Explosion-proof robot MOS tube optimization driving control method and system based on intelligent algorithm

Abstract

The invention relates to the technical field of industrial robots and intelligent control, and discloses an explosion-proof robot MOS tube optimization driving control method and system based on an intelligent algorithm. The method comprises the following steps: acquiring multi-point temperature monitoring data of a power device to construct a virtual thermal gradient model; according to the virtual thermal gradient model, operation boundary parameter optimization calculation is carried out, and a dynamic safety threshold value is determined; acquiring real-time output state data of the driving system and generating a preliminary steady-state control strategy by combining the real-time output state data with the dynamic safety threshold; monitoring load efficiency feedback data, and triggering multi-target optimization of driving parameters by taking the dynamic safety threshold as a boundary constraint when an efficiency decline feature is identified, so as to obtain a to-be-verified driving parameter set; performing online prediction verification on the to-be-verified driving parameter set to generate a verification result identifier; and according to the verification result identifier, performing configuration updating or configuration keeping processing on a control register of the driving system during operation. The method effectively improves the system stability.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of industrial robots and intelligent control technology, and particularly relates to a MOS tube optimization driving control method and system for an explosion-proof robot based on an intelligent algorithm. BACKGROUND

[0002] At present, with the continuous improvement of industrial automation level, explosion-proof robots are increasingly widely used in high-risk environments such as flammable and explosive environments, and bear key tasks such as inspection, operation and emergency rescue. The stability and reliability of the driving system of the explosion-proof robot are directly related to the production safety and equipment life.

[0003] In one prior art, the driving control of the explosion-proof robot often relies on a general SCADA (Supervisory Control And Data Acquisition) system to perform macro state monitoring and logic scheduling. This traditional control architecture usually uses fixed driving parameters or static safety thresholds to manage core power devices (such as MOS tubes), that is, only when the temperature or current exceeds the preset hard red line, an alarm or shutdown protection is triggered, and there is a lack of dynamic perception and fine control capability for the microscopic running state of the device. However, power devices will inevitably generate heat accumulation under high-frequency switching action, and this thermal effect and the working efficiency of the device have a complex nonlinear coupling relationship. When the robot performs a high-load or long-endurance task, if only static thresholds are used for management, the vicious cycle of "temperature rise leading to increased internal resistance, increased internal resistance causing efficiency decline, and efficiency decline further exacerbating temperature rise" cannot be broken in time. This neglect of the thermal-efficiency coupling mechanism can cause performance degradation of the driving components before reaching the hard alarm threshold, seriously affecting the response speed and running stability of the robot in complex working conditions.

[0004] Therefore, the prior art has the technical problem of insufficient reliability of the driving system. SUMMARY

[0005] The present application provides a MOS tube optimization driving control method and system for an explosion-proof robot based on an intelligent algorithm to solve the technical problem of insufficient reliability of the driving system in the prior art.

[0006] In a first aspect, to solve the above technical problem, the present application provides a MOS tube optimization driving control method for an explosion-proof robot based on an intelligent algorithm, comprising:

[0007] Obtaining multi-point temperature monitoring data of a power device and performing thermal field gradient feature extraction processing to construct a virtual thermal gradient model;

[0008] According to the virtual thermal gradient model, performing running boundary parameter optimization calculation to determine a dynamic safety threshold;

[0009] Real-time output state data of the driving system is acquired, and the working mode switching matching is performed in combination with the dynamic safety threshold to generate a preliminary steady-state control strategy.

[0010] Load efficiency feedback data in dynamic operation is monitored, and when the efficiency recession feature is identified, the driving parameter multi-objective optimization is triggered with the dynamic safety threshold as the boundary constraint to obtain a to-be-verified driving parameter set.

[0011] The to-be-verified driving parameter set is subjected to online prediction verification based on the virtual thermal gradient model to generate a verification result identifier.

[0012] According to the verification result identifier, the control register of the driving system is subjected to runtime configuration update or configuration retention processing.

[0013] In a second aspect, the present application provides an explosion-proof robot MOS tube optimization driving control system based on an intelligent algorithm, comprising:

[0014] A thermal field perception and modeling module is configured to acquire multi-point temperature monitoring data of a power device, perform thermal field gradient feature extraction processing, and construct a virtual thermal gradient model.

[0015] A threshold optimization module is configured to perform operating boundary parameter optimization calculation based on the virtual thermal gradient model to determine a dynamic safety threshold.

[0016] A strategy initial matching module is configured to acquire real-time output state data of the driving system, and perform working mode switching matching in combination with the dynamic safety threshold to generate a preliminary steady-state control strategy.

[0017] A parameter optimization module is configured to monitor load efficiency feedback data in dynamic operation, and when the efficiency recession feature is identified, the driving parameter multi-objective optimization is triggered with the dynamic safety threshold as the boundary constraint to obtain a to-be-verified driving parameter set.

[0018] A simulation verification module is configured to perform online prediction verification on the to-be-verified driving parameter set based on the virtual thermal gradient model to generate a verification result identifier.

[0019] An execution control module is configured to perform runtime configuration update or configuration retention processing on the control register of the driving system according to the verification result identifier.

[0020] Compared with the prior art, the present application has the following beneficial effects:

[0021] (1) This invention constructs a virtual thermal gradient model by acquiring multi-point temperature monitoring data of power devices and uses a genetic algorithm to periodically optimize and calculate dynamic safety thresholds. This mechanism breaks the limitation of traditional SCADA systems that rely solely on single-point static temperature thresholds for passive protection. It can quantify the thermal field distribution trend and operational safety boundary of power devices in the explosion-proof sealed space in real time. This achieves refined active thermal management of the explosion-proof robot drive system and maximizes the performance potential of power devices while ensuring safety.

[0022] (2) This invention monitors the load efficiency feedback during dynamic operation in real time and triggers multi-objective optimization of drive parameters based on particle swarm optimization when efficiency degradation characteristics are detected. This method is different from simple frequency reduction and current limiting protection. Instead, it targets the device characteristic drift caused by thermal effects and intelligently finds the optimal drive waveform parameters that can minimize switching losses. This effectively breaks the vicious cycle of "temperature rise leading to efficiency decline, and efficiency decline further aggravating temperature rise" and significantly improves the energy efficiency ratio and operational stability of the robot under high load continuous operation.

