 watts, identify the power load data point as a power load fluctuation point (P_fluctuation_point). Record the power value (P_fluctuation_value) and corresponding time identifier (Time_ID_fluctuation) for each fluctuation point. For example, if the power difference ΔP_i of a data point is 30 watts, exceeding the threshold of 25 watts, then this data point is identified as a power load fluctuation point, with a power value P_fluctuation_value of 100 watts and a time identifier Time_ID_fluctuation of 10 milliseconds. Finally, in step S35, all identified power load fluctuation points are arranged in order according to their time identifiers to form a fluctuation load sequence data (Fluctuation_Sequence).In fluctuating load sequence data, the power value (P_fluctuation_value), time stamp (Time_ID_fluctuation), and power difference (ΔP_fluctuation) are recorded for each fluctuation point. For example, fluctuating load sequence data may contain records such as: (Time_ID_fluctuation_1=10ms, P_fluctuation_value_1=100W, ΔP_fluctuation_1=30W), (Time_ID_fluctuation_2=50ms, P_fluctuation_value_2=120W, ΔP_fluctuation_2=40W), etc. This data is stored in a dedicated fluctuating load database for subsequent power supply adaptive regulation analysis.

[0037] Preferably, step S3, which maps the power ripple response of the mechanical support structure response model using fluctuating load sequence data, includes: The switching contact points of the mechanical support structure response model are identified by using fluctuating load sequence data, and their response characteristics under power fluctuations are detected; when load fluctuations occur in the fluctuating load sequence data, the contact resistance value of the switching contact points is recorded. For each load fluctuation point in the fluctuating load sequence data, if the contact resistance value shows a linear increasing trend when the load fluctuation point appears, then the relationship between the load fluctuation point and the switch contact point is determined to be a response synchronization relationship. The load fluctuation points are mapped to the switch contact area of ​​the mechanical support structure response model, and the power change amplitude, change frequency and duration of the switch contact area are recorded. Based on the power change amplitude and duration of the fluctuation point, the ripple response intensity of the switch contact point is classified and ripple response mapping data is generated. When the power change exceeds 15% of the rated power and lasts for more than 10 milliseconds, it is marked as high-intensity ripple response data; When the power change is between 10% and 15% of the rated power and the duration is between 5 and 10 milliseconds, it is marked as medium intensity ripple response data. Data with a power change of less than 10% of the rated power and a duration of less than 5 milliseconds is classified as low-intensity ripple response data.

[0038] In this embodiment of the invention, switch contact points in the response model of a mechanical support structure are identified using fluctuating load sequence data. A high-precision resistance meter is used to monitor the contact resistance of the switch contact points in real time, achieving a measurement accuracy of 0.01 ohms and a sampling frequency of 1000 times / second. When a load fluctuation point appears in the fluctuating load sequence data, the contact resistance value (R_contact) of the switch contact point at that moment is recorded. For example, when a certain load fluctuation point (Time_ID_fluctuation = 20ms) occurs, the recorded contact resistance value is R_contact = 0.1 ohms. For each load fluctuation point in the fluctuating load sequence data, the changing trend of the contact resistance value is analyzed. Linear regression analysis is used to fit the contact resistance value before and after the occurrence of the load fluctuation point. If the fitting result shows that the contact resistance value increases linearly when the load fluctuation point occurs, it is determined that the relationship between the load fluctuation point and the switch contact point is a response synchronization relationship. For example, through linear regression analysis, it is found that the contact resistance value increases linearly before and after the load fluctuation point, with a slope of 0.005 ohms / millisecond, thus it is determined that the load fluctuation point and the switch contact point have a response synchronization relationship. Load fluctuation points are mapped to the switch contact area of ​​the mechanical support structure response model. The power change amplitude (ΔP_contact), frequency (f_contact), and duration (T_contact) of the switch contact area are recorded. The power change amplitude is obtained by calculating the difference between the power value at the load fluctuation point and the power value at the previous data point; the frequency is calculated by counting the number of load fluctuation points per unit time; and the duration is obtained by recording the difference between the start and end times of the load fluctuation point. For example, a load fluctuation point might have a power change amplitude of ΔP_contact = 50 watts, a frequency of f_contact = 2 Hz, and a duration of T_contact = 15 milliseconds. Based on the power change amplitude and duration of the fluctuation point, the ripple response intensity of the switch contact point is graded. The data is classified using a preset grading standard: when the power change exceeds 15% of the rated power (P_nominal) and lasts for more than 10 milliseconds, it is marked as high-intensity ripple response data; when the power change is between 10% and 15% of the rated power and lasts for 5 to 10 milliseconds, it is marked as medium-intensity ripple response data; and when the power change is less than 10% of the rated power and lasts for less than 5 milliseconds, it is marked as low-intensity ripple response data. For example, if the power supply's rated power is P_nominal = 500 watts, and the power change at a certain load fluctuation point is ΔP_contact = 80 watts (exceeding 15% of the rated power) and lasts for T_contact = 15 milliseconds (exceeding 10 milliseconds), then it is marked as high-intensity ripple response data, and the graded ripple response data is stored as ripple response mapping data.

