Automobile inflator pump and dynamic voltage monitoring control system thereof

Abstract

The invention relates to the technical field of automobile inflator pump monitoring, and discloses an automobile inflator pump and a dynamic voltage monitoring control system thereof. The system comprises a voltage data acquisition module, a dynamic noise suppression module, a voltage fluctuation modeling module, a self-adaptive compensation control module and the like. The voltage data acquisition module acquires and processes multi-dimensional data, the dynamic noise suppression module removes noise, the voltage fluctuation modeling module extracts voltage and load correlation characteristics, and the adaptive compensation control module optimizes control parameters to stabilize the voltage. In addition, the system is further provided with a digital twinborn simulation module monitoring gas circuit, a distributed updating module optimization model based on federated learning, a redundancy protection circuit guarantee safety, and a fault mode library for providing preventive maintenance suggestions. The system can improve the running stability, safety and efficiency of the automobile inflator pump, and effectively overcomes the defects of a traditional inflator pump in the aspects of voltage monitoring control and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of automotive air pump monitoring, and particularly to an automotive air pump and its dynamic voltage monitoring and control system. Background Art

[0002] In the field of modern automobiles, the automotive air pump is an important device to ensure the normal air pressure of tires, and its performance and stability are crucial. With the continuous development of automotive electronic systems, the voltage stability of in-vehicle power supplies faces many challenges, which have a significant impact on the reliable operation of automotive air pumps.

[0003] Traditional automotive air pumps have many deficiencies in voltage monitoring and control. On the one hand, their monitoring of the in-vehicle power supply voltage is often inaccurate and unable to obtain voltage data in real time and comprehensively. For example, it can only monitor the amplitude of the voltage, ignoring information such as the frequency and phase of voltage fluctuations, resulting in the inability to detect potential voltage anomalies in a timely manner. On the other hand, there are also defects in the monitoring of the motor working current and ambient temperature. The change of the motor working current is closely related to the load of the air pump. If it cannot be accurately monitored, it is difficult to accurately evaluate the working state of the air pump. The ambient temperature also affects the performance of the air pump. For example, low temperature will increase the viscosity of the lubricating oil in the air pump, resulting in an increase in the motor load. However, traditional systems are difficult to effectively monitor and cope with the impact brought by changes in ambient temperature.

[0004] In terms of signal processing, traditional systems lack effective noise suppression means. The complex electromagnetic environment inside the vehicle will generate a large amount of noise, interfering with the acquisition of signals such as voltage and current, making the obtained data have errors, and thus affecting subsequent analysis and control decisions. At the same time, the ability to model and analyze voltage fluctuations is limited, unable to deeply explore the non-linear correlation characteristics between voltage transient fluctuations and load changes, difficult to accurately predict the voltage change trend, and unable to take measures in advance to ensure the stable operation of the air pump.

[0005] In terms of control strategies, traditional automotive air pumps usually adopt fixed control parameters and cannot be dynamically adjusted according to real-time data such as voltage, current, and temperature. When the in-vehicle power supply voltage fluctuates or the air pump load changes, this fixed-parameter control method will cause the motor speed to be unstable, which not only affects the inflation efficiency but also may shorten the service life of the motor. For example, when the voltage is low, the motor speed may decrease and the inflation time may be extended; when the voltage is too high, the motor may overheat severely or even burn out due to overload.

[0006] In addition, traditional air pumps also have deficiencies in terms of fault warning and maintenance. Lacking effective fault mode recognition and prediction mechanisms, they can often only be repaired after a fault occurs, unable to detect potential problems in advance and take preventive measures, which brings inconvenience to users, increases the usage cost, and poses safety hazards. To sum up, the existing automotive air pump technology has obvious defects in aspects such as voltage monitoring, signal processing, control strategies, and fault warning. There is an urgent need for a more advanced and intelligent dynamic voltage monitoring and control system to solve these problems in order to improve the overall performance and reliability of automotive air pumps. Summary of the Invention

[0007] The purpose of the present invention is to provide a dynamic voltage monitoring and control system for an automotive air pump to solve the problems raised in the above background technology.

[0008] To achieve the above purpose, the present invention provides the following technical solution: A dynamic voltage monitoring and control system for an automotive air pump, the system includes:

[0009] A voltage data acquisition module, which is used to collect vehicle power supply voltage, motor working current, and ambient temperature data through multi-channel sensors, and generate a multi-dimensional real-time monitoring data set;

[0010] A dynamic noise suppression module, which is used to perform non-linear signal decomposition on the multi-dimensional real-time monitoring data set, and use the combined algorithm of empirical mode decomposition and Kalman filtering to separate high-frequency noise components and steady-state voltage characteristics, and generate a denoised time series signal sequence;

[0011] A voltage fluctuation modeling module, which is used to construct a deep generative model based on variational autoencoders, perform latent variable space mapping on the denoised time series signal sequence, extract non-linear correlation characteristics of voltage transient fluctuations and load changes, and generate a dynamic voltage state coding vector;

[0012] An adaptive compensation control module, which is used to design a policy optimization network based on reinforcement learning, and dynamically adjust the PWM duty cycle and motor speed control parameters according to the dynamic voltage state coding vector to generate an anti-disturbance voltage stabilization instruction.

[0013] Preferably, the non-linear signal decomposition includes:

[0014] Perform empirical mode decomposition on the vehicle power supply voltage signal, and generate a set of intrinsic mode function components through local extreme point interpolation;

[0015] Construct a Kalman filter state equation, input the high-frequency intrinsic mode components as observation noise, and iteratively update the system state estimation value;

[0016] Use a sliding window entropy value detection algorithm to identify the mutation points of the steady-state voltage component and segment continuous steady signal segments.