[0023] (3) This invention performs virtual prediction verification by inputting the set of driving parameters to be verified into an online simulation model, and strictly determines whether to solidify the strategy and update the configuration based on the status flag of the verification result. This digital twin closed-loop logic of "virtual verification first, then physical execution" ensures that only the optimized parameters confirmed by the model will be written into the control register, completely avoiding the risk of system crash caused by "bad parameters" entering the system due to algorithm convergence problems. Thus, it solves the problem of insufficient robustness of adaptive control strategies in the prior art and realizes the safe self-healing and continuous evolution of the driving control strategy. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the process for optimizing the drive control of the MOS transistor of an explosion-proof robot based on intelligent algorithms, provided in the first embodiment of the present invention.

[0025] Figure 2 This is a schematic diagram of the structure of the explosion-proof robot MOS transistor optimized drive control system based on intelligent algorithms provided in the second embodiment of the present invention. Detailed Implementation

[0026] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] Reference Figure 1The first embodiment of the present invention provides an optimized drive control method for MOSFETs of explosion-proof robots based on intelligent algorithms, including the following steps:

[0028] S11: Acquire multi-point temperature monitoring data of power devices, extract thermal gradient features, and construct a virtual thermal gradient model.

[0029] S12, Based on the virtual thermal gradient model, perform optimization calculations of the running boundary parameters to determine the dynamic safety threshold;

[0030] S13, acquire the real-time output status data of the drive system, and combine it with the dynamic safety threshold to perform working mode switching matching and generate a preliminary steady-state control strategy.

[0031] S14, monitor the load efficiency feedback data during dynamic operation, and when efficiency degradation characteristics are identified, trigger multi-objective optimization of driving parameters with the dynamic safety threshold as the boundary constraint to obtain the set of driving parameters to be verified;

[0032] S15, perform online prediction verification on the set of driving parameters to be verified based on the virtual thermal gradient model, and generate a verification result identifier;

[0033] S16, based on the verification result identifier, perform runtime configuration update or configuration retention processing on the control register of the drive system.

[0034] It is particularly important to note that, considering the high computational demands of this invention, including virtual thermal gradient model construction, global optimization using genetic algorithms, local iteration using particle swarm optimization, and online differential equation solving, this embodiment preferably employs a heterogeneous dual-core processing platform in its hardware architecture to ensure that the explosion-proof robot can perform the aforementioned intelligent calculations in real-time and stably without watchdog resets or control interruption losses while executing millisecond-level low-level motor drive control. Specifically, the core processor of the drive controller adopts a dual-core system-on-a-chip containing a digital signal processing core and a general-purpose processing core, such as an ARM Cortex-M7 core combined with an M4 core, or a digital signal processor combined with a field-programmable gate array architecture. The main core focuses on executing high-frequency pulse width modulation refresh, proportional-integral-derivative control of the current and speed loops, and hardware overcurrent protection logic, possessing the highest interrupt priority to ensure the absolute real-time performance of the low-level drive. The coprocessor, acting as a computational acceleration engine, runs in a low-priority background thread of the real-time operating system, specifically responsible for executing model construction, parameter optimization, and online verification tasks in S11 to S16. The two exchange dynamic thresholds and waveform parameters through shared memory or dual-port random access memory. This hardware and software co-design completely resolves the conflict of computing resources between complex intelligent algorithms and high real-time drive control, ensuring the engineering feasibility of the solution.

[0035] In step S11, multi-point temperature monitoring data of the power device is acquired, and thermal gradient feature extraction processing is performed to construct a virtual thermal gradient model, including:

[0036] Multi-channel temperature readings of the power device surface and heat dissipation area are collected to obtain the multi-point temperature monitoring data;

[0037] The multi-point temperature monitoring data are subjected to spatial location mapping processing to obtain a temperature spatial distribution matrix;

[0038] Based on the temperature spatial distribution matrix, the heat flux density and temperature rise rate are calculated, and the thermal field gradient features are extracted.

[0039] The virtual thermal gradient model is established by performing multidimensional numerical fitting based on the thermal field gradient characteristics.

[0040] In one implementation, for the acquisition of the multi-point temperature monitoring data, this embodiment constructs a multi-channel temperature sensing array covering the core power devices on the sealed drive circuit board of the explosion-proof robot. The system selects high-sensitivity NTC thermistors as sensors, and its circuit design conforms to the GB3836.4 standard. Spark risks are isolated by encapsulation or metal explosion-proof housings to ensure long-term reliability in high-risk environments. These sensors are deployed in the heat source center area, heat diffusion edge area, and heat dissipation end area of ​​the power devices, respectively.

[0041] It should be noted that, in order to resolve the contradiction between the thermal conduction hysteresis of the NTC sensor and the requirements of transient protection, this embodiment adopts an electrothermal dual-frequency sampling architecture. This architecture includes a high-frequency electrical channel synchronized with the PWM carrier frequency of the motor control and a low-frequency thermal channel set based on the device's thermal time constant. The high-frequency electrical channel acquires the instantaneous current value in real time and performs current square time integration (IS) in each PWM interrupt service routine. 2 The accumulation operation of t) is used to capture transient heat energy; the low-frequency thermal channel collects surface temperature at a lower frequency. At the end of each thermal sampling cycle, the system reads and resets the integral accumulation value of the high-frequency channel, uses it as a feedforward compensation amount combined with the current temperature reading, and generates the dynamically corrected multi-point temperature monitoring data. Regarding the determination of the sampling cycle of the low-frequency thermal channel, this embodiment abandons subjective empirical estimation and instead determines it based on the Nyquist sampling theorem combined with thermal transient testing.

[0042] Specifically, during the R&D phase, a step power excitation is applied to the device, and the thermal impedance response curve is recorded to extract the minimum thermal time constant τ; the sampling frequency is set to 5 to 10 times the cutoff frequency corresponding to this time constant (i.e., This ensures that the temperature rise inflection point signal during the heat conduction process can be captured without distortion.

[0043] In one implementation, for the spatial location mapping process, this embodiment first establishes a two-dimensional Cartesian coordinate system with the geometric center of the power device as the origin, and pre-stores the precise physical coordinates of each sensor on the circuit board. To transform the finite discrete sensor readings into a continuous temperature spatial distribution matrix, this embodiment employs an inverse distance-weighted interpolation method. This method is based on the physical law that heat diffusion attenuates with distance. For any grid point in the matrix without a sensor, the formula for estimating its temperature is:

[0044]

[0045] Where N is the total number of sensors, T i Let d be the measured temperature of the i-th sensor. i Let be the Euclidean distance from the grid point to the i-th sensor, and p be the power parameter. Through this step, the system can obtain a high-resolution dense temperature matrix.