[0039] Preferably, in step S4, the heat accumulation data collected during the power supply operation, and the heat accumulation concentration areas of the power supply determined and recorded through the heat accumulation data, include: Temperature sensors are installed at the locations of the heat sink, power module, capacitors, and transformer inside the power supply; each temperature sensor collects temperature data at fixed time intervals of 1 second. For each sensor location, the difference between the current power supply operating temperature and the previous power supply operating temperature is used to obtain the power supply operating temperature difference. Identify areas of sustained temperature rise in the power supply operating temperature difference and mark them as potential areas of concentrated heat accumulation; If the temperature difference of the power supply operating temperature in the potential heat accumulation zone exceeds the preset temperature rise value, where the preset temperature rise value is 2 degrees Celsius per second, it is determined to be a heat accumulation zone of the power supply.

[0040] In this embodiment of the invention, during implementation, temperature sensors are first installed at key locations inside the power supply, specifically including heat sinks, power modules, capacitors, and transformers. These temperature sensors are high-precision thermistor temperature sensors with a measurement range of 0℃ to 100℃ and an accuracy of ±0.1℃. Each temperature sensor collects temperature data at fixed time intervals of 1 second and stores the collected temperature data (T_sensor) in the form of a time series. Each data point includes a timestamp (Time_stamp) and a corresponding temperature value (T_value). For each sensor location, the difference between the current power supply operating temperature and the previously collected power supply operating temperature is calculated to obtain the power supply operating temperature difference (ΔT). Specifically, for the temperature data (T_value_i) at each time point, the difference between it and the temperature data at the previous time point (T_value_i-1) is calculated, i.e., ΔT_i = T_value_i - T_value_i-1. For example, if the temperature value at a certain time point is 45℃ and the temperature value at the previous time point is 44℃, then the temperature difference ΔT_i is 1℃. Next, regions with continuously rising temperatures within the power supply operating temperature difference are identified. A sliding window algorithm is used to analyze the temperature difference sequence, with a window size of 5 time points (5 seconds) and a step size of 1 time point (1 second). Within each sliding window, it is determined whether the temperature difference is continuously positive. If all temperature differences within the window are positive, the region is marked as a potential heat accumulation concentration area (Potential_Heat_Area). For example, if the temperature differences within a certain sliding window are 0.5℃, 0.3℃, 0.2℃, 0.4℃, and 0.1℃, then the region is marked as a potential heat accumulation concentration area. Finally, it is determined whether the power supply operating temperature difference within the potential heat accumulation concentration area exceeds a preset temperature rise value. The preset temperature rise value is set to increase by 2℃ per second. Within the potential heat accumulation concentration area, if the temperature difference ΔT at any time point exceeds 2℃, then the region is determined to be a power supply heat accumulation concentration area (Heat_Critical_Area). For example, if the temperature difference ΔT at a certain point in time is 2.5℃ within a potential heat accumulation concentration area, then that area is identified as a power source heat accumulation concentration area. The location, start time, and end time of the heat accumulation concentration area are recorded and stored for subsequent power source adaptive control analysis.

[0041] Preferably, step S4, which maps the heat dissipation response of the mechanical support structure response model based on the concentrated heat accumulation area of ​​the power source, includes: Measure the temperature change rate and peak temperature of each heat accumulation zone, and classify the heat accumulation zones into heat dissipation demand levels based on the temperature change rate and peak temperature of each zone: If the rate of temperature change exceeds 3 degrees Celsius per second or the peak temperature exceeds 80 degrees Celsius, it is classified as a high heat dissipation requirement area: If the temperature change rate is between 1 and 3 degrees Celsius per second or the peak temperature is between 60 and 80 degrees Celsius, it is classified as a medium heat dissipation requirement zone. If the rate of temperature change is less than 1 degree Celsius per second or the peak temperature is below 60 degrees Celsius, it is classified as a low heat dissipation requirement zone. Identify heat sinks, thermal pads, ventilation holes, and heat dissipation channels in the mechanical support structure response model to obtain the heat dissipation hardware structure characteristics; The heat dissipation demand regions of high heat dissipation demand region, medium heat dissipation demand region, and low heat dissipation demand region are respectively matched with the heat dissipation hardware structure features to generate heat dissipation response mapping data.

