Uniform magnetic field design system and method based on intelligent perception

Abstract

The invention discloses a uniform magnetic field design system and method based on intelligent perception, and relates to the technical field of uniform magnetic fields. Local deviation of a magnetic field is detected and analyzed in real time through an intelligent perception system and a deep reinforcement learning algorithm, and refined magnetic field compensation is realized by dynamically adjusting and compensating coil current; pID (Proportion Integration Differentiation) control is combined with intelligent feedback, a rapidly changing local magnetic field is rapidly and accurately adjusted, local nonuniformity, thermal noise and a potential interference source are identified through quantum simulation analysis, coil parameters are optimized in combination with deep reinforcement learning, and current and coil spacing are adjusted to maximize the uniformity of the magnetic field. The uniformity and the stability of the magnetic field are remarkably improved, meanwhile, the resistance to thermal noise and external disturbance is enhanced by monitoring and analyzing the change of the magnetic field in real time, recognizing an external interference source and performing compensation adjustment, the dependence on manual adjustment is greatly reduced, and different magnetic field requirements can be quickly met.

Description

Technical Field

[0001] The present invention relates to the technical field of uniform magnetic field, and in particular to a uniform magnetic field design system and method based on intelligent perception. Background Art

[0002] Magnetic field simulation devices have important application value in aerospace engineering such as geomagnetic navigation. An ideal magnetic field simulation device should generate a highly uniform magnetic field in a sufficiently large spatial range. Therefore, the uniformity of the magnetic field has become a key indicator for evaluating the performance of magnetic field simulation devices. In recent years, with the widespread application of magnetic field simulation devices in various fields, the uniformity of their magnetic field distribution has attracted more and more attention from scientific researchers. The Helmholtz coil is one of the earliest classic designs used to generate a uniform magnetic field. Its principle is to generate a uniform magnetic field by passing current in the same direction through two sets of coaxial coils with appropriate spacing. However, as the distance from the center position increases, the uniformity of the magnetic field decreases rapidly, limiting its application in large space scenes. In response to the shortcomings of the Helmholtz coil, scholars have proposed a variety of improvement plans, such as using a three-coil or four-coil array to keep the magnetic field uniform in a larger space, expand the range of the uniform magnetic field, and change the coil spacing or the coil diameter ratio to improve the magnetic field uniformity.

[0003] Current uniform magnetic field design methods gradually introduce intelligent sensing and optimization algorithms to improve the uniformity and stability of the magnetic field. However, in large-scale coil groups, current uniform magnetic field design methods often have edge effects, thermal noise and external interference, resulting in local magnetic field unevenness. Existing intelligent sensing methods mainly rely on preset sensing points and compensation mechanisms, and are not precise enough in responding to local changes. At the same time, the optimization algorithms used, such as genetic algorithms and particle swarm optimization, are difficult to quickly adapt to instantaneous changes in the magnetic field when facing real-time data due to computational complexity and algorithm response time issues, and the adjustment efficiency is low under the condition of multi-variable coupling.

[0004] In order to improve the response accuracy of local changes, the traditional solution adopts adding compensation coils to adjust the local magnetic field distribution, or manually adjusts the local magnetic field. However, the compensation method lacks real-time performance and cannot effectively compensate for the rapidly changing local magnetic field. Manual adjustment has a slow response speed and low accuracy. In order to cope with magnetic field changes, the traditional solution optimizes parameters through a large number of offline simulations, but these parameters lack flexible adjustment capabilities during actual operation. Some methods of adjusting parameters using preset adjustment curves cannot dynamically respond to real-time perception data, resulting in limited adjustment effects. Therefore, there is an urgent need for a uniform magnetic field system based on intelligent perception and its design method to solve such problems. Summary of the invention

[0005] In view of the above existing problems, the present invention is proposed.