[0046] It is particularly important to note that the value of the power parameter p is determined through an offline finite element simulation calibration process. A three-dimensional thermal simulation model of the drive board is constructed to generate an ideal temperature field distribution under rated load. Using the simulated temperature at the sensor coordinate points as input, different candidate p values ​​(e.g., increasing from 1.0 to 3.0 in steps of 0.1) are used for interpolation reconstruction. The root mean square error (RMSE) between each reconstruction matrix and the ideal temperature field is calculated. The p value corresponding to the minimum RMSE (usually close to 2, conforming to the inverse square law of energy) is selected as the system preset parameter.

[0047] In one implementation, for the extraction of the thermal field gradient features, this embodiment performs numerical differential calculations in both the spatial and temporal domains to quantify the flow and accumulation of heat. In the spatial domain, this embodiment uses the Sobel operator or the central difference method to calculate the partial derivatives of the temperature matrix along the horizontal and vertical axes, thereby obtaining the spatial gradient vector. The magnitude of this gradient vector characterizes the intensity of the local heat flux density, and its direction characterizes the dominant path of heat diffusion. In the temporal domain, this embodiment combines the temperature matrices of the current and previous moments to calculate the first derivative in the time dimension. This derivative is the temperature rise rate, directly reflecting the dynamic balance between the current heat generation power and the heat dissipation capacity.

[0048] In one implementation, for establishing the virtual thermal gradient model, this embodiment constructs a spatiotemporally coupled polynomial fitting model. The model function is defined as:

[0049] T model (x,y,t)=a(t)x 2 +b(t)y 2 +c(t)xy+d(t)x+e(t)y+f(t)

[0050] Here, a(t) to f(t) are coefficients that change dynamically with time. It should be noted that these coefficients are not purely mathematical fitting parameters without physical meaning, but rather deeply integrate the aforementioned extracted thermal field gradient features.

[0051] Specifically, this embodiment uses a constraint-based solution method, directly assigning the coefficients d(t) and e(t) of the linear term in the polynomial to the spatial gradient value at the center point (0,0). and This locks in the tilt of the model surface at the center of the heat source, giving the model a clear physical meaning of heat flow. After this constraint assignment, the coefficients d(t) and e(t) are determined by real-time monitoring data and no longer participate in the subsequent recursive least squares update process. For the remaining coefficients a(t), b(t), c(t), and f(t), this embodiment abandons the traditional static least squares method and instead uses recursive least squares (RLS) with a forgetting factor for real-time updates. This algorithm aims to minimize the sum of squared errors between the current temperature matrix and the model's predicted values, iteratively correcting the coefficient vector. The forgetting factor λ in the RLS algorithm is linked to the thermal time constant τ of the power device (e.g., ...). T s (For the update cycle). This algorithm mechanism gives the evolution of model coefficients the same thermal inertia as the physical device. That is, the current coefficient update depends not only on the current temperature sampling, but also on the smoothing constraint of the historical thermal state, thus mathematically realizing the organic coupling of spatial distribution and time response.

[0052] In step S12, based on the virtual thermal gradient model, the operational boundary parameters are optimized to determine the dynamic safety threshold, including:

[0053] Based on the virtual thermal gradient model, thermal accumulation trend extrapolation is performed to obtain future thermal field evolution data;

[0054] A global safety fitness function is constructed based on the future thermal field evolution data. The genetic algorithm is used to iteratively optimize the preset initial operating boundary parameters to obtain the optimal combination of boundary parameters.

[0055] Based on the optimal combination of boundary parameters, a safety boundary quantization mapping is performed to obtain the dynamic safety threshold.

[0056] In one implementation, for the extrapolation of the heat accumulation trend, this embodiment uses the Taylor series expansion method for short-term prediction based on the virtual thermal gradient model established in S11 and its included first-order time-dimension derivative. Specifically, the system extracts the model coefficient vector and its rate of change at the current moment, calculates the predicted coefficient vector for future moments, and reconstructs the temperature spatial distribution matrix for future moments based on the predicted coefficient vector. This matrix is ​​the future thermal field evolution data.

[0057] It should be noted that the prediction time window in the above calculation process is a key preset parameter, and its value is determined based on the thermal capacity characteristics of the power device. The thermal time constant of the device is obtained through thermal transient testing, and the prediction time window is set to 10% to 20% of this thermal time constant. This setting is based on the fact that within this time range, heat conduction is approximately linear, and linear extrapolation using the first derivative can ensure that the prediction error is within an acceptable range, for example, less than 2%, thus achieving a balance between computational efficiency and prediction accuracy.

[0058] In one implementation, regarding the construction of the global safety fitness function, this embodiment aims to find the optimal balance between performance output and thermal safety risk. The function is defined as a performance gain term minus a risk penalty term. Specifically, the mathematical expression is:

[0059]

[0060] Where x represents the operating boundary parameters to be optimized, including the maximum switching frequency and peak current limit; P out (x) represents the theoretical output power under this parameter, P rated Rated power; T max (x) represents the predicted maximum temperature calculated based on the aforementioned future thermal field evolution data; T limit w1 represents the physical tolerance temperature of the device material, for example, 150 degrees Celsius; w2 and w1 are weighting coefficients.

[0061] It is worth noting that the temperature normalization factor T in the formula... scale It is a key parameter for controlling the sensitivity of the penalty term. Its value is usually set to 10% to 20% of the difference between the device's maximum operating temperature and the ambient temperature, for example, 10 degrees Celsius to 20 degrees Celsius. This ensures that the penalty value increases exponentially when the predicted temperature approaches the limit, thereby forming an effective soft constraint boundary.

[0062] Furthermore, this implementation employs an objective mapping method based on a lookup table for determining the weighting coefficients w1 and w2. The system pre-defines a task mode and weighting mapping table in its firmware. The construction of this table is based on rigorous offline bench testing and multi-objective Pareto optimality analysis. On a standard test bench, for each defined task mode, the system iterates through the combinations of weighting coefficients between zero and one with a preset step size (e.g., 0.1), recording the corresponding steady-state output power and temperature rise curves. For the "emergency rescue mode," the weighting combination that maximizes output power without triggering hard safety protection at the device junction temperature (e.g., w1 = 0.8, w2 = 0.2) is selected from the test data as the fixed parameters, prioritizing power response. For the "routine inspection mode," the weighting combination that minimizes the temperature rise rate and meets the minimum cruise speed requirement (e.g., w1 = 0.3, w2 = 0.7) is selected as the fixed parameters, prioritizing equipment lifespan. During operation, the system reads the current task status register and automatically matches the corresponding weighting coefficients, thus achieving objective and dynamic optimization guidance.