[0042] In this embodiment of the invention, during implementation, the temperature change rate (ΔT_rate) and peak temperature (T_peak) of each heat accumulation concentration area are first measured. Temperature data of the heat accumulation concentration areas are collected using installed temperature sensors, and the temperature change rate is calculated. Specifically, within each heat accumulation concentration area, a continuous time period (e.g., 10 seconds) is selected, and the linear temperature change rate within that time period is calculated, i.e., ΔT_rate = (T_final - T_initial) / Δt, where T_final is the temperature at the end of the time period, T_initial is the temperature at the beginning of the time period, and Δt is the length of the time period. Simultaneously, the highest temperature value of the heat accumulation concentration area during the monitoring period is recorded as the peak temperature T_peak. For example, if the temperature of a heat accumulation concentration area rises from 50℃ to 65℃ in 10 seconds, then ΔT_rate = (65-50) / 10 = 1.5℃ / s; if the peak temperature of this area is 70℃, then T_peak = 70℃. Next, based on the temperature change rate and peak temperature of the heat accumulation concentration areas, they are classified into different heat dissipation demand levels. The specific classification criteria are as follows: if the temperature change rate exceeds 3℃ / s or the peak temperature exceeds 80℃, it is classified as a high heat dissipation demand area; if the temperature change rate is between 1-3℃ / s or the peak temperature is between 60-80℃, it is classified as a medium heat dissipation demand area; and if the temperature change rate is less than 1℃ / s or the peak temperature is less than 60℃, it is classified as a low heat dissipation demand area. For example, the ΔT_rate of the aforementioned heat accumulation concentration area is 1.5℃ / s, and the T_peak is 70℃. According to the classification criteria, this area is classified as a medium heat dissipation demand area. Subsequently, the heat dissipation hardware structure features in the mechanical support structure response model are identified. The mechanical support structure is modeled using 3D modeling software, and the positions and dimensions of heat sinks, thermal pads, ventilation holes, and heat dissipation channels are marked. The heatsink measures 10cm × 5cm × 1cm, the thermal pad is 0.5mm thick, the vent diameter is 2cm, and the cross-sectional area of ​​the heat dissipation channel is 5cm². These heat dissipation hardware structural features are recorded and stored for subsequent heat dissipation requirement matching. Finally, high-heat-demand areas, medium-heat-demand areas, and low-heat-demand areas are matched with the heat dissipation hardware structural features to generate heat dissipation response mapping data. Specifically, for high-heat-demand areas, the heatsink and heat dissipation channel are matched, and the heat dissipation power of the heatsink (P_heat_sink) and the airflow velocity of the heat dissipation channel (V_airflow) are calculated. For medium-heat-demand areas, the thermal pad and vent are matched, and the thermal resistance of the thermal pad (R_thermal_pad) and the ventilation volume of the vent (Q_ventilation) are calculated.For areas with low heat dissipation requirements, only ventilation holes are matched, and the airflow of the ventilation holes is calculated. For example, for areas with high heat dissipation requirements, the heat dissipation power of the heat sink is calculated as P_heat_sink=k×A×ΔT, where k is the heat dissipation coefficient, A is the area of ​​the heat sink, and ΔT is the temperature difference; the airflow velocity of the heat dissipation channel is calculated as V_airflow=Q / A, where Q is the airflow. The correspondence between the matched heat dissipation requirement areas and the characteristics of the heat dissipation hardware structure, as well as the relevant parameter values, are recorded as heat dissipation response mapping data for subsequent power supply adaptive control analysis.

[0043] Preferably, step S5 involves multimodal data fusion based on structural vibration mapping data, ripple response mapping data, and heat dissipation response mapping data, and determining the power supply adaptive control strategy, including: Frequency domain analysis was performed on the structural vibration mapping data to extract vibration frequency components and identify the primary and secondary vibration frequencies. Based on the primary and secondary vibration frequencies, the structural layout vibration impact was assessed on the mechanical support structure response model to obtain structural layout vibration impact data. Time-domain analysis was performed on the ripple response mapping data to extract the ripple amplitude and ripple frequency; based on the ripple amplitude and ripple frequency, the component ripple influence was evaluated on the mechanical support structure response model to obtain component ripple influence data; Heat flow analysis is performed on the heat dissipation response mapping data to extract the temperature change rate and heat dissipation demand level; the heat dissipation structure impact assessment is performed on the mechanical support structure response model based on the temperature change rate and heat dissipation demand level to obtain heat dissipation structure impact data. The vibration impact data of structural layout, the ripple impact data of components, and the impact data of heat dissipation structure are multimodally weighted and fused according to a weight of 4:3:3 to obtain multimodal fused data; Calculate the fusion index of the multimodal fusion data. If the fusion index is less than 30, it indicates that the power supply is in good condition and a low-level power supply control strategy is determined. If the fusion index is between 30 and 70, it indicates that the power supply is in a moderate condition and a medium-level power supply control strategy is determined. If the fusion index is greater than 70, it indicates that the power supply is in a poor condition and a high-level power supply control strategy is determined.