[0017] Preferably, the construction of the deep generation model includes:

[0018] Introduce a causal convolutional layer in the encoder to capture the temporal causal relationship of the voltage sequence and output the parameters of the latent variable probability distribution;

[0019] In the decoder, use a gated recurrent unit to reconstruct the voltage fluctuation waveform and superimpose an attention mechanism to focus on the characteristics at the moment of load mutation;

[0020] Optimize the parameters of the encoder and decoder through an adversarial training strategy to minimize the weighted sum of the reconstruction error and the discriminator loss function.

[0021] Preferably, the design of the policy optimization network includes:

[0022] Define the state space as a multi-dimensional vector of the voltage ripple coefficient, current harmonic distortion rate, and temperature compensation factor;

[0023] Construct a double-delay deep deterministic policy gradient algorithm framework and use a target network and an experience replay pool to optimize the convergence of the control strategy;

[0024] Design the reward function as a linear combination of the reciprocal of the absolute value of the voltage deviation and the energy loss index to drive the policy network to generate an optimal compensation instruction.

[0025] Preferably, the system further includes:

[0026] A digital twin simulation module for establishing a motor-air pump coupling model based on physical equations and real-time simulating the dynamic relationship between the air chamber pressure and the motor torque;

[0027] Input the measured voltage data into the simulation model to generate a virtual pressure curve and trigger an air circuit leakage warning signal through residual analysis.

[0028] Preferably, the generation of the multi-dimensional real-time monitoring data set includes:

[0029] Synchronously collect the power bus CAN signal, Hall sensor current pulse, and NTC thermistor temperature data;

[0030] Use a tensor decomposition algorithm to fill in the missing values of the heterogeneous data and construct a three-dimensional spatio-temporal data matrix;

[0031] Eliminate the sensor baseline drift through the moving average difference method and enhance the identifiability of the data transient characteristics.

[0032] Preferably, the system further includes:

[0033] Construct a distributed update module based on federated learning and jointly train a voltage prediction model through in-vehicle edge devices and cloud servers;

[0034] Design a differential privacy mechanism to encrypt local gradient parameters, and use a model aggregation algorithm to generate a globally shared compensation control strategy.

[0035] Preferably, the generation of the anti-disturbance voltage stabilization instruction includes:

[0036] Establish a multi-objective optimization model with voltage stability, motor life loss, and energy efficiency as constraints;

[0037] Use the non-smooth Newton method to solve the optimization problem with inequality constraints, and generate the Pareto front solution set of PWM frequency and duty cycle;

[0038] Select the optimal compromise solution through the fuzzy decision-making layer and output the control parameters of the pulse width modulation signal.

[0039] Preferably, the system further includes:

[0040] Design a redundant protection circuit based on optocoupler isolation to monitor the junction temperature and switching loss of the IGBT module in real time;

[0041] When detecting an overcurrent or overheating state, automatically switch to the standby drive circuit and reset the control strategy parameters.

[0042] Preferably, the system further includes:

[0043] Construct a fault mode library based on the time series clustering algorithm to perform feature pattern matching on historical voltage fluctuation data;

[0044] Use the dynamic time warping algorithm to calculate the similarity between the real-time signal and the fault template, and generate preventive maintenance recommendation instructions.

[0045] Compared with the prior art, the beneficial effects of the present invention are:

[0046] The dynamic voltage monitoring and control system of the automotive air pump of the present invention has many significant beneficial effects. In terms of data acquisition and processing, the voltage data acquisition module synchronously acquires the vehicle power supply voltage, motor working current, and ambient temperature data with a multi-channel sensor, fills in the missing values using the tensor decomposition algorithm, eliminates the baseline drift using the moving average difference method, and constructs an accurate three-dimensional spatio-temporal data matrix to ensure that the data input into the system is comprehensive and accurate. The dynamic noise suppression module uses the combined algorithm of empirical mode decomposition and Kalman filtering to effectively separate the high-frequency noise components and the steady-state voltage characteristics, generating a denoised time series signal sequence, greatly improving the data quality and laying a foundation for subsequent accurate analysis.

[0047] The voltage fluctuation modeling module constructs a deep generative model based on variational autoencoders. Through causal convolutional layers, gated recurrent units, and attention mechanisms, it can accurately extract the non-linear correlation features between voltage transient fluctuations and load changes, generate dynamic voltage state encoding vectors, realize in-depth analysis and effective prediction of voltage fluctuations, and provide a strong basis for the formulation of control strategies. The adaptive compensation control module designs a policy optimization network based on reinforcement learning, and dynamically adjusts the PWM duty cycle and motor speed control parameters according to the dynamic voltage state encoding vector. This process realizes the optimization of control strategies by reasonably defining the state space, constructing a double-delayed deep deterministic policy gradient algorithm framework, and designing a scientific reward function, not only ensuring voltage stability, but also taking into account motor life loss and energy efficiency. For example, when there is a voltage fluctuation, it can quickly adjust the PWM duty cycle to make the motor operate efficiently under a stable voltage, reduce the additional losses caused by voltage instability, extend the service life of the motor, and at the same time reduce energy consumption.

[0048] The digital twin simulation module in the system establishes a motor-air pump coupling model to real-time simulate the dynamic relationship between the air chamber pressure and the motor torque. Through residual analysis, it can timely detect air leakage problems in the air circuit, send out warning signals, effectively avoid the reduction of inflation efficiency and equipment damage caused by air leakage, and ensure the normal operation of the inflator pump. The distributed update module based on federated learning collaboratively trains the voltage prediction model through in-vehicle edge devices and cloud servers, encrypts local gradient parameters using differential privacy mechanisms, and uses model aggregation algorithms to generate globally shared compensation control strategies, which not only protects data privacy but also continuously optimizes the model with the powerful computing power of the cloud, improving the adaptability and performance of the system.