[0063] In one implementation, real-number encoding is used for the iterative optimization process of the genetic algorithm. The preset initial operating boundary parameters are derived from the official technical specifications provided by the power device manufacturer. This specification data, including the absolute maximum rated current and the maximum switching frequency, is stored in the system's read-only memory.

[0064] It is particularly important to note that, to address the computational limitations of embedded controllers and prevent blockage of the main control loop, this embodiment employs a state machine-based time-sliced ​​asynchronous computation mechanism in its software architecture. Each population iteration of the genetic algorithm is decomposed into several small computational subtasks, such as calculating the fitness of a single individual or performing a crossover mutation. The system utilizes the idle task hook function of the real-time operating system to execute only one subtask in each idle time slice, and uses static variables to store the algorithm context. The final result is output only after all subtasks have been executed. The asynchronous sliced ​​tasks of the genetic algorithm are managed through task priority in the RTOS; for example, fitness calculation is performed once per idle time slice, with the time consumption controlled within 10μs. The online simulation uses the explicit Euler method, utilizing the SIMD instructions of the DSP for parallel computation, ensuring that the simulation step is completed within 0.1ms.

[0065] Regarding the algorithm parameters, the population size was set to 20 to 30, and the maximum number of iterations was set to 10 to 20. These two values ​​were not chosen arbitrarily, but were determined through convergence efficiency inflection point analysis in offline simulations. Specifically, a digital twin model of the driving system was built on a PC, the genetic algorithm was run and iterated through different population sizes and iteration numbers, and a "fitness gain - computation time" curve was plotted. The parameter combination corresponding to the inflection point where the curve slope drops sharply was selected. This combination represents the engineering approximate optimal solution that achieves the maximum fitness improvement with minimal computational resource consumption. The reason for choosing a genetic algorithm instead of a particle swarm optimization algorithm is that S12 aims to find the global safe boundary, i.e., the discrete feasible region in the solution space, and the genetic algorithm has advantages in global search and preventing getting trapped in local optima.

[0066] In one implementation, regarding the safety boundary quantization mapping, this embodiment directly maps the optimal combination of boundary parameters output by the genetic algorithm to the limiting threshold of the control system. Specifically, the system writes the optimal maximum switching frequency into the frequency upper limit register of the pulse width modulation generator and the optimal peak current into the reference voltage digital-to-analog converter register of the current protection comparator. These two physical quantities constitute the dynamic safety threshold. In this way, the system transforms the abstract thermal field prediction result into specific circuit electrical limits, realizing proactive thermal safety protection.

[0067] In step S13, real-time output status data of the drive system is acquired, and operating mode switching is matched in conjunction with the dynamic safety threshold to generate a preliminary steady-state control strategy, including:

[0068] The output current and output voltage values ​​of the drive system are collected to obtain the real-time output status data;

[0069] Based on the real-time output status data and the dynamic safety threshold, safety margin calculation and risk classification are performed to obtain the operational risk level label;

[0070] Based on the operational risk level label, a preset control strategy library is retrieved, and mode switching matching is performed to obtain the target working mode;

[0071] The target operating mode is analyzed for control logic to generate the preliminary steady-state control strategy, and the preliminary steady-state control strategy is written into the control register of the drive system.

[0072] In one implementation, for the acquisition of the real-time output status data, this embodiment deploys a high-frequency Hall current sensor and a precision resistor voltage divider network in the motor drive circuit of the explosion-proof robot. The system reads the bus voltage and phase current respectively through the independent analog-to-digital conversion channel of the microcontroller at a sampling rate synchronized with the pulse width modulation carrier frequency. To eliminate the interference of switching noise, this embodiment explicitly adopts a moving average filtering method. In specific implementation, the system opens a first-in-first-out queue in memory as a sliding window, pushes new data acquired each time into the queue and removes the oldest data, and calculates the arithmetic mean of the data in the queue as the current sampled value.

[0073] It should be noted that the size of this sliding window is determined based on the ratio of the pulse width modulation carrier frequency to the analog-to-digital conversion sampling frequency. For example, if the carrier frequency is 20 kHz and the sampling frequency is 200 kHz, the window size is set to 10. This setting is based on ensuring that the window exactly covers a complete switching cycle, thereby physically canceling out the current ripple component caused by inductor charging and discharging, resulting in a smooth and true DC component. These two filtered values ​​and the calculated instantaneous power together constitute the real-time output status data.

[0074] It should be noted that the safety margin calculation and risk classification processing are implemented based on the normalized distance algorithm. In this embodiment, the dynamic safety threshold determined in S12 is used as the safety boundary under the current operating condition. The system calculates the safety margin of real-time data relative to this boundary, using the formula: one minus the ratio of real-time current to the upper limit of dynamic current. Subsequently, the system compares the calculated safety margin with preset classification thresholds. The preset classification thresholds include warning thresholds and critical thresholds. These two thresholds are determined based on the safe operating area curves of power devices and statistical analysis of long-term reliability test data. For example, a warning threshold set to 0.2 represents a 20% dynamic response margin, and a critical threshold set to 0.05 represents an imminent triggering of physiological limits. If the safety margin is greater than the warning threshold, it is judged as low risk; if it is between the two, it is judged as medium risk; and if it is less than the critical threshold, it is judged as high risk. This judgment result is the operating risk level label.

[0075] In one implementation, for the mode switching matching, this embodiment pre-constructs a preset control strategy configuration library stored in non-volatile memory. This library is organized using a key-value pair structure, where the key is the operating risk level label, and the value is the corresponding predefined operating mode. It is particularly important to note that the construction process of this strategy library is not based on experience, but rather on rigorous offline multi-condition bench testing. Specifically, the construction method involves simulating different thermal loads and load torques on a test bench to find the optimal control boundary of the system under different safety margins. For the "low-risk" label, the parameter set that achieves maximum dynamic response without triggering overcurrent protection is determined through testing and defined as the aggressive response mode; for the "high-risk" label, the minimum power consumption parameter set that maintains the device junction temperature from rising is determined through testing and defined as the flexible derating mode. These experimentally verified parameter sets are structured and stored, thus completing the construction of the strategy library. During system runtime, the target operating mode can be determined by retrieving the library based on the current label.