[0044] In this embodiment of the invention, the structural vibration mapping data is first analyzed in the frequency domain. The vibration signals in the structural vibration mapping data are processed using Fast Fourier Transform (FFT) technology to extract the vibration frequency components. Specifically, FFT analysis is performed on each vibration signal data segment to obtain its spectrum. In the spectrum, the frequency component with the largest amplitude is identified as the primary vibration frequency (f_main), and the frequency component with the second largest amplitude is identified as the secondary vibration frequency (f_secondary). For example, the FFT analysis result of a certain vibration signal shows that the primary vibration frequency f_main is 50 Hz and the secondary vibration frequency f_secondary is 100 Hz. Next, the structural layout vibration impact is assessed based on the primary and secondary vibration frequencies to evaluate the mechanical support structure response model. Modal analysis technology is used, combining the primary and secondary vibration frequencies, to analyze the modal response of the mechanical support structure at these frequencies. The structural layout vibration impact data (Vibration_Layout_Data) is calculated, including the displacement response (D_main and D_secondary) and stress response (σ_main and σ_secondary) of each component at the primary and secondary vibration frequencies. For example, a component's displacement response D_main at the primary vibration frequency is 0.1 mm, and its stress response σ_main is 100 MPa; at the secondary vibration frequency, its displacement response D_secondary is 0.05 mm, and its stress response σ_secondary is 50 MPa. Time-domain analysis is performed on the ripple response mapping data. Peak detection technology is used to extract the ripple amplitude (A_ripple) and ripple frequency (f_ripple). Specifically, in the time series of the ripple response mapping data, peaks and troughs are detected, the amplitude difference between adjacent peaks or troughs is calculated as the ripple amplitude, and the reciprocal of the time interval between adjacent peaks or troughs is calculated as the ripple frequency. For example, the ripple amplitude A_ripple of a certain ripple response data is 0.5 volts, and the ripple frequency f_ripple is 100 Hz. The ripple impact of components is assessed in the mechanical support structure response model based on the ripple amplitude and ripple frequency. Finite element analysis (FEA) technology is used, combined with ripple amplitude and ripple frequency, to calculate the dynamic response of each component under ripple excitation. The component ripple impact data (Ripple_Part_Data) is obtained, including the vibration displacement (D_ripple) and stress change (Δσ_ripple) of each component. For example, the vibration displacement (D_ripple) of a certain component under ripple excitation is 0.02 mm, and the stress change (Δσ_ripple) is 20 MPa. Heat flux analysis is performed on the heat dissipation response mapping data. Using heat flux analysis techniques, the temperature change rate (ΔT_rate) and heat dissipation demand level (Heat_Demand_Level) are extracted.The specific operation is as follows: In the heat dissipation response mapping data, the rate of temperature change over time is calculated as the temperature change rate; the heat dissipation demand level is determined according to the heat dissipation demand level classification standard. For example, if the temperature change rate of a certain heat dissipation response data is 2 degrees Celsius per second, the heat dissipation demand level is in the medium heat dissipation demand area. The impact of the heat dissipation structure on the mechanical support structure response model is evaluated based on the temperature change rate and the heat dissipation demand level. The thermal-structure coupling analysis technique is used to calculate the impact of the heat dissipation structure on the mechanical support structure by combining the temperature change rate and the heat dissipation demand level. The heat dissipation structure impact data (Heat_Structure_Data) is obtained, including the temperature distribution (T_distribution) and thermal stress (σ_thermal) of each component. For example, the temperature distribution T_distribution of a certain component is 60 degrees Celsius, and the thermal stress σ_thermal is 80 MPa. The structural layout vibration impact data (Vibration_Layout_Data), component ripple impact data (Ripple_Part_Data), and heat dissipation structure impact data (Heat_Structure_Data) are multimodally weighted and fused according to a weight of 4:3:3. The specific steps are as follows: Calculate the weighted fusion data (Multi_Mode_Data) as Multi_Mode_Data = 4 × Vibration_Layout_Data + 3 × Ripple_Part_Data + 3 × Heat_Structure_Data. For example, if the structural layout vibration influence data is 10, the component ripple influence data is 8, and the heat dissipation structure influence data is 6, then the weighted fusion data Multi_Mode_Data = 4 × 10 + 3 × 8 + 3 × 6 = 86. Calculate the fusion index (Fusion_Index) of the multimodal fusion data. The formula for calculating the fusion index is Fusion_Index = Multi_Mode_Data / (4 + 3 + 3). For example, if the above weighted fusion data Multi_Mode_Data is 86, then the fusion index Fusion_Index = 86 / 10 = 8.6. The power supply control strategy is determined based on the integration index: if the integration index is less than 30, it indicates that the power supply is operating well, and a low-level power supply control strategy is determined; if the integration index is between 30 and 70, it indicates that the power supply is operating moderately, and a medium-level power supply control strategy is determined; if the integration index is greater than 70, it indicates that the power supply is operating poorly, and a high-level power supply control strategy is determined.