[0049] The redundant protection circuit monitors the junction temperature and switching losses of the IGBT module in real time. When detecting an overcurrent or overheating state, it automatically switches to the standby drive circuit and resets the control strategy parameters, greatly improving the safety and reliability of the system, and preventing the entire inflator pump system from crashing due to IGBT module failures. The fault mode library analyzes historical voltage fluctuation data through time series clustering algorithms, calculates the similarity between real-time signals and fault templates using dynamic time warping algorithms, and generates preventive maintenance advice instructions to help users detect potential faults in advance, perform maintenance in a timely manner, reduce maintenance costs, and improve the usage efficiency and safety of the automotive inflator pump. In short, the present invention comprehensively improves the performance of the automotive inflator pump from multiple aspects such as data acquisition and processing, voltage monitoring and analysis, control strategy optimization, to fault warning and maintenance, and has extremely high practical value and market prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is the working principle diagram of a dynamic voltage monitoring and control system for an automotive inflator pump according to the present invention;

[0051] Figure 2 It is a flowchart for the non-linear signal decomposition of the dynamic noise suppression module;

[0052] Figure 3 It is a flowchart for the construction of the deep generation model of the voltage fluctuation modeling module;

[0053] Figure 4 It is a flowchart for the generation of the multi-dimensional real-time monitoring data set;

[0054] Figure 5 It is a schematic diagram of the device structure of the vehicle charging pump;

[0055] Figure 6 It is Figure 5 Schematic diagram of the structure after omitting the baffle;

[0056] Figure 7 Schematic diagram of the structure after partial removal of the main body shell;

[0057] Figure 8 It is a schematic diagram of the structure of the main body shell and the tool slot. Specific implementation manners

[0058] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0059] Please refer to Figures 1-8 , the present invention provides a dynamic voltage monitoring and control system for a vehicle inflator, and its overall implementation solution is as follows:

[0060] The dynamic voltage monitoring and control system of this vehicle inflator mainly consists of a voltage data acquisition module, a dynamic noise suppression module, a voltage fluctuation modeling module, and an adaptive compensation control module.

[0061] The voltage data acquisition module collects vehicle power supply voltage, motor working current, and ambient temperature data through multi-channel sensors, and generates a multi-dimensional real-time monitoring data set. The multi-channel sensors are respectively connected to the vehicle power supply, the motor, and the ambient temperature detection points to achieve synchronous acquisition of relevant data. The collected data includes power bus CAN signals, Hall sensor current pulses, and NTC thermistor temperature data. These heterogeneous data are processed by specific algorithms to construct a three-dimensional spatio-temporal data matrix, thereby providing a comprehensive and accurate data basis for subsequent analysis.

[0062] The dynamic noise suppression module receives the multi-dimensional real-time monitoring data set generated by the voltage data acquisition module and performs nonlinear signal decomposition on it. The module uses the empirical mode decomposition and Kalman filtering joint algorithm to separate the high-frequency noise component and the steady-state voltage characteristics, and then generates a noise-reduced time series signal sequence. Through this process, the noise interference in the data is removed, making the true characteristics of the voltage signal more prominent, providing reliable data for subsequent precise analysis.

[0063] The voltage fluctuation modeling module builds a deep generative model based on the variational autoencoder to perform latent variable space mapping on the denoised time series signal sequence. Through this operation, the nonlinear correlation characteristics of voltage transient fluctuations and load changes are extracted to generate a dynamic voltage state encoding vector. The encoder and decoder in the deep generative model work together, using technologies such as causal convolutional layers, gated recurrent units, and attention mechanisms to mine the complex features in the voltage signal, providing a key basis for subsequent control strategies.

[0064] The adaptive compensation control module is designed based on a reinforcement learning strategy optimization network. According to the dynamic voltage state encoding vector generated by the voltage fluctuation modeling module, the PWM duty cycle and motor speed control parameters are dynamically adjusted to generate anti-disturbance voltage stabilization instructions. This module optimizes the control strategy by defining a suitable state space, building an effective algorithm framework, and designing a reasonable reward function, so that the automobile air pump can operate stably under different voltage and load conditions, ensuring the normal operation of the air pump and the life and energy efficiency of the motor.

[0065] The implementation of the present invention is further described below in conjunction with Examples 1 to 6.

[0066] Embodiment 1:

[0067] In this embodiment, specific implementation details of the voltage data acquisition module and the dynamic noise suppression module are further described.

[0068] For the voltage data acquisition module, the power bus CAN signal, Hall sensor current pulse and NTC thermistor temperature data are collected synchronously. The power bus CAN signal contains a variety of information about the vehicle power supply, such as voltage amplitude, voltage change trend, etc. These signals are collected and transmitted to the system through the CAN bus interface. The Hall sensor uses the Hall effect to convert the motor operating current into an easily detectable current pulse signal. The frequency and amplitude of these pulse signals are related to the motor current. By counting and analyzing the pulse signals, the motor operating current data can be accurately obtained. The NTC thermistor is sensitive to ambient temperature, and its resistance value will change with the change of ambient temperature. By measuring the resistance value of the NTC thermistor and according to its characteristic curve, the ambient temperature data can be calculated.

[0069] These heterogeneous data collected may have problems of missing values and baseline drift. The tensor decomposition algorithm is used to fill in the missing values of the heterogeneous data. The tensor decomposition algorithm is an algorithm that can decompose a multi-dimensional tensor into the product form of multiple low-dimensional tensors. By performing tensor decomposition on the collected three-dimensional spatio-temporal data matrix, the internal structure and correlation of the data can be utilized to reasonably estimate and fill in the missing data. For example, assuming that the voltage data at a certain moment is missing, the tensor decomposition algorithm can predict the value of the missing voltage data based on other relevant data before and after that moment and the structural characteristics of the entire data matrix, thus ensuring the integrity of the data.