[0006] Therefore, the present invention provides a uniform magnetic field design system and method based on intelligent sensing to solve the problems of slow response speed, poor dynamic adaptability and insufficient optimization accuracy in local compensation, real-time adjustment and fine optimization of the existing uniform magnetic field design based on intelligent sensing.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] In one aspect, the present invention provides a uniform magnetic field system based on intelligent sensing, comprising:

[0009] The intelligent sensing system, based on the sensor array deployed in the magnetic field area, collects multi-dimensional data including magnetic field strength, direction, temperature and external interference in real time, and uses data fusion technology to integrate and denoise the data of multiple sensors. It is also responsible for transmitting the collected data to the compensation module and optimization module in real time;

[0010] The sensor array consists of multiple sensors distributed in the magnetic field area, which is used to collect magnetic field parameters at multiple points and monitor environmental parameters;

[0011] Quantum simulation system, used to simulate and analyze the collected magnetic field data at the microscopic level, dynamically simulate the evolution of the magnetic field in different time and space, and identify magnetic field inhomogeneity and potential interference sources;

[0012] The compensation module adjusts the current or position of the local compensation coil in real time according to the feedback from the intelligent sensing system;

[0013] The optimization module uses deep reinforcement learning to optimize the system coil design and compensation strategy. The optimization adjustment is made in real time based on real-time feedback.

[0014] On the other hand, the present invention provides a design method of a uniform magnetic field system based on intelligent perception, comprising:

[0015] Step S1: Intelligent sensing system deployment and data collection.

[0016] Deploy intelligent perception systems and sensor arrays to build a magnetic field distribution model. The sensor array collects multi-dimensional perception data including magnetic field strength, direction, and ambient temperature in real time. The perception data is processed using multi-sensor fusion technology, and denoised and standardized.

[0017] Step S2, microscopic analysis of magnetic field based on quantum simulation,

[0018] The pre-processed sensing data is input into the quantum simulation system for microscopic analysis, simulating the dynamic evolution of the magnetic field, identifying local inhomogeneous areas, thermal noise influence points and potential interference sources in the magnetic field, and outputting local inhomogeneity and interference source distribution areas.

[0019] Step S3, deep reinforcement learning algorithm optimization,

[0020] Based on the results of quantum simulation analysis, the coil parameters including current and coil spacing are optimized using deep reinforcement learning algorithms.

[0021] Output optimized coil parameters, form a preliminary optimization plan, and establish a fast response model;

[0022] Step S4: local real-time dynamic compensation.

[0023] Arrange local compensation modules in key areas identified by quantum simulation analysis and connect them to the intelligent perception system.

[0024] The intelligent perception system feeds back the local deviation data in the key area to the compensation module.

[0025] According to the feedback data, the compensation module dynamically adjusts the current of the compensation coil to perform local magnetic field compensation, thereby achieving real-time response and correction to the rapidly changing local magnetic field;

[0026] Step S5, finite element simulation and closed-loop verification,

[0027] The optimized coil parameters and compensation strategy are input into the finite element simulation software for overall verification.

[0028] Simulate and analyze the uniformity, stability and resistance to external interference of the magnetic field.

[0029] The simulation results are fed back to the optimization module. If the uniformity index is not met, return to step S3 to repeatedly adjust the optimization parameters for iterative optimization.

[0030] Furthermore, the magnetic field distribution model is constructed in step S1 as follows:

[0031] The data collected by the sensor are spatially interpolated and integrated. Assume that the sensor array consists of N sensors, and the data collected by each sensor i is the magnetic field intensity B i , direction θ i and ambient temperature T i , Where B(x,y,z) represents the magnetic field intensity at the position (x,y,z) in the three-dimensional coordinate system, and w i represents the weight factor, that is, the contribution of each sensor to the location. The weight is determined based on the distance between the sensor and the location and its measurement accuracy. i represents the magnetic field strength measurement value of the i-th sensor;

[0032] Weight Among them, α represents the attenuation factor, di represents the distance from the i-th sensor to the position (x, y, z), Where (x i ,y i ,z i ) is the position coordinate of sensor i, and represents the contribution of the i-th and j-th sensors to the (x, y, z) position, and the weight increases with the distance d i and d j Increase and decrease rapidly;