[0045] Preferably, step S5, which involves dynamically adjusting and controlling the power supply according to the power supply adaptive control strategy, includes: The power supply is dynamically adjusted and controlled according to the low-level power regulation strategy, so that the output power of the power module is maintained at 90% of the rated power, the cooling fan is running at 40% of the rated speed, the heat sink heat dissipation area is maintained at 60% of the rated area, and the load distributor maintains the load distribution ratio as even. The power supply is dynamically adjusted and controlled according to the power supply level regulation strategy. The output power of the power module is gradually reduced to 70% of the rated power, with each adjustment step being 5% of the rated power and an adjustment interval of 2 seconds. The speed of the cooling fan is gradually increased to 60% of the rated speed, with each adjustment step being 10% of the rated speed and an adjustment interval of 3 seconds. The heat dissipation area of ​​the heat sink is gradually increased to 80% of the rated area, with each adjustment step being 10% of the rated area and an adjustment interval of 4 seconds. The load distributor dynamically adjusts the load distribution ratio based on real-time monitoring data, prioritizing the allocation to modules with lower loads, with each adjustment step being 5% of the total load and an adjustment interval of 5 seconds. According to the power supply low-level regulation strategy, the power supply is dynamically adjusted and controlled. The output power of the power module is immediately reduced to 50% of the rated power, with each adjustment step being 10% of the rated power and an adjustment interval of 1 second; the speed of the cooling fan is immediately increased to 100% of the rated speed, with each adjustment step being 20% ​​of the rated speed and an adjustment interval of 2 seconds; the heat dissipation area of ​​the heat sink is immediately increased to 100% of the rated area, with each adjustment step being 20% ​​of the rated area and an adjustment interval of 3 seconds; the load distributor urgently adjusts the load distribution ratio, giving priority to the backup module, with each adjustment step being 10% of the total load and an adjustment interval of 2 seconds.

[0046] In this embodiment of the invention, when the fusion index is less than 30, a low-level control strategy is triggered. The power management system (PMS) adjusts the output power of the power module to maintain it at 90% of the rated power (P_nominal). For example, if the rated power of the power supply is 500 watts, the target power is adjusted to 450 watts. The speed of the cooling fan is adjusted to 40% of the rated speed (RPM_nominal) via a speed controller. For example, if the rated speed is 3000 rpm, the target speed is adjusted to 1200 rpm. The heat dissipation area of ​​the heat sink is maintained at 60% of the rated area (A_nominal) via an adjustable heat sink device. For example, if the rated area is 100 square centimeters, the target area is adjusted to 60 square centimeters. The load distributor maintains a uniform load distribution ratio through a load balancing algorithm, meaning that each module receives the same proportion of the total load. When the fusion index is between 30 and 70, a medium-level control strategy is triggered. The output power of the power module is gradually reduced to 70% of the rated power via the PMS, with each adjustment step being 5% of the rated power and an adjustment interval of 2 seconds. For example, starting from 90% of the rated power, the power is reduced by 25 watts each time (assuming a rated power of 500 watts) until it reaches 350 watts. The cooling fan speed is gradually increased to 60% of the rated speed via a speed controller, with each adjustment step being 10% of the rated speed and an adjustment interval of 3 seconds. For example, starting from 40% of the rated speed (1200 rpm), the speed is increased by 300 rpm each time until it reaches 1800 rpm. The heat dissipation area of ​​the heat sink is gradually increased to 80% of the rated area via an adjustable heat sink device, with each adjustment step being 10% of the rated area and an adjustment interval of 4 seconds. For example, starting from 60% of the rated area (60 square centimeters), the area is increased by 10 square centimeters each time until it reaches 80 square centimeters. The load distributor dynamically adjusts the load distribution ratio based on real-time monitoring data, prioritizing allocation to modules with lower loads, with each adjustment step being 5% of the total load and an adjustment interval of 5 seconds. When the fusion index is greater than 70, a higher-level control strategy is triggered. The power module's output power is immediately reduced to 50% of its rated power via the PMS, with adjustments in increments of 10% of the rated power and intervals of 1 second. For example, starting from the current power, the power is reduced by 50 watts each time (assuming a rated power of 500 watts) until it reaches 250 watts. The cooling fan speed is immediately increased to 100% of its rated speed via the speed controller, with adjustments in increments of 20% of the rated speed and intervals of 2 seconds. For example, starting from the current speed, the speed is increased by 600 rpm each time (assuming a rated speed of 3000 rpm) until it reaches 3000 rpm. The heatsink's heat dissipation area is immediately increased to 100% of its rated area via the adjustable heatsink device, with adjustments in increments of 20% of the rated area and intervals of 3 seconds.For example, starting from the current area, increase by 20 square centimeters each time (assuming the rated area is 100 square centimeters) until it reaches 100 square centimeters. The load distributor urgently adjusts the load distribution ratio, prioritizing the allocation of load to the standby module, with each adjustment step being 10% of the total load and an adjustment interval of 2 seconds.