[0070] Meanwhile, the sensor baseline drift is eliminated by the moving average difference method. The moving average difference method first performs moving average processing on the data, calculates the average value of the data within a certain time window, and then subtracts the moving average value from the current data to obtain the difference data. This can effectively eliminate the baseline drift phenomenon of the sensor caused by various factors and enhance the identifiability of the transient characteristics of the data. For example, if the baseline of the sensor slowly rises after long-term use, after being processed by the moving average difference method, the true change trend of the data can be highlighted, facilitating the subsequent accurate analysis of signals such as voltage and current.

[0071] The specific process of non-linear signal decomposition in the dynamic noise suppression module is as follows: First, the empirical mode decomposition is performed on the vehicle power supply voltage signal. The empirical mode decomposition is a decomposition method based on the local characteristic time scale of the signal. By finding the local extreme points of the signal and performing cubic spline interpolation on these extreme points, a set of intrinsic mode function components is generated. These intrinsic mode function components represent the characteristics of the signal at different time scales, including high-frequency noise components and steady-state voltage characteristic components.

[0072] Then, the Kalman filter state equation is constructed, and the high-frequency intrinsic mode components are used as the observation noise input. The Kalman filter is an algorithm that uses the linear system state equation to optimally estimate the system state through the system input and output observation data. In this system, the high-frequency intrinsic mode components are regarded as the observation noise, and the effective suppression of the noise is achieved by continuously iterating and updating the system state estimate value. Assume the system state equation is , and the observation equation is , where is the system state vector at time is the state transition matrix, is the control input matrix, is the control input vector, is the process noise vector, is the observation vector, is the observation matrix, is the observation noise vector. In this implementation, the high-frequency eigenmode components are corresponded to , and through the iterative calculation of the Kalman filter, the system state estimation value is continuously optimized, so as to separate the steady-state voltage characteristics.

[0073] Finally, the sliding window entropy detection algorithm is used to identify the mutation points of the steady-state voltage component and segment the continuous steady signal segments. The sliding window entropy detection algorithm calculates the entropy value of the signal within a certain time window. The entropy value can reflect the uncertainty and complexity of the signal. When the entropy value mutates, it indicates that the signal may have an abnormality or a state change. By continuously sliding the window and calculating the entropy value, the mutation points of the steady-state voltage component can be identified, and then the continuous and steady voltage signal can be segmented into multiple segments, which is convenient for more detailed analysis and processing of the voltage signals in different segments later.

[0074] Embodiment 2:

[0075] This embodiment focuses on the construction process of the deep generative model in the voltage fluctuation modeling module.

[0076] When constructing a deep generative model based on a variational autoencoder, a causal convolutional layer is introduced in the encoder part to capture the temporal causal relationship of the voltage sequence. The causal convolutional layer is a special convolutional layer, and its convolutional operation is only performed at the current time and previous times, and will not be affected by the data at future times. Therefore, it can well capture the temporal causal relationship of the voltage sequence. For example, when processing data where voltage changes over time, the causal convolutional layer can predict the voltage change trend at the current time based on the voltage values at past times, so as to better understand the internal law of the voltage sequence.

[0077] The encoder processes the denoised temporal signal sequence through the causal convolutional layer and outputs the parameter of the latent variable probability distribution. Suppose the output parameter of the latent variable probability distribution is (mean) and (logarithm of variance). These parameters represent the distribution characteristics of the voltage sequence in the latent variable space. By analyzing these parameters, some features that are difficult to directly observe from the original data in the voltage sequence can be mined.

[0078] The decoder uses a gated recurrent unit to reconstruct the voltage fluctuation waveform. The gated recurrent unit (GRU) is a special recurrent neural network unit, which can effectively solve the problems of gradient vanishing and gradient explosion in traditional recurrent neural networks and can better capture the long-term dependence relationship in time series. In the decoder, the GRU reconstructs the voltage fluctuation waveform step by step according to the parameter of the latent variable probability distribution output by the encoder. For example, the GRU can simulate the voltage change at different times according to the feature information in the latent variable, so as to restore a voltage fluctuation waveform close to the real one.

[0079] Meanwhile, the decoder superimposed attention mechanism focuses on the features at the moment of load mutation. The attention mechanism enables the model to automatically focus on the importance of different parts when processing data. In voltage fluctuation modeling, the voltage features at the moment of load mutation are crucial for analyzing the stability and performance of the system. Through the attention mechanism, when reconstructing the voltage fluctuation waveform, the model can pay more attention to the features at the moment of load mutation and enhance the expression of these key features in the model. For example, when the load suddenly increases or decreases, the attention mechanism will increase the weight of the voltage change features at that moment, enabling the model to capture these changes more accurately and reflect them in the reconstructed waveform.

[0080] Optimize the parameters of the encoder and decoder through an adversarial training strategy. The adversarial training strategy introduces a discriminator, whose role is to distinguish whether the reconstructed voltage fluctuation waveform comes from real data or data generated by the model. There is a mutual game among the encoder, decoder, and discriminator. The encoder and decoder try to generate a reconstructed waveform closer to the real data to deceive the discriminator; while the discriminator tries to accurately distinguish between real data and generated data. Optimize the parameters of the encoder and decoder by minimizing the weighted sum of the reconstruction error and the discriminator loss function. Assume the reconstruction error is , the discriminator loss function is , and the weighting coefficients are and , then the optimization objective is . During the training process, continuously adjust the parameters of the encoder, decoder, and discriminator to make the reconstructed waveform more accurate, and at the same time improve the model's ability to extract the non-linear correlation features of voltage transient fluctuations and load changes, and finally generate an accurate dynamic voltage state coding vector.

[0081] Example 3:

[0082] This example details the design of the policy optimization network in the adaptive compensation control module and the generation process of anti-disturbance voltage stabilization instructions.