[0033] Perform weighted fusion. in, represents the magnetic field strength after fusion, β i represents the fusion weight, weight Among them, σ i represents the standard deviation of the measurement noise of sensor i;

[0034] Use Kalman filtering to smooth the data and perform normalization. The data smoothing method is:

[0035] in, represents the estimated magnetic field strength at time k, represents the estimated magnetic field strength at time k-1, K k Represents the Kalman gain, which indicates the degree of correction of the estimated value by the current measurement. Among them, P k-1 represents the covariance of the prediction error, and R represents the covariance of the measurement noise;

[0036] standardization: Where B′ represents the standardized magnetic field intensity, μ B represents the average value of magnetic field strength, σ B Represents the standard deviation of the magnetic field strength.

[0037] Furthermore, in step S2, the quantum simulation system performs microscopic analysis of the magnetic field as follows:

[0038] The preprocessed sensory data Input the quantum simulation system for initialization, where N is the number of sensors. During initialization, the data is mapped to the quantum state space;

[0039] Where ψ0 represents the initial quantum state, that is, the initial distribution of the magnetic field, N represents the number of sensors, represents the phase factor, φ i Dependent on the sensor position and magnetic field direction, |B i ,θ i ,Ti > represents the data collected by sensor i;

[0040] The dynamic evolution of the magnetic field is simulated by using the time-dependent Schrödinger equation to simulate the dynamic evolution of the magnetic field. The evolution of the quantum state is controlled by the Hamiltonian H(t) to simulate the time variation and disturbance effects of the magnetic field. The evolution formula is: Among them, ψ(t) represents the quantum state at time t, represents the reduced Planck constant, represents partial derivative;

[0041] H(t)=H0+H disturb (t), represents the total Hamiltonian, which is composed of the initial system Hamiltonian H0 and the perturbation term H disturb (t) is composed of the disturbance term H disturb (t) represents the dynamic influence of thermal noise and external interference,

[0042] Where M represents the number of disturbance sources, γ k (t) represents the disturbance intensity, which varies with time, σ k represents the Pauli matrix, which represents the local perturbation form of the magnetic field.

[0043] Furthermore, in step S2, the microscopic analysis method of the magnetic field by the quantum simulation system also includes:

[0044] To identify inhomogeneities, thermal noise and interference sources, the probability density function P(x, y, z) of the magnetic field distribution is calculated based on the local evolution of the observed state, and the inhomogeneous areas and interference sources are identified. The probability density P(x, y, z) = |<x,y,z|ψ(t)> | 2 , where P(x,y,z) represents the probability of magnetic field distribution at position (x,y,z), and |x,y,z> represents the position state in spatial coordinates;

[0045] Inhomogeneity identification ΔB(x,y,z)=|B target -∫ V P(x,y,z)B(x,y,z)dV|, where ΔB(x,y,z) represents the deviation from the target magnetic field at the position (x,y,z), and B target represents the target magnetic field strength, ∫ V Indicates that the volume V is integrated to calculate the average value of the local magnetic field distribution, thereby identifying the distribution area of ​​local inhomogeneity and interference sources.

[0046] Furthermore, in step S3, based on the results of quantum simulation analysis, the coil parameters are optimized by deep reinforcement learning as follows:

[0047] The quantum simulation results are used as the environment state input for reinforcement learning. The initial parameters are set, including the current I0 and the coil spacing d0, and the initial magnetic field uniformity index is calculated. The state space is defined as:

[0048] s t ={B(x,y,z),ΔB(x,y,z),I,d}, where, s t represents the state vector, that is, the system state at the current time t, B(x,y,z) represents the magnetic field strength at the position (x,y,z), ΔB(x,y,z) represents the magnetic field inhomogeneity deviation at the position (x,y,z), I represents the current coil current, and d represents the current coil spacing;