[0076] It is worth noting that, regarding the control logic parsing and register writing, this embodiment transforms the abstract target operating mode into specific low-level control parameters. Specifically, the system reads the data structure corresponding to the target operating mode, which contains a set of offline-calibrated proportional-integral-derivative controller coefficients, pulse width modulation dead time, and maximum duty cycle limit. To ensure the validity of these parameters, the offline calibration process employs a step response test method. In a laboratory environment, for each operating mode, a step current command is injected into the drive system, and the system's overshoot and settling time are recorded. By iteratively adjusting the controller coefficients, the current overshoot in the flexible derating mode is less than 5% and oscillation-free to ensure thermal stability; while the rise time in the aggressive response mode is less than one millisecond to ensure dynamic response. The determined optimal coefficients are then stored in the strategy library. The system directly writes these parsed specific parameters into the control register of the drive system, completing the loading and execution of the preliminary steady-state control strategy, thereby realizing a logical closed loop from macroscopic risk assessment to microscopic hardware control.

[0077] In step S14, load efficiency feedback data during dynamic operation is monitored, and when efficiency degradation characteristics are identified, multi-objective optimization of driving parameters is triggered using the dynamic safety threshold as a boundary constraint to obtain a set of driving parameters to be verified, including:

[0078] The system collects real-time input voltage, input current, output voltage, and output current of the drive system, performs power calculation and ratio calculation, and obtains real-time energy efficiency ratio data.

[0079] The real-time energy efficiency ratio data is compared with a preset degradation threshold;

[0080] If the real-time energy efficiency ratio data is lower than the preset degradation threshold, feature extraction processing is performed to obtain efficiency degradation features.

[0081] The dynamic safety threshold is set as the search space boundary of the particle swarm optimization algorithm, and a local fitness function is constructed based on the efficiency decay characteristics.

[0082] The optimal driving waveform parameters are obtained by performing constrained iterative optimization using the particle swarm optimization algorithm.

[0083] The optimal driving waveform parameters are encapsulated to obtain the set of driving parameters to be verified.

[0084] In one implementation, for acquiring the real-time energy efficiency ratio data, this embodiment fully utilizes the existing hardware sampling circuit and vector control algorithm data within the explosion-proof robot motor controller. The system collects the DC bus voltage and bus current in real time, and obtains the input power through product calculation. For the calculation of output power, in order to adapt to the embedded low-level driver environment and ensure high real-time performance, this embodiment abandons the traditional power frequency measurement method and instead adopts instantaneous power theory based on coordinate transformation.

[0085] Specifically, by utilizing the three-phase current and voltage acquired in real time in the motor control algorithm (FOC), and through Clarke and Park transformations, the d-axis voltage v in the synchronous rotating coordinate system is obtained. d q-axis voltage v q and the corresponding d-axis current i d q-axis current i q According to the formula:

[0086] P out =1.5(v) d i d +v q i q )

[0087] The output mechanical power is calculated directly within the digital signal processor (DSP). The instantaneous power loss data is the difference between the input power and the output power. The system further calculates the real-time energy efficiency ratio. The time window of this calculation process is strictly synchronized with the pulse width modulation control cycle in the preliminary steady-state control strategy to eliminate calculation errors caused by asynchrony between input and output data.

[0088] It should be noted that the preset degradation threshold is determined based on statistical process control principles, rather than a fixed value. Specifically, the system maintains a sliding window containing the historical energy efficiency mean and standard deviation for the most recent several operating cycles, and sets the degradation threshold to the historical energy efficiency mean minus three times the standard deviation. When the real-time energy efficiency ratio data falls below this threshold, it indicates a statistically significant decrease in system efficiency, rather than random fluctuation. At this point, the system extracts the efficiency degradation feature, which includes the current energy efficiency deviation value and the rate of energy efficiency decline, used to quantify the urgency of optimization. It should be noted that the sliding window size of the statistical process control is adapted to the system sampling frequency; for example, the window capacity is set to 100 points, corresponding to data within 1 second, and the update cycle is strictly synchronized with the control cycle (e.g., 10ms) to avoid lag.

[0089] In one implementation, for multi-objective optimization of the driving parameters, this embodiment initiates a particle swarm optimization algorithm to find driving waveform parameters that can reduce losses. To address the limitation of computing power in embedded systems, the particle swarm optimization algorithm is not executed in the high-frequency interrupt service routine of motor control, but rather runs in a low-priority background task of the embedded operating system, utilizing idle time slices of the control cycle for asynchronous computation. During algorithm initialization, the currently used driving parameters in the preliminary steady-state control strategy are used as the center position of the initial particle swarm, generating an initial population with small perturbations around it, thereby accelerating convergence. It is particularly important to note that, to avoid the risk of motor squealing or control loop instability that may be caused by dynamically adjusting the switching frequency, the particle position vector in this embodiment consists only of the dead time and gate drive current—two key parameters—and does not include the carrier frequency.

[0090] It is worth noting that during the optimization process, the system uses the dynamic safety threshold determined in S12 as the hard search space boundary of the particle swarm optimization algorithm. Any particle attempting to cross this safety boundary will be forcibly pulled back to the boundary value. The construction of the local fitness function is based on a simplified theoretical model of switching loss, and the function is defined as a weighted sum of switching loss, conduction loss, and safety penalty term. Regarding the determination of the weight coefficients, this embodiment adopts a calibration method based on order-of-magnitude normalization. Specifically, under rated load conditions, the orders of magnitude of the theoretical switching loss and conduction loss are calculated separately. If the numerical order of magnitude of the switching loss is k times that of the conduction loss, then the weight coefficient of the conduction loss is set to k times the weight coefficient of the switching loss. Through this normalization process, the natural differences in the numerical magnitudes of different loss components are eliminated, ensuring that the particle swarm optimization algorithm can optimize the two types of losses in a balanced manner, preventing the algorithm from excessively favoring the larger loss term and ignoring the optimization space of the other term.