[0083] In terms of the design of the policy optimization network, first define the state space as a multi-dimensional vector of voltage ripple coefficient, current harmonic distortion rate, and temperature compensation factor. The voltage ripple coefficient is an important indicator to measure the degree of voltage fluctuation, which reflects the voltage fluctuation magnitude within a cycle. The current harmonic distortion rate is used to measure the content of harmonic components in the current. Harmonics will have an adverse impact on the operating performance of the motor, so the current harmonic distortion rate is also one of the key parameters to evaluate the system performance. The temperature compensation factor is a parameter introduced considering the influence of environmental temperature on the motor performance. Different environmental temperatures will cause changes in parameters such as the resistance and inductance of the motor, thereby affecting the operating state of the motor. Through the temperature compensation factor, these influences can be corrected.

[0084] Construct the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm framework, and use the target network and experience replay pool to optimize the convergence of the control strategy. The target network is a copy of the policy optimization network, and its parameters are updated relatively slowly. By using the target network, the fluctuations in the policy optimization process can be reduced, and the stability of the algorithm can be improved. The experience replay pool stores the experience data generated during the policy optimization process, including information such as states, actions, rewards, and next states. During training, a batch of data is randomly sampled from the experience replay pool for training, which can break the correlation between the data, prevent the model from falling into local optimal solutions, and accelerate the convergence of the control strategy.

[0085] Design the reward function as a linear combination of the reciprocal of the absolute value of the voltage deviation and the energy loss index. Assume the voltage deviation is and the energy loss index is , the reward function can be expressed as , where is the weight coefficient, which is used to balance the importance of the voltage deviation and the energy loss in the reward function. When the voltage deviation is smaller, the value of is larger and the reward is higher; at the same time, when the energy loss index is lower, the value of is smaller and the reward is also higher. Through such a design of the reward function, the policy network is driven to generate the optimal compensation instruction, so that the system can minimize the energy loss while ensuring voltage stability.

[0086] In terms of generating anti-disturbance voltage stabilization instructions, establish a multi-objective optimization model with voltage stability, motor life loss, and energy efficiency as constraints. Voltage stability is the key to ensuring the normal operation of the vehicle air pump, motor life loss is related to the long-term use cost of the air pump, and energy efficiency affects the vehicle's energy consumption. In the multi-objective optimization model, these factors are used as constraints, and their mutual relationships are comprehensively considered.

[0087] Use the nonsmooth Newton method to solve the optimization problem with inequality constraints and generate the Pareto front solution set of the PWM frequency and duty cycle. The nonsmooth Newton method is an effective method for solving nonsmooth optimization problems. In this system, there are some nonsmooth constraint conditions and objective functions in the multi-objective optimization problem, so using the nonsmooth Newton method can solve it more efficiently. Through this method, a series of combinations of PWM frequency and duty cycle that satisfy different trade-off relationships can be obtained, and these combinations form the Pareto front solution set. The Pareto front solution set represents the set of all solutions where a certain objective cannot be further optimized without sacrificing other objectives.

[0088] Finally, the optimal compromise solution is selected through the fuzzy decision-making layer, and the control parameters of the pulse-width modulation signal are output. The fuzzy decision-making layer evaluates and selects each solution in the Pareto front solution set according to the actual requirements and operating conditions of the system. The fuzzy decision-making layer uses fuzzy logic to comprehensively consider multiple evaluation indicators, such as voltage stability, motor life loss, and energy efficiency. By setting appropriate membership functions and fuzzy rules, each solution is evaluated fuzzily, and finally an optimal compromise solution is selected to output the control parameters of the pulse-width modulation signal, such as PWM frequency and duty cycle, so as to achieve stable control of the automotive air pump.

[0089] Embodiment 4:

[0090] The digital twin simulation module establishes a motor-air pump coupling model based on physical equations to simulate the dynamic relationship between the air chamber pressure and the motor torque in real time. The motor-air pump coupling model comprehensively considers the electromagnetic characteristics of the motor and the mechanical characteristics of the air pump. From the perspective of the motor, according to Faraday's law of electromagnetic induction and the structural parameters of the motor, the electromagnetic torque equation of the motor is established , where is the electromagnetic torque of the motor, is the motor torque constant, is the armature current of the motor. For the air pump, according to the ideal gas state equation and mechanical principles, the relationship equation between the air chamber pressure and the motor torque is established. Assuming the volume of the air pump is , the initial pressure of the gas is , the temperature is , according to the ideal gas state equation ( is the amount of substance of the gas, is the molar gas constant), combined with the working process of the air pump, considering the piston movement, intake and exhaust processes of the air pump, etc., the equation for the change of the air chamber pressure with time is established. At the same time, the load torque generated by the air pump on the motor is related to the air chamber pressure and the structural parameters of the air pump. Through these physical equations, a motor-air pump coupling model is established.

[0091] The measured voltage data is input into the simulation model to generate a virtual pressure curve. The measured voltage data is obtained through the voltage data acquisition module, and these voltage data reflect the real-time state of the vehicle-mounted power supply. Substituting these voltage data into the motor-air pump coupling model, the model calculates the operating state of the motor, such as motor speed, current, etc., and then calculates the change of the air chamber pressure with time according to the working principle of the air pump to generate a virtual pressure curve.

[0092] Trigger the air circuit leakage warning signal through residual analysis. Residual analysis is to compare the virtual pressure curve with the actually measured air chamber pressure curve, calculate the difference between the two, that is, the residual. If there is a leakage in the air circuit, the actual air chamber pressure will be lower than the pressure value under normal conditions, resulting in the residual between the virtual pressure curve and the actual pressure curve exceeding the normal range. Set a residual threshold. When the residual is greater than this threshold, the system determines that there may be a leakage in the air circuit and triggers the air circuit leakage warning signal. For example, under normal conditions, the residual between the virtual pressure curve and the actual pressure curve is within When the residual exceeds , the system immediately issues a warning signal to remind the user to check whether there is a leakage in the air circuit, so as to perform maintenance in time and avoid the reduction of the working efficiency of the air pump or the inability to work properly due to the air circuit leakage.