[0049] Using the policy network π(a t |s t )Select action a t , optimize the coil parameters, maximize the reward function R, and the strategy update formula is a t =π θ (s t ), where a t Indicates the action at the current moment, that is, the adjustment value of the current I and the coil spacing d, π θ represents the policy network, with parameter θ;

[0050] The reward function is R = -α∑ (x,y,z) |ΔB(x,y,z)|-β(I 2 +d 2 ), where R represents the reward value, the goal is to maximize R, α and β represent weight parameters, which control the influence of uniformity and parameter adjustment respectively, and |ΔB(x,y,z)| represents the absolute value of non-uniformity;

[0051] According to the current strategy π θ , execute the action to adjust the current and coil spacing, and obtain the new state s t+1 and reward R, the state transition formula is s t+1 =f(s t ,a t ), where s t+1 represents the new state after executing the action, and f represents the state transition function.

[0052] Furthermore, in step S3, the method of optimizing coil parameters by using deep reinforcement learning also includes:

[0053] Using the gradient boosting strategy network parameter θ, the strategy gradient update formula is Among them, θ t+1 represents the updated policy parameter, η represents the learning rate, represents the expected gradient of reward;

[0054] A fast response model is established. When the environment changes, the coil parameters are adjusted. The fast response model is: in, represents the estimated action of fast response, g represents the fast response network, which adjusts the response speed to the magnetic field state through training, optimizes the current and spacing of the coils using deep reinforcement learning methods, forms a preliminary optimization plan, and establishes a fast response model to realize real-time parameter adjustment.

[0055] Furthermore, in step S4, the key area is the local inhomogeneity and interference source distribution area.

[0056] Further, in step S4, the current of the compensation coil is dynamically adjusted according to the feedback data to perform local magnetic field compensation in the following manner:

[0057] The intelligent sensing system feeds back the data to the compensation control module based on the local magnetic field deviation ΔB(x,y,z) collected in real time. The deviation formula is ΔB(x,y,z)=B measured (x,y,z)-B target , where ΔB(x,y,z) represents the magnetic field deviation at position (x,y,z), B measured (x, y, z) represents the actual measured magnetic field strength, B target Indicates the target magnetic field strength;

[0058] According to the feedback magnetic field deviation, the current adjustment amount ΔI that needs to be compensated is calculated. The compensation current and the deviation are controlled by proportional integral differential PID. Where ΔI(t) represents the compensation current adjustment at time t, K p ,K i ,K d are proportional, integral and differential gain coefficients respectively, ΔB(t) represents the magnetic field deviation at the current moment, represents the integral of the magnetic field deviation, that is, the cumulative deviation, It represents the derivative of the magnetic field deviation, i.e. the rate of change;

[0059] The calculated current adjustment ΔI(t) is applied to the compensation coil, and the adjusted current I(t) is used to compensate for the magnetic field deviation. The current adjustment process is expressed as I(t)=I base +ΔI(t), where I(t) represents the current of the compensation coil at the current moment, I base Indicates the initial setting current of the compensation coil.

[0060] Furthermore, in step S4, the local magnetic field compensation method further includes:

[0061] Calculate the adjusted local magnetic field B based on the adjusted current I(t) comp (x, y, z), the compensation effect is fed back to the intelligent sensing system, and the magnetic field compensation formula is: Among them, B comp (x, y, z) represents the magnetic field strength after compensation, k represents a constant, R represents the radius of the compensation coil, and r represents the distance from the center of the coil to the compensation position;

[0062] Continuously monitor ΔB(x,y,z) and adjust the compensation current according to the latest deviation. The dynamic feedback method is:

[0063] ΔB new =ΔB old -B comp (x,y,z), where ΔB new Indicates the new magnetic field deviation after compensation, ΔB old Indicates the magnetic field deviation before compensation; dynamically adjusts the current of the compensation coil according to the real-time feedback data to achieve fast and accurate compensation of the local magnetic field, thereby improving the overall uniformity and stability of the magnetic field.