[0047] This specification also provides a power adaptive control system based on multimodal data fusion, used to execute the power adaptive control method based on multimodal data fusion as described above. The power adaptive control system based on multimodal data fusion includes: The mechanical support structure response model construction module is used to acquire internal mechanical structure data of the power supply, record power supply operating status data, analyze the stress response characteristics of the internal mechanical structure data of the power supply under different power supply operating status data, and establish a mechanical support structure response model. The electrical vibration impact mapping module is used to collect the power supply electrical parameters during the power supply operation process; perform parameter correlation detection on the power supply electrical parameters to obtain electrical parameter correlation data; and perform electrical vibration impact mapping on the mechanical support structure response model through the electrical parameter correlation data to generate structural vibration mapping data. The power ripple response mapping module is used to collect power load data during power supply operation; to perform time series labeling on the power load data and to monitor fluctuating loads to obtain fluctuating load sequence data; and to perform power ripple response mapping on the mechanical support structure response model using the fluctuating load sequence data to generate ripple response mapping data. The heat dissipation response mapping module is used to collect heat accumulation data during power supply operation, identify and record the concentrated heat accumulation area of ​​the power supply through the heat accumulation data, and perform heat dissipation response mapping on the mechanical support structure response model based on the concentrated heat accumulation area of ​​the power supply to generate heat dissipation response mapping data. The power adaptive control module is used to perform multimodal data fusion based on structural vibration mapping data, ripple response mapping data, and heat dissipation response mapping data, and to determine the power adaptive control strategy; and to dynamically adjust and control the power supply according to the power adaptive control strategy.

[0048] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0049] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A power supply adaptive control method based on multimodal data fusion, characterized in that, Includes the following steps: Step S1: Obtain data on the internal mechanical structure of the power supply and record the power supply operating status data; The stress response characteristics of the internal mechanical structure of the power supply under different power supply operating conditions were analyzed, and a response model of the mechanical support structure was established. Step S2: Collect the power supply electrical parameters during the power supply operation; perform parameter correlation detection on the power supply electrical parameters to obtain electrical parameter correlation data; use the electrical parameter correlation data to perform electrical vibration influence mapping on the mechanical support structure response model to generate structural vibration mapping data; Step S3: Collect power load data during power supply operation; perform time series labeling on the power load data and monitor fluctuating load to obtain fluctuating load sequence data; Power ripple response mapping is performed on the mechanical support structure response model using fluctuating load sequence data to generate ripple response mapping data. Step S4: Collect heat accumulation data during power supply operation, and determine and record the concentrated heat accumulation area of ​​the power supply based on the heat accumulation data; Based on the heat accumulation concentration area of ​​the power source, the heat dissipation response model of the mechanical support structure is mapped to generate heat dissipation response mapping data. Step S5: Perform multimodal data fusion based on structural vibration mapping data, ripple response mapping data, and heat dissipation response mapping data, and determine the power supply adaptive control strategy; The power supply is dynamically adjusted and controlled based on the power supply adaptive regulation strategy.

2. The power supply adaptive control method based on multimodal data fusion according to claim 1, characterized in that, Step S2, which involves parametric correlation detection of the power supply's electrical parameters, includes: Power supply electrical parameters are classified into voltage, current, and frequency according to parameter type; The voltage, current, and frequency are divided into multiple equal-length time segments in chronological order. Within each time period, the changing trends of voltage and current are analyzed synchronously to obtain voltage-current parameter correlation data; if both voltage and current rise or fall within the same time period, it is marked as a change in the same direction; if voltage rises while current falls, or voltage falls while current rises, it is marked as a change in opposite directions. Within each time period, the changing trends of frequency and current are analyzed synchronously to obtain frequency-current parameter correlation data; if both frequency and current rise or fall within the same time period, it is marked as a change in the same direction; if frequency rises while current falls, or frequency falls while current rises, it is marked as a change in opposite directions. Electrical parameter correlation data is determined based on pressure-current parameter correlation data and frequency-current parameter correlation data.

3. The power supply adaptive control method based on multimodal data fusion according to claim 2, characterized in that, Step S2 involves mapping the electrical vibration effects on the mechanical support structure response model using electrical parameter correlation data, including: Identify the input-end fixed support and the output-end insulating support in the mechanical support structure response model; The pressure-flow fluctuation in the pressure-flow parameter correlation data is evaluated to obtain the pressure-flow fluctuation degree value; Calculate the pressure flow vibration intensity factor for the fixed support at the input end based on the pressure flow fluctuation value; The pressure flow vibration intensity factor is mapped to the mechanical support structure response model, the degree of influence of the pressure flow factor is recorded, and the input vibration mapping data is generated. Frequency flow changes are detected in the frequency flow parameter correlation data to obtain the degree of frequency flow change value; The frequency current vibration intensity factor is calculated for the output end insulating bracket based on the frequency current variation value. The frequency-current vibration intensity factor is mapped to the mechanical support structure response model, the degree of influence of the frequency-current factor is recorded, and the output vibration mapping data is generated. The vibration mapping data at the input and output ends is mapped to the mechanical support structure response model, and the electrical vibration characteristics are detected to generate structural vibration mapping data.