[0093] Example 5:

[0094] This example elaborates on the specific implementation of the distributed update module based on federated learning.

[0095] Build a distributed update module based on federated learning, and jointly train the voltage prediction model through in-vehicle edge devices and cloud servers. The in-vehicle edge device has certain computing and storage capabilities and can initially process the collected data locally. In this system, the in-vehicle edge device collects relevant data of the automotive air pump, including voltage data, motor working current data, ambient temperature data, etc.

[0096] The cloud server has powerful computing resources and storage capabilities. The in-vehicle edge device encrypts the feature information of the local data and uploads it to the cloud server instead of directly uploading the original data, which can protect the privacy of user data. The cloud server receives the data features uploaded by multiple in-vehicle edge devices and uses the federated learning algorithm for model training. The federated learning algorithm allows each participating party to jointly train a global model without sharing the original data. During the training process, the cloud server updates the information of the model parameters uploaded by each in-vehicle edge device and generates a globally shared compensation control strategy through the model aggregation algorithm.

[0097] Design a differential privacy mechanism to encrypt the local gradient parameters. The differential privacy mechanism is a technology for protecting data privacy during data publishing and analysis. When the in-vehicle edge device trains the voltage prediction model, the calculated local gradient parameters contain sensitive information of the local data. To protect this privacy, before uploading the local gradient parameters, the differential privacy mechanism is used to encrypt them. The differential privacy mechanism adds a certain amount of noise to the gradient parameters, so that even if an attacker obtains the encrypted gradient parameters, they cannot accurately infer the information of the original data. Suppose the original gradient parameter is , and the added noise is , the encrypted gradient parameters is , where is random noise generated according to parameters such as the privacy budget of differential privacy. By reasonably adjusting the distribution and intensity of the noise, while ensuring the model training effect, data privacy can be maximally protected.

[0098] The model aggregation algorithm is used to generate a globally shared compensation control strategy. The model aggregation algorithm is one of the key technologies in federated learning. It integrates the local model parameters trained by each vehicle-mounted edge device to generate globally shared model parameters. Common model aggregation algorithms include the FedAvg algorithm, etc. In this system, the cloud server receives the encrypted local gradient parameters uploaded by multiple vehicle-mounted edge devices, performs operations such as weighted averaging on these parameters according to the model aggregation algorithm, and obtains the parameter update of the global model. Then, based on the updated global model parameters, a globally shared compensation control strategy is generated. This compensation control strategy can be sent to each vehicle-mounted edge device, enabling the automotive air pump to be controlled according to the globally optimized strategy, improving the overall performance and adaptability of the system.

[0099] Embodiment 6:

[0100] In the design of the redundant protection circuit based on optocoupler isolation, the optocoupler isolation technology uses the transmission of optical signals between the light-emitting diode and the photosensitive element to achieve electrical isolation, effectively avoiding electrical interference between different circuits and improving the stability and safety of the system. This redundant protection circuit is used to monitor the junction temperature and switching loss of the IGBT module in real time. The IGBT (Insulated Gate Bipolar Transistor), as a key component in the power drive circuit of the automotive air pump, its working state directly affects the performance and reliability of the air pump. Excessive junction temperature will cause the performance of the IGBT module to decline or even be damaged; while excessive switching loss will not only reduce the system efficiency but may also cause overheating problems.

[0101] Therefore, special temperature sensors and current sensors are used in the circuit to obtain the junction temperature data of the IGBT module and the current data reflecting the switching loss. The temperature sensor is closely attached to the heat sink or the chip of the IGBT module to measure the junction temperature in real time. The current sensor is connected in series in the main circuit of the IGBT to accurately collect the working current, and then the switching loss is calculated through relevant formulas. Taking the common switching loss calculation formula as an example, where represents the switching loss, is the collector-emitter voltage, is the collector current, is the switching frequency, and are the turn-on time and turn-off time respectively. Through the measured , and the known switching frequency and other parameters, the switching loss can be calculated.

[0102] When an overcurrent or overheat state is detected, the redundant protection circuit will automatically switch to the standby drive circuit. This process is implemented by the control logic circuit, which continuously monitors the data collected by the sensors and compares it with the preset overcurrent threshold and overheat threshold. Once the monitored data exceeds the threshold, the control logic circuit immediately issues a switching signal, and through a switching device such as a relay or an electronic switch, switches the main drive circuit to the standby drive circuit. At the same time, the system will reset the control strategy parameters to adapt to the characteristics of the standby drive circuit, ensuring that the inflator pump can continue to operate stably. For example, the parameters of the standby drive circuit may be slightly different from those of the main drive circuit. Resetting the control strategy parameters can adjust the PWM duty cycle, motor speed control parameters, etc., to ensure that the motor can still work properly under the standby drive.

[0103] When constructing a fault mode library based on the time series clustering algorithm, a large amount of historical voltage fluctuation data is first collected. These data cover the voltage changes of the automotive inflator pump under various working conditions, including normal working states, different degrees of fault states, etc. The time series clustering algorithm divides the voltage fluctuation data with similar change trends into the same category to form different fault mode categories. Common time series clustering algorithms such as the dynamic time warping (DTW) clustering algorithm, whose core idea is to calculate the similarity distance between two time series and classify the sequences with close distances into the same category.