[0064] The beneficial effects of the present invention are:

[0065] The present invention detects and analyzes the local deviation of the magnetic field in real time through an intelligent sensing system and a deep reinforcement learning algorithm, dynamically adjusts the compensation coil current to achieve refined magnetic field compensation, combines PID control with intelligent feedback, and makes rapid and accurate adjustments to the rapidly changing local magnetic field.

[0066] The present invention identifies local inhomogeneities, thermal noise and potential interference sources through quantum simulation analysis, optimizes coil parameters in combination with deep reinforcement learning, adjusts current and coil spacing to maximize the uniformity of the magnetic field, and significantly improves the uniformity and stability of the magnetic field.

[0067] The present invention monitors and analyzes the magnetic field changes in real time, identifies external interference sources and performs compensation adjustments, thereby enhancing the ability to resist thermal noise and external disturbances.

[0068] The present invention greatly reduces the reliance on manual adjustment, and forms a fully automatic parameter adjustment and optimization process through an intelligent sensing system, reinforcement learning optimization and a rapid response model, which can quickly adapt to different magnetic field requirements. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0070] Figure 1 It is a schematic diagram of the structure of the uniform magnetic field design system of the present invention;

[0071] Figure 2 It is a schematic flow chart of the design method of the uniform magnetic field design system of the present invention. DETAILED DESCRIPTION

[0072] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0073] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0074] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0075] Example 1, reference Figure 1 , this embodiment provides a uniform magnetic field design system based on intelligent perception, including:

[0076] The intelligent sensing system, based on the sensor array deployed in the magnetic field area, collects multi-dimensional data including magnetic field strength, direction, temperature and external interference in real time, and uses data fusion technology to integrate and denoise the data of multiple sensors. It is also responsible for transmitting the collected data to the compensation module and optimization module in real time;

[0077] The sensor array consists of multiple sensors distributed in the magnetic field area, which is used to collect magnetic field parameters at multiple points and monitor environmental parameters;

[0078] Quantum simulation system, used to simulate and analyze the collected magnetic field data at the microscopic level, dynamically simulate the evolution of the magnetic field in different time and space, and identify magnetic field inhomogeneity and potential interference sources;

[0079] The compensation module adjusts the current or position of the local compensation coil in real time according to the feedback from the intelligent sensing system;

[0080] The optimization module uses deep reinforcement learning to optimize the system coil design and compensation strategy. The optimization adjustment is made in real time based on real-time feedback.

[0081] Example 2, reference Figure 2 This embodiment provides a design method for a uniform magnetic field design system based on intelligent perception, comprising the following steps:

[0082] Step S1: Intelligent sensing system deployment and data collection.

[0083] Deploy intelligent perception systems and sensor arrays to build a magnetic field distribution model. The sensor array collects multi-dimensional perception data including magnetic field strength, direction, and ambient temperature in real time. The perception data is processed using multi-sensor fusion technology, and denoised and standardized.

[0084] The method of constructing the magnetic field distribution model in step S1 is as follows:

[0085] The data collected by the sensor are spatially interpolated and integrated. Assume that the sensor array consists of N sensors, and the data collected by each sensor i is the magnetic field intensity B i , direction θ i and ambient temperature T i , Where B(x,y,z) represents the magnetic field intensity at the position (x,y,z) in the three-dimensional coordinate system, and w i represents the weight factor, that is, the contribution of each sensor to the location. The weight is determined based on the distance between the sensor and the location and its measurement accuracy. i represents the magnetic field strength measurement value of the i-th sensor;

[0086] Weight Among them, α represents the attenuation factor, d i represents the distance from the i-th sensor to the position (x, y, z), Where (x i ,y i ,z i ) is the position coordinate of sensor i, and represents the contribution of the i-th and j-th sensors to the (x, y, z) position, and the weight increases with the distance d i and d j Increase and decrease rapidly;

[0087] Perform weighted fusion. in, represents the magnetic field strength after fusion, β i represents the fusion weight, weight Among them, σ i represents the standard deviation of the measurement noise of sensor i;