[0091] It should be noted that the calibration method based on order-of-magnitude normalization aims to avoid the optimization process being dominated by a single component due to differences in the physical dimensions of losses and typical numerical values. In specific implementation, the switching loss P is first calculated under rated operating conditions. sw_nom and conduction loss P cond_nom The theoretical approximation; let its ratio be k' = P sw_nom / P cond_nom Then, in the actual construction of the fitness function F... local =W sw ·P sw +W cond ·P cond When +Penalty, let W cond =W sw ·k'; This means that near the saturation point, the contributions of the two losses to the objective function are adjusted to the same order of magnitude, thus ensuring that the particle swarm optimization algorithm can effectively optimize both simultaneously; weight base W sw It can be fine-tuned according to optimization preferences, for example, the default setting is 1.0.

[0092] For example, in this embodiment, the optimal driving waveform parameters output by the particle swarm optimization algorithm are serialized and packaged according to a predefined data structure. This data structure includes parameter type identifiers, parameter values, and cyclic redundancy check codes, forming a structured data packet, i.e., the set of driving parameters to be verified. This packaging format is designed to strictly match the input interface of the online simulation model in subsequent steps, facilitating direct loading for virtual verification.

[0093] In step S15, the set of driving parameters to be verified is subjected to online prediction verification based on the virtual thermal gradient model, and a verification result identifier is generated, including:

[0094] A preset equivalent circuit model of power devices is retrieved and combined with the virtual thermal gradient model to construct an online simulation environment that includes thermal-electric coupling characteristics.

[0095] The set of driving parameters to be verified is loaded into the online simulation environment to perform virtual operation testing and output virtual response data.

[0096] The stability fluctuation index is calculated based on the virtual response data, and a preset stability threshold is obtained.

[0097] The stability fluctuation index is compared numerically with the preset stability threshold.

[0098] If the stability fluctuation index is less than the stability threshold, a verification pass flag is generated;

[0099] If the stability fluctuation index is greater than or equal to the stability threshold, a verification failure flag is generated.

[0100] In one implementation, this embodiment employs a lightweight multiphysics coupling architecture for constructing the online simulation environment. The system first retrieves a pre-set equivalent circuit model of the power device from the firmware storage area. It is particularly important to note that the parameters of this circuit model are not arbitrarily set, but rather derived from the SPICE model datasheet provided by the power device manufacturer, or extracted during the R&D phase through thermal transient impedance testing conforming to JESD51-14 standards, i.e., structure function analysis. The model includes three pairs of thermal resistance-thermal capacity time constants from the device junction to the case, as well as the on-resistance characteristics that vary with temperature.

[0101] In this embodiment, the equivalent circuit model of the power device is pre-installed in the non-volatile memory of the controller in a parameterized form. Specifically, during the system production stage, for the selected specific model of MOSFET, key steady-state and transient parameters (including the thermal resistance-thermal capacitance network parameters and the coefficients of on-resistance as a function of junction temperature) are extracted based on the SPICE model datasheet. These parameters are organized into one or more lookup tables or fitting coefficient tables and are fixed together with the model solver code (a discretization solver program based on the forward Euler method). During runtime, the simulation verification module calls the solver, using the current case temperature provided by the virtual thermal gradient model as the starting temperature of the thermal network and the calculated losses of the electrical part as the input of the thermal network, to iteratively solve for the junction temperature. At the same time, the calculated junction temperature is fed back to the lookup table or calculation module of the electrical part in real time to complete the single-step thermal-electric coupling calculation. This model ensures that several cycles of closed-loop simulation prediction are completed on the embedded processor within milliseconds.

[0102] Subsequently, the system calls the virtual thermal gradient model established in S11, mapping its real-time thermal field state to the thermal network boundary conditions of the circuit model. By using the loss output of the electrical model as the input of the thermal model, and feeding back the temperature output of the thermal model to influence the physical parameters of the electrical model, a closed-loop simulation system capable of reflecting real-world operating conditions is constructed. This reduced-order model design based on the extraction of real physical parameters ensures the physical realism of the simulation while significantly reducing computational complexity, enabling it to run in the background tasks of an embedded processor.

[0103] In one implementation, for the virtual operation test, the system uses the set of driving parameters to be verified generated in S14 as the driving excitation signal for the simulation environment. The coupled equations are discretized and solved using the forward Euler method or the Runge-Kutta method to simulate the system's dynamic response after applying the new set of parameters within a short-term prediction window, such as ten pulse width modulation cycles. The data stream output from this simulation process is the virtual response data, which includes the predicted phase current waveform, bus voltage ripple, and transient change curves of the device junction temperature.

[0104] It should be noted that the calculation of the stability fluctuation index aims to quantify whether the new parameters have caused unexpected system oscillations or instability, rather than simply comparing the similarity with historical data. This resolves the logical paradox that "the greater the optimization magnitude, the higher the deviation, leading to misjudgment." In this embodiment, second-order difference operations are performed on the current waveform in the virtual response data to extract the variance or peak value of the current change rate as a quantitative index characterizing the system's stability. The preset stability threshold is determined based on the safe operating area of ​​the power devices and the electromagnetic compatibility requirements of the system. Specifically, the current ripple amplitude of the system in the critical stable state is measured through offline experiments, and 80% of this amplitude is set as the safe threshold.

[0105] It is worth noting that this embodiment strictly adheres to the principle of prioritizing stability in the generation logic of the verification results. The system compares the calculated stability fluctuation index with the preset stability threshold. If the fluctuation index is greater than or equal to the threshold, it indicates that although the parameters optimized by the particle swarm optimization algorithm may have high theoretical energy efficiency, they have led to the deterioration of the system's dynamic performance, such as aggressive overshoot or oscillation, posing a risk of loss of control. In this case, a verification failure flag is directly generated. If the fluctuation index is less than the threshold, it indicates that the system response is stable. In this case, the system assumes that the optimized parameters are valid and generates a verification success flag. This mechanism ensures that only physically stable control strategies are allowed to enter the subsequent update process, constructing a security firewall to prevent "bad parameters" from accidentally entering the system.

[0106] In step S16, based on the verification result identifier, the control register of the drive system is subjected to runtime configuration update or configuration retention processing, including:

[0107] Identify the specific type of the verification result identifier;

[0108] If the verification result is identified as a verification pass identifier, then the driver parameter set to be verified is encoded into instructions to generate a runtime configuration update instruction, and the runtime configuration update instruction is written into the control register of the driver system to complete the runtime configuration update.

[0109] If the verification result is marked as a verification failure, then the configuration retention process is executed to keep the current driver parameters unchanged and record the failure log.