[0104] For each type of fault mode, its characteristic mode is extracted, and these characteristic modes are stored in the fault mode library as fault templates. For example, in a certain type of fault mode, the voltage fluctuation may show a periodic downward trend, and the downward amplitude and period have certain rules, and this rule is the characteristic mode of this fault mode. During actual operation, the dynamic time warping algorithm is used to calculate the similarity between the real-time signal and the fault template. The dynamic time warping algorithm finds the best matching path by elastically matching two time series on the time axis, thereby calculating the similarity between them. Suppose the real-time voltage fluctuation sequence is , and the fault template sequence is , and the similarity calculated by the DTW algorithm is . The smaller the similarity value, the more similar the two sequences are.

[0105] When the calculated similarity is lower than the preset threshold, the system determines that the real-time signal matches a certain fault template, indicating that there may be a corresponding fault. At this time, the system generates a preventive maintenance advice instruction according to the maintenance advice preset in the fault mode library. For example, if it is detected that the real-time voltage fluctuation is similar to a template indicating a motor winding short circuit fault, the system will issue an instruction to recommend checking and repairing the motor winding, so as to take measures before the fault deteriorates further, reduce the maintenance cost, and improve the reliability and service life of the automotive air pump.

[0106] An automotive air pump, comprising an air pump main body 1 made of high-strength engineering plastic. The interior of the air pump main body 1 adopts a modular design and is provided with an air pump assembly 2 and a detachable lithium battery pack 3;

[0107] The interior of the air pump main body 1 is also integrated with a tire pressure detection module that is connected to the air pump assembly 2 through a high-pressure air pipe air circuit. This tire pressure detection module uses a digital pressure sensor with a detection accuracy of ±0.1 PSI;

[0108] The tire pressure detection module is electrically connected to a high-brightness LED display module 4 through a PCB circuit board. An opening 30 for the display surface of the display module 4 to protrude is provided on the outer surface of the air pump main body 1, and a waterproof rubber ring is provided at the edge of the opening 30.

[0109] Furthermore, the exterior of the air pump main body 1 adopts an ergonomic design and is provided with a notch 10 for placing tool components and an inflation nozzle therein. The notch 10 is located on the side of the housing at a convenient operation position. The notch 10 is an embedded structure with a depth of 15 - 20 mm. A number of tool grooves for placing tool components and grooves for placing the inflation nozzle are respectively provided in the notch 10. The tool grooves include a special wrench groove, a tire repair tool groove, and a fuse groove, and the shapes of the grooves match the corresponding tools;

[0110] A detachable baffle 5 is provided outside the notch 10. The baffle 5 is fixed by a magnetic attraction method, and a shock-proof sponge layer is provided on the inner side of the baffle 5.

[0111] Preferably, the bottom of the air pump main body 1 is provided with an anti-slip rubber pad, and the top is provided with a portable handle; the battery pack 3 supports Type-C fast charging and is provided with a battery power indicator; the display module 4 can switch to display real-time tire pressure, preset tire pressure, and battery power information.

[0112] Its working principle can be divided into the following parts:

[0113] 1. Air pump inflation system

[0114] Air pump assembly 2: Driven by a motor, a piston or diaphragm structure compresses air to generate high-pressure air flow. The maximum air pressure can reach 150 PSI, suitable for inflating automotive tires, bicycles, balls, etc.

[0115] Inflation nozzle: Connected to the air pump through a high-pressure air pipe, using a standard valve nozzle interface to ensure a tight connection with the tire valve and prevent air leakage.

[0116] 2. Tire pressure detection and display system

[0117] Tire pressure detection module: Built-in digital pressure sensor (accuracy ±0.1 PSI), which monitors the tire pressure in real time and transmits the data to the control circuit.

[0118] Display module 4: Displays the real-time tire pressure, preset tire pressure, and battery level through a high-brightness LED screen. Users can adjust the inflation target value according to their needs to achieve precise inflation.

[0119] 3. Power supply system

[0120] Removable battery pack 3: Provides power supply, supports Type-C fast charging, and has a battery level indicator to facilitate users to know the remaining battery level.

[0121] 4. Auxiliary function design

[0122] Tool storage notch 10: Embedded structure, containing special wrenches, tire repair tools, fuses, etc. in grooves, convenient for emergency use.

[0123] Shock-proof baffle 5: Adopts a magnetic detachable design, with shock-proof sponge inside to protect tools and reduce vibration noise.

[0124] Workflow:

[0125] Startup detection: Connect the tire nozzle, and the tire pressure detection module automatically reads the current tire pressure and displays it.

[0126] Set target value: The user sets the required tire pressure (such as 36 PSI) through the display module.

[0127] Automatic inflation: The air pump starts and continues to inflate until the set value is reached and then stops automatically to avoid over-inflation.

[0128] Storage management: After use, the tools and nozzles can be put back into the notch, and the baffle closes to keep it tidy.

[0129] In summary, through the built-in tire pressure detection module, this inflator can not only inflate the tires, but also monitor the tire pressure to ensure that the tires are within the appropriate pressure range, improving driving safety. Moreover, the display module 4 is exposed through the opening 30, allowing users to directly view the tire pressure data outside the inflator, with simple and intuitive operation. At the same time, the notch 10 provides a convenient storage solution, giving a dedicated storage space for the tool components and the inflation nozzle, keeping the device clean and making the inflator more practical, enabling users to quickly access the required tools and nozzles.

[0130] Meanwhile, the battery pack 3 and the air pump assembly 2 are electrically connected through the electrical interface terminals 12, providing necessary power for the air pump assembly. The air pump assembly 2 and the tire pressure detection module are connected through an air path, enabling the air pump assembly 2 to supply compressed air to the tire pressure detection module to monitor and adjust the tire pressure. The tire pressure detection module and the display module 4 are electrically connected, and the tire pressure detection module can transmit the monitored tire pressure information to the display module 4 for users to view.

[0131] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article or device.