[0088] Use Kalman filtering to smooth the data and perform normalization. The data smoothing method is:

[0089] in, represents the estimated magnetic field strength at time k, represents the estimated magnetic field strength at time k-1, K k Represents the Kalman gain, which indicates the degree of correction of the estimated value by the current measurement. Among them, P k-1 represents the covariance of the prediction error, and R represents the covariance of the measurement noise;

[0090] standardization: Where B′ represents the standardized magnetic field intensity, μ B represents the average value of magnetic field strength, σ B represents the standard deviation of magnetic field strength;

[0091] Specifically, the intelligent sensing system is used to collect multi-dimensional data of the magnetic field in real time, and the microscopic evolution of the magnetic field is analyzed in combination with quantum simulation to effectively identify local uneven areas, thermal noise influencing points and interference sources. The coil parameters are optimized through a deep reinforcement learning algorithm, and it learns how to adjust the current and coil spacing to maximize the uniformity of the magnetic field, thereby achieving real-time response and compensation of the local magnetic field, reducing unevenness and improving the stability of the system.

[0092] Step S2, microscopic analysis of magnetic field based on quantum simulation,

[0093] The pre-processed sensory data is input into the quantum simulation system for microscopic analysis.

[0094] Simulate the dynamic evolution of the magnetic field, identify local inhomogeneous areas, thermal noise influence points and potential interference sources in the magnetic field,

[0095] Output local inhomogeneity and interference source distribution area,

[0096] In step S2, the quantum simulation system performs microscopic analysis of the magnetic field as follows:

[0097] The preprocessed sensory data Input the quantum simulation system for initialization, where N is the number of sensors. During initialization, the data is mapped to the quantum state space;

[0098] Where ψ0 represents the initial quantum state, that is, the initial distribution of the magnetic field, N represents the number of sensors, represents the phase factor, φ i Dependent on the sensor position and magnetic field direction, |B i ,θ i ,Ti > represents the data collected by sensor i;

[0099] The dynamic evolution of the magnetic field is simulated by using the time-dependent Schrödinger equation to simulate the dynamic evolution of the magnetic field. The evolution of the quantum state is controlled by the Hamiltonian H(t) to simulate the time variation and disturbance effects of the magnetic field. The evolution formula is: Among them, ψ(t) represents the quantum state at time t, represents the reduced Planck constant, represents partial derivative;

[0100] H(t)=H0+H disturb (t), represents the total Hamiltonian, which is composed of the initial system Hamiltonian H0 and the perturbation term H disturb (t) is composed of the disturbance term H disturb (t) represents the dynamic influence of thermal noise and external interference,

[0101] Where M represents the number of disturbance sources, γ k (t) represents the disturbance intensity, which varies with time, σ k represents the Pauli matrix, which represents the local perturbation form of the magnetic field;

[0102] The microscopic analysis method of the magnetic field by the quantum simulation system in step S2 also includes:

[0103] To identify inhomogeneities, thermal noise and interference sources, the probability density function P(x, y, z) of the magnetic field distribution is calculated based on the local evolution of the observed state, and the inhomogeneous areas and interference sources are identified. The probability density P(x, y, z) = |<x,y,z|ψ(t)> | 2 , where P(x,y,z) represents the probability of magnetic field distribution at position (x,y,z), and |x,y,z> represents the position state in spatial coordinates;

[0104] Inhomogeneity identification ΔB(x,y,z)=|B target -∫ V P(x,y,z)B(x,y,z)dV|, where ΔB(x,y,z) represents the deviation from the target magnetic field at the position (x,y,z), and B target represents the target magnetic field strength, ∫ V It means integrating the volume V and calculating the average value of the local magnetic field distribution, so as to identify the distribution area of ​​local inhomogeneity and interference source;

[0105] Specifically, through intelligent sensing systems and deep reinforcement learning, the coil parameters and compensation strategies are dynamically adjusted according to real-time data feedback, and PID control is used to accurately adjust the current to achieve dynamic compensation for rapid changes in the local magnetic field. The dynamic response enables adaptation to sudden changes in the environment and maintains control of the magnetic field uniformity in multiple variable coupling scenarios, overcoming the lag and inaccurate adjustment problems of traditional methods when faced with rapid changes.