4. The power supply adaptive control method based on multimodal data fusion according to claim 1, characterized in that, Step S3, which involves time-series labeling of the power load data and monitoring of fluctuating loads, includes: Based on the start time of power supply operation, the power load data is divided into 1-millisecond time intervals, with each time interval corresponding to a power load data point. Each power load data point is assigned a time stamp, which is in milliseconds, to record the specific position of the data point in the time series. Analyze each power load data point in the time series one by one, and calculate the power difference between adjacent power load data points; When the power difference exceeds 5% of the rated power of the power supply, the power load data point is identified as a power load fluctuation point, and the power value and corresponding time identifier of the power load data point are recorded. All identified power load fluctuation points are arranged in order of time to form fluctuating load sequence data; in the fluctuating load sequence data, the power value, time mark and power difference of each fluctuation point are recorded.

5. The power supply adaptive control method based on multimodal data fusion according to claim 1, characterized in that, Step S3, which maps the power ripple response of the mechanical support structure response model using fluctuating load sequence data, includes: The switching contact points of the mechanical support structure response model are identified by using fluctuating load sequence data, and their response characteristics under power fluctuations are detected; when load fluctuations occur in the fluctuating load sequence data, the contact resistance value of the switching contact points is recorded. For each load fluctuation point in the fluctuating load sequence data, if the contact resistance value shows a linear increasing trend when the load fluctuation point appears, then the relationship between the load fluctuation point and the switch contact point is determined to be a response synchronization relationship. The load fluctuation points are mapped to the switch contact area of ​​the mechanical support structure response model, and the power change amplitude, change frequency and duration of the switch contact area are recorded. Based on the power change amplitude and duration of the fluctuation point, the ripple response intensity of the switch contact point is classified and ripple response mapping data is generated. When the power change exceeds 15% of the rated power and lasts for more than 10 milliseconds, it is marked as high-intensity ripple response data; When the power change is between 10% and 15% of the rated power and the duration is between 5 and 10 milliseconds, it is marked as medium intensity ripple response data. Data with a power change of less than 10% of the rated power and a duration of less than 5 milliseconds is classified as low-intensity ripple response data.

6. The power supply adaptive control method based on multimodal data fusion according to claim 1, characterized in that, In step S4, heat accumulation data is collected during the power supply's operation. The concentrated heat accumulation areas of the power supply are identified and recorded based on the heat accumulation data, including: Temperature sensors are installed at the locations of the heat sink, power module, capacitors, and transformer inside the power supply; each temperature sensor collects temperature data at fixed time intervals of 1 second. For each sensor location, the difference between the current power supply operating temperature and the previous power supply operating temperature is used to obtain the power supply operating temperature difference. Identify areas of sustained temperature rise in the power supply operating temperature difference and mark them as potential areas of concentrated heat accumulation; If the temperature difference of the power supply operating temperature in the potential heat accumulation zone exceeds the preset temperature rise value, where the preset temperature rise value is 2 degrees Celsius per second, it is determined to be a heat accumulation zone of the power supply.

7. The power supply adaptive control method based on multimodal data fusion according to claim 1, characterized in that, Step S4, which maps the heat dissipation response of the mechanical support structure response model based on the concentrated heat accumulation area of ​​the power source, includes: Measure the temperature change rate and peak temperature of each heat accumulation zone, and classify the heat accumulation zones into heat dissipation demand levels based on the temperature change rate and peak temperature of each zone: If the rate of temperature change exceeds 3 degrees Celsius per second or the peak temperature exceeds 80 degrees Celsius, it is classified as a high heat dissipation requirement area: If the temperature change rate is between 1 and 3 degrees Celsius per second or the peak temperature is between 60 and 80 degrees Celsius, it is classified as a medium heat dissipation requirement zone. If the rate of temperature change is less than 1 degree Celsius per second or the peak temperature is below 60 degrees Celsius, it is classified as a low heat dissipation requirement zone. Identify heat sinks, thermal pads, ventilation holes, and heat dissipation channels in the mechanical support structure response model to obtain the heat dissipation hardware structure characteristics; The heat dissipation demand regions of high heat dissipation demand region, medium heat dissipation demand region, and low heat dissipation demand region are respectively matched with the heat dissipation hardware structure features to generate heat dissipation response mapping data.