[0132] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A dynamic voltage monitoring and control system for an automotive air pump, characterized in that, The system includes: A voltage data acquisition module, which is used to collect vehicle power supply voltage, motor operating current and ambient temperature data through multi-channel sensors, and generate a multi-dimensional real-time monitoring data set; A dynamic noise suppression module, which is used to perform non-linear signal decomposition on the multi-dimensional real-time monitoring data set, and adopt a joint algorithm of empirical mode decomposition and Kalman filter to separate high-frequency noise components and steady-state voltage characteristics, and generate a denoised time series signal sequence; A voltage fluctuation modeling module, which is used to construct a deep generation model based on variational autoencoder, perform latent variable space mapping on the denoised time series signal sequence, extract non-linear correlation characteristics of voltage transient fluctuations and load changes, and generate a dynamic voltage state coding vector; An adaptive compensation control module, which is used to design a policy optimization network based on reinforcement learning, dynamically adjust the PWM duty cycle and motor speed control parameters according to the dynamic voltage state coding vector, and generate an anti-disturbance voltage stabilization instruction.

2. The dynamic voltage monitoring and control system of an automotive air pump as claimed in claim 1, wherein, The non-linear signal decomposition includes: Performing empirical mode decomposition on the vehicle power supply voltage signal, and generating a set of intrinsic mode function components through local extreme point interpolation; Constructing a Kalman filter state equation, inputting high-frequency intrinsic mode components as observation noise, and iteratively updating the system state estimate value; Adopting a sliding window entropy value detection algorithm to identify the mutation points of steady-state voltage components and segment continuous steady signal segments.

3. The dynamic voltage monitoring and control system of an automotive air pump according to claim 1, characterized in that The construction of the deep generation model includes: Introducing a causal convolutional layer in the encoder to capture the temporal causal relationship of the voltage sequence and output the parameters of the latent variable probability distribution; Adopting a gated recurrent unit in the decoder to reconstruct the voltage fluctuation waveform, and superimposing an attention mechanism to focus on the characteristics at the moment of load mutation; Optimizing the parameters of the encoder and decoder through an adversarial training strategy, and minimizing the weighted sum of the reconstruction error and the discriminator loss function.

4. The dynamic voltage monitoring and control system of an automotive air pump as claimed in claim 1, wherein, The design of the policy optimization network includes: Defining the state space as a multi-dimensional vector of voltage ripple coefficient, current harmonic distortion rate and temperature compensation factor; Constructing a double-delay deep deterministic policy gradient algorithm framework, and adopting a target network and an experience replay pool to optimize the convergence of the control strategy; Designing the reward function as a linear combination of the reciprocal of the absolute value of voltage deviation and the energy loss index, and driving the policy network to generate an optimal compensation instruction.

5. The dynamic voltage monitoring and control system of an automotive air pump according to claim 1, characterized in that, The system further includes: A digital twin simulation module, which is used to establish a motor-air pump coupling model based on physical equations and real-time simulate the dynamic relationship between the air chamber pressure and the motor torque; Inputting the measured voltage data into the simulation model to generate a virtual pressure curve, and triggering an air circuit leakage warning signal through residual analysis.

6. The dynamic voltage monitoring and control system of an automotive air pump according to claim 1, characterized in that The generation of the multi-dimensional real-time monitoring data set includes: Synchronously collecting power bus CAN signals, Hall sensor current pulses and NTC thermistor temperature data; Adopting a tensor decomposition algorithm to fill in the missing values of heterogeneous data and construct a three-dimensional spatio-temporal data matrix; Eliminating the baseline drift of the sensor through the moving average difference method and enhancing the identifiability of data transient characteristics.

7. The dynamic voltage monitoring and control system of an automotive air pump according to claim 1, characterized in that The system further includes: Constructing a distributed update module based on federated learning, and jointly training a voltage prediction model through vehicle-mounted edge devices and cloud servers; Design a differential privacy mechanism to encrypt local gradient parameters, and use a model aggregation algorithm to generate a globally shared compensation control strategy.

8. The dynamic voltage monitoring and control system of an automotive air pump according to claim 1, characterized in that, The generation of the anti-disturbance voltage stabilization instruction includes: Establish a multi-objective optimization model with voltage stability, motor life loss, and energy efficiency as constraints; Use the non-smooth Newton method to solve the optimization problem with inequality constraints, and generate the Pareto front solution set of the PWM frequency and duty cycle; Select the optimal compromise solution through the fuzzy decision layer and output the control parameters of the pulse width modulation signal; The system also includes: Design a redundant protection circuit based on optocoupler isolation to monitor the junction temperature and switching loss of the IGBT module in real time; When an overcurrent or overheat state is detected, automatically switch to the standby drive circuit and reset the control strategy parameters; The system also includes: Construct a fault mode library based on the time series clustering algorithm to perform feature pattern matching on historical voltage fluctuation data; Use the dynamic time warping algorithm to calculate the similarity between the real-time signal and the fault template, and generate preventive maintenance recommendation instructions.

9. An automotive air pump, comprising an air pump main body (1), characterized in that, Inside the air pump main body (1), there is an air pump assembly (2) and a battery pack (3). Inside the air pump main body (1), there is also a tire pressure detection module that is pneumatically connected to the air pump assembly (2). The tire pressure detection module is electrically connected to a display module (4). An opening (30) for the display surface of the display module (4) to protrude is provided on the outer surface of the air pump main body (1). A notch (10) for placing tool parts and inflation nozzles is also provided on the outside of the air pump main body (1).

10. The automotive air pump according to claim 9, characterized in that, The notch (10) is an embedded structure. Inside the notch (10), there are also several tool grooves for placing tool parts and grooves for placing inflation nozzles. A detachable baffle (5) is provided outside the notch (10).

Images