[0110] In one implementation, for identifying the verification result identifier, this embodiment allocates a dedicated status flag register in the microcontroller's memory. The system reads the Boolean or enumerated value of this register to determine the conclusion of the preceding online prediction verification step. If the identification result is a verification failure identifier, it indicates that although the current local optimization parameters may theoretically have energy efficiency advantages, they have failed the safety test in terms of dynamic stability, posing a risk of system oscillation or overheating and runaway. At this time, the system triggers configuration hold processing, which physically involves locking the write permission of the pulse width modulation generator's configuration register and resetting the intermediate variables of the optimization algorithm, forcibly maintaining the driving parameters from the previous moment unchanged. Simultaneously, the system records the failed parameter combination and corresponding operating data to the error log area of ​​non-volatile memory as a negative sample or fault diagnosis basis for subsequent algorithm iterations.

[0111] In one implementation, if the identification result is a verification pass flag, the generation of the instruction encoding and runtime configuration update instruction in this embodiment follows the underlying communication protocol specification of the driver chip, such as a serial peripheral interface or internal integrated circuit bus protocol. The system maps the physical quantities in the set of driver parameters to be verified into binary machine code. For example, the microsecond-level dead time is multiplied by the clock frequency coefficient to convert it into a 16-bit count value; the drive current amplitude is converted into a control word for the digital-to-analog converter. Subsequently, the system adds a frame header, register address bits, and cyclic redundancy check code to these data, encapsulating them into a complete communication data frame, which is the runtime configuration update instruction.

[0112] It is worth noting that, for the write operation to the control register, to prevent waveform distortion or bridge arm shoot-through during parameter switching, this embodiment employs a hardware-level double-buffering mechanism to ensure atomicity. The system first writes the runtime configuration update instruction to the shadow register, at which point the parameters have not yet taken effect. Only when the current pulse width modulation period ends, i.e., when a counter overflows or underflows, does the hardware logic automatically load the value from the shadow register into the active register, i.e., the control register. This mechanism ensures that all drive parameters, such as frequency, dead time, and current, are updated simultaneously, guaranteeing the continuity and safety of the control waveform.

[0113] In one implementation, this embodiment employs an experience-based update strategy for updating the preset control strategy configuration library. The configuration library is a lookup table stored in electrically erasable read-only memory or flash memory, recording the optimal combination of drive parameters under different operating conditions. When a new parameter set is verified and successfully executed, the system checks the energy efficiency improvement brought by that parameter set. If the improvement exceeds a preset update threshold, for example, more than one percent, the system overwrites the parameter set into the storage unit corresponding to the operating condition in the configuration library, using it as a new best practice sample. In this way, the system can continuously accumulate successful control experience, ensuring that the final drive control configuration not only adapts to the current situation but also provides better initial values ​​for future rapid cold starts or similar operating conditions.

[0114] It should be noted that, to ensure the robustness of the strategy library updates and avoid the overwriting of inferior parameters due to chance or special operating conditions, this embodiment sets conservative update conditions. First, the calculation benchmark for the energy efficiency improvement is the historical average energy efficiency of the parameter set stored under the corresponding operating condition index in the current strategy library when executing the current task. Second, the update trigger must meet the following conditions: after a single successful verification, the calculated improvement exceeds the update threshold (e.g., 1%), and in the subsequent N consecutive (e.g., 3) similar operating condition cycles (judged based on load current and thermal state), the parameter set can pass online verification and maintain energy efficiency improvement. After the above conditions are met, the system replaces the original parameters in the library with the new parameter set. At the same time, the replaced old parameter set will be archived as a historical version for rollback in special circumstances.

[0115] In summary, this invention overcomes the physical lag of traditional sensors in explosion-proof sealed environments by constructing a virtual thermal gradient model that integrates spatiotemporal characteristics. Utilizing an asynchronous collaborative architecture of genetic algorithms and particle swarm optimization, it achieves refined control from global safety boundaries to microscopic driving waveforms while ensuring the real-time performance of the embedded system. In particular, the introduction of an online prediction and verification mechanism based on reduced-order equivalent circuits and a strict token passing strategy for verification results ensures that only parameters confirmed as safe and efficient through virtual simulation are written into the hardware. This effectively breaks the vicious cycle of heat accumulation and efficiency degradation in power devices, improving the operational stability and adaptability of the explosion-proof robot under high-load conditions.

[0116] Reference Figure 2 The second embodiment of the present invention provides an optimized drive control system for explosion-proof robot MOSFETs based on intelligent algorithms, comprising:

[0117] The thermal field sensing and modeling module is used to acquire multi-point temperature monitoring data of power devices, extract thermal field gradient features, and construct a virtual thermal gradient model.

[0118] The threshold optimization module is used to perform optimization calculations of the running boundary parameters based on the virtual thermal gradient model to determine the dynamic safety threshold.

[0119] The initial strategy matching module is used to acquire the real-time output status data of the drive system and combine it with the dynamic safety threshold to perform working mode switching matching and generate an initial steady-state control strategy.

[0120] The parameter optimization module is used to monitor load efficiency feedback data during dynamic operation, and when efficiency degradation characteristics are identified, it triggers multi-objective optimization of driving parameters with the dynamic safety threshold as the boundary constraint to obtain the set of driving parameters to be verified.

[0121] The simulation verification module is used to perform online prediction verification of the driving parameter set to be verified based on the virtual thermal gradient model, and generate a verification result identifier.

[0122] The execution control module is used to perform runtime configuration updates or configuration retention processing on the control registers of the drive system based on the verification result identifier.

[0123] It should be noted that the intelligent algorithm-based explosion-proof robot MOS transistor optimization drive control system provided in this embodiment of the invention is used to execute all the process steps of the intelligent algorithm-based explosion-proof robot MOS transistor optimization drive control method in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0124] This invention also provides an electronic device. The electronic device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a smart algorithm-based optimized drive control program for MOSFETs in explosion-proof robots. When the processor executes the computer program, it implements the steps in the various embodiments of the smart algorithm-based optimized drive control method for MOSFETs in explosion-proof robots described above, for example... Figure 1 Step S11 is shown. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the above system embodiments, such as the thermal field sensing and modeling module.

[0125] For example, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the electronic device.

[0126] The electronic device may be a desktop computer, laptop, handheld computer, or smart tablet, etc. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. It may include more or fewer components than described above, or combine certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, etc.