8. The power supply adaptive control method based on multimodal data fusion according to claim 1, characterized in that, Step S5 involves multimodal data fusion based on structural vibration mapping data, ripple response mapping data, and heat dissipation response mapping data, and determining the power supply adaptive control strategy, including: Frequency domain analysis was performed on the structural vibration mapping data to extract vibration frequency components and identify the primary and secondary vibration frequencies. Based on the primary and secondary vibration frequencies, the structural layout vibration impact was assessed on the mechanical support structure response model to obtain structural layout vibration impact data. Time-domain analysis was performed on the ripple response mapping data to extract the ripple amplitude and ripple frequency; based on the ripple amplitude and ripple frequency, the component ripple influence was evaluated on the mechanical support structure response model to obtain component ripple influence data; Heat flow analysis is performed on the heat dissipation response mapping data to extract the temperature change rate and heat dissipation demand level; the heat dissipation structure impact assessment is performed on the mechanical support structure response model based on the temperature change rate and heat dissipation demand level to obtain heat dissipation structure impact data. The vibration impact data of structural layout, the ripple impact data of components, and the impact data of heat dissipation structure are multimodally weighted and fused according to a weight of 4:3:3 to obtain multimodal fused data; Calculate the fusion index of the multimodal fusion data. If the fusion index is less than 30, it indicates that the power supply is in good condition and a low-level power supply control strategy is determined. If the fusion index is between 30 and 70, it indicates that the power supply is in a moderate condition and a medium-level power supply control strategy is determined. If the fusion index is greater than 70, it indicates that the power supply is in a poor condition and a high-level power supply control strategy is determined.

9. The power supply adaptive control method based on multimodal data fusion according to claim 8, characterized in that, Step S5, which involves dynamically adjusting and controlling the power supply according to the power supply adaptive control strategy, includes: The power supply is dynamically adjusted and controlled according to the low-level power regulation strategy, so that the output power of the power module is maintained at 90% of the rated power, the cooling fan is running at 40% of the rated speed, the heat sink heat dissipation area is maintained at 60% of the rated area, and the load distributor maintains the load distribution ratio as even. The power supply is dynamically adjusted and controlled according to the power supply level regulation strategy. The output power of the power module is gradually reduced to 70% of the rated power, with each adjustment step being 5% of the rated power and an adjustment interval of 2 seconds. The speed of the cooling fan is gradually increased to 60% of the rated speed, with each adjustment step being 10% of the rated speed and an adjustment interval of 3 seconds. The heat dissipation area of ​​the heat sink is gradually increased to 80% of the rated area, with each adjustment step being 10% of the rated area and an adjustment interval of 4 seconds. The load distributor dynamically adjusts the load distribution ratio based on real-time monitoring data, prioritizing the allocation to modules with lower loads, with each adjustment step being 5% of the total load and an adjustment interval of 5 seconds. According to the power supply low-level regulation strategy, the power supply is dynamically adjusted and controlled. The output power of the power module is immediately reduced to 50% of the rated power, with each adjustment step being 10% of the rated power and an adjustment interval of 1 second; the speed of the cooling fan is immediately increased to 100% of the rated speed, with each adjustment step being 20% ​​of the rated speed and an adjustment interval of 2 seconds; the heat dissipation area of ​​the heat sink is immediately increased to 100% of the rated area, with each adjustment step being 20% ​​of the rated area and an adjustment interval of 3 seconds; the load distributor urgently adjusts the load distribution ratio, giving priority to the backup module, with each adjustment step being 10% of the total load and an adjustment interval of 2 seconds.

10. A power supply adaptive control system based on multimodal data fusion, characterized in that, For executing the power adaptive regulation method based on multimodal data fusion as described in claim 1, the power adaptive regulation system based on multimodal data fusion includes: The mechanical support structure response model construction module is used to acquire internal mechanical structure data of the power supply, record power supply operating status data, analyze the stress response characteristics of the internal mechanical structure data of the power supply under different power supply operating status data, and establish a mechanical support structure response model. The electrical vibration impact mapping module is used to collect the power supply electrical parameters during the power supply operation process; perform parameter correlation detection on the power supply electrical parameters to obtain electrical parameter correlation data; and perform electrical vibration impact mapping on the mechanical support structure response model through the electrical parameter correlation data to generate structural vibration mapping data. The power ripple response mapping module is used to collect power load data during power supply operation; to perform time series labeling on the power load data and to monitor fluctuating loads to obtain fluctuating load sequence data; and to perform power ripple response mapping on the mechanical support structure response model using the fluctuating load sequence data to generate ripple response mapping data. The heat dissipation response mapping module is used to collect heat accumulation data during power supply operation, identify and record the concentrated heat accumulation area of ​​the power supply through the heat accumulation data, and perform heat dissipation response mapping on the mechanical support structure response model based on the concentrated heat accumulation area of ​​the power supply to generate heat dissipation response mapping data. The power adaptive control module is used to perform multimodal data fusion based on structural vibration mapping data, ripple response mapping data, and heat dissipation response mapping data, and to determine the power adaptive control strategy; and to dynamically adjust and control the power supply according to the power adaptive control strategy.

Images