[0127] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the electronic device, connecting all parts of the electronic device via various interfaces and lines.

[0128] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, RAM, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0129] Wherein, if the modules / units integrated in the electronic device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0130] It should be noted that the system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the system embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0131] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for driving and controlling MOS tubes of an explosion-proof robot based on an intelligent algorithm, characterized in that, The method comprises the following steps: acquiring multi-point temperature monitoring data of a power device, performing thermal field gradient feature extraction processing, and constructing a virtual thermal gradient model; performing running boundary parameter optimization calculation according to the virtual thermal gradient model to determine a dynamic safety threshold; acquiring real-time output state data of a driving system, and combining the dynamic safety threshold to perform working mode switching matching to generate a preliminary steady-state control strategy; monitoring load efficiency feedback data in dynamic operation, and when an efficiency recession feature is identified, triggering driving parameter multi-objective optimization with the dynamic safety threshold as a boundary constraint to obtain a to-be-verified driving parameter set; performing online prediction verification of the to-be-verified driving parameter set based on the virtual thermal gradient model to generate a verification result identifier; performing runtime configuration update or configuration retention processing on a control register of the driving system according to the verification result identifier.

2. The intelligent algorithm-based explosion-proof robot MOS tube optimization driving control method according to claim 1, characterized in that, The method comprises the following steps: acquiring multi-point temperature monitoring data of a power device, performing thermal field gradient feature extraction processing, and constructing a virtual thermal gradient model; acquiring multi-channel temperature readings of the surface and heat dissipation area of the power device to obtain the multi-point temperature monitoring data; performing spatial position mapping processing on the multi-point temperature monitoring data to obtain a temperature spatial distribution matrix; performing thermal flow density and temperature rise rate calculation according to the temperature spatial distribution matrix to extract thermal field gradient features; 3. The intelligent algorithm-based explosion-proof robot MOS tube optimization driving control method according to claim 1, characterized in that, performing multi-dimensional numerical fitting processing according to the thermal field gradient features to establish the virtual thermal gradient model. The method comprises the following steps: performing thermal accumulation trend extrapolation processing according to the virtual thermal gradient model to obtain future thermal field evolution data; constructing a global safety fitness function according to the future thermal field evolution data, and performing iterative optimization processing on preset initial running boundary parameters by using a genetic algorithm to obtain an optimal boundary parameter combination; 4. The intelligent algorithm-based explosion-proof robot MOS tube optimization driving control method according to claim 1, characterized in that, performing safety boundary quantitative mapping according to the optimal boundary parameter combination to obtain the dynamic safety threshold. The method comprises the following steps: acquiring output current values and output voltage values of the driving system to obtain the real-time output state data; performing safety margin calculation and risk classification processing according to the real-time output state data and the dynamic safety threshold to obtain a running risk level label; performing mode switching matching by retrieving a preset control strategy library according to the running risk level label to obtain a target working mode; 5. The intelligent algorithm-based explosion-proof robot MOS tube optimization driving control method according to claim 1, characterized in that, performing control logic analysis on the target working mode to generate the preliminary steady-state control strategy, and writing the preliminary steady-state control strategy into a control register of the driving system. The method comprises the following steps: acquiring real-time input voltage, input current, output voltage, and output current of the driving system, and performing power calculation and ratio calculation to obtain real-time energy efficiency ratio data; The real-time energy efficiency ratio data is compared with a preset degradation threshold value; If the real-time energy efficiency ratio data is lower than the preset degradation threshold value, a feature extraction process is performed to obtain an efficiency degradation feature; The dynamic safety threshold value is set as the search space boundary of a particle swarm optimization algorithm, and a local fitness function is constructed according to the efficiency degradation feature; The particle swarm optimization algorithm is used to perform limited iterative optimization on the driving waveform parameters to obtain optimal driving waveform parameters; The optimal driving waveform parameters are packaged to obtain the to-be-verified driving parameter set.

6. The intelligent algorithm based explosion-proof robot MOS tube optimization driving control method according to claim 1, characterized in that, The to-be-verified driving parameter set is subjected to online prediction verification based on the virtual thermal gradient model to generate a verification result identifier, including: A preset power device equivalent circuit model is called and combined with the virtual thermal gradient model to construct an online simulation environment containing thermal-electric coupling characteristics; The to-be-verified driving parameter set is loaded into the online simulation environment for virtual running test, and virtual response data is output; A stability fluctuation index is calculated according to the virtual response data, and a preset stability threshold value is obtained; The stability fluctuation index is compared with the preset stability threshold value in numerical value; If the stability fluctuation index is less than the stability threshold value, a verification pass identifier is generated; If the stability fluctuation index is greater than or equal to the stability threshold value, a verification failure identifier is generated.

7. The intelligent algorithm based explosion-proof robot MOS tube optimization driving control method according to claim 1, characterized in that, According to the verification result identifier, the control register of the driving system is subjected to runtime configuration update or retention processing, including: The specific type of the verification result identifier is identified; If the verification result identifier is a verification pass identifier, the to-be-verified driving parameter set is subjected to instruction encoding to generate a runtime configuration update instruction, and the runtime configuration update instruction is written into the control register of the driving system to complete the runtime configuration update; If the verification result identifier is a verification failure identifier, configuration retention processing is performed to maintain the current driving parameters unchanged, and failure logs are recorded.

8. An intelligent algorithm-based anti-explosion robot MOS tube optimization driving control system, characterized in that, It includes: A thermal field perception and modeling module is used to obtain multi-point temperature monitoring data of a power device and perform thermal field gradient feature extraction processing to construct a virtual thermal gradient model; A threshold optimization module is used to perform running boundary parameter optimization calculation according to the virtual thermal gradient model to determine a dynamic safety threshold value; A strategy initial matching module is used to obtain real-time output state data of a driving system and combine the dynamic safety threshold value to perform working mode switching matching to generate an initial steady-state control strategy; A parameter optimization module is used to monitor load efficiency feedback data in dynamic running, and when an efficiency degradation feature is identified, a driving parameter multi-objective optimization is triggered with the dynamic safety threshold value as a boundary constraint to obtain a to-be-verified driving parameter set; An simulation verification module is used to perform online prediction verification based on the virtual thermal gradient model on the to-be-verified driving parameter set to generate a verification result identifier; An execution control module is used to perform runtime configuration update or configuration retention processing on the control register of the driving system according to the verification result identifier.

Images