[0106] Step S3, deep reinforcement learning algorithm optimization,

[0107] Based on the results of quantum simulation analysis, the coil parameters including current and coil spacing are optimized using deep reinforcement learning algorithms.

[0108] Output optimized coil parameters, form a preliminary optimization plan, and establish a fast response model;

[0109] In step S3, based on the results of quantum simulation analysis, the coil parameters are optimized using deep reinforcement learning as follows:

[0110] The quantum simulation results are used as the environment state input for reinforcement learning. The initial parameters are set, including the current I0 and the coil spacing d0, and the initial magnetic field uniformity index is calculated. The state space is defined as:

[0111] s t ={B(x,y,z),ΔB(x,y,z),I,d}, where, s t represents the state vector, that is, the system state at the current time t, B(x,y,z) represents the magnetic field strength at the position (x,y,z), ΔB(x,y,z) represents the magnetic field inhomogeneity deviation at the position (x,y,z), I represents the current coil current, and d represents the current coil spacing;

[0112] Using the policy network π(a t |s t )Select action a t , optimize the coil parameters, maximize the reward function R, and the strategy update formula is a t =π θ (s t ), where a t Indicates the action at the current moment, that is, the adjustment value of the current I and the coil spacing d, π θ represents the policy network, with parameter θ;

[0113] The reward function is R = -α∑ (x,y,z) |ΔB(x,y,z)|-β(i 2 +d 2), where R represents the reward value, the goal is to maximize R, α and β represent weight parameters, which control the influence of uniformity and parameter adjustment respectively, and |ΔB(x,y,z)| represents the absolute value of non-uniformity;

[0114] According to the current strategy π θ , execute the action to adjust the current and coil spacing, and obtain the new state s t+1 and reward R, the state transition formula is s t+1 =f(s t ,a t ), where s t+1 represents the new state after executing the action, and f represents the state transition function;

[0115] In step S3, the method of optimizing coil parameters by using deep reinforcement learning also includes:

[0116] Using the gradient boosting strategy network parameter θ, the strategy gradient update formula is Among them, θ t+1 represents the updated policy parameter, η represents the learning rate, represents the expected gradient of reward;

[0117] A fast response model is established. When the environment changes, the coil parameters are adjusted. The fast response model is: in, represents the estimated action of fast response, g represents the fast response network, which adjusts the response speed to the magnetic field state through training, optimizes the current and spacing of the coils using deep reinforcement learning methods, forms a preliminary optimization plan, and establishes a fast response model to achieve real-time parameter adjustment;

[0118] Specifically, the quantum simulation system's analysis of microscopic disturbances in the magnetic field enables the system to identify in advance interference sources that may cause inhomogeneities. Combined with real-time adjustment of the compensation module, it can effectively suppress the impact of external interference on the magnetic field distribution.

[0119] Step S4: local real-time dynamic compensation.

[0120] Arrange local compensation modules in key areas identified by quantum simulation analysis and connect them to the intelligent perception system.

[0121] The intelligent perception system feeds back the local deviation data in the key area to the compensation module.

[0122] According to the feedback data, the compensation module dynamically adjusts the current of the compensation coil to perform local magnetic field compensation, thereby achieving real-time response and correction to the rapidly changing local magnetic field;

[0123] In step S4, the key area is the local inhomogeneity and interference source distribution area;

[0124] In step S4, the current of the compensation coil is dynamically adjusted according to the feedback data to perform local magnetic field compensation in the following manner:

[0125] The intelligent sensing system feeds back the data to the compensation control module based on the local magnetic field deviation ΔB(x,y,z) collected in real time. The deviation formula is ΔB(x,y,z)=B measured (x,y,z)-B target , where ΔB(x,y,z) represents the magnetic field deviation at position (x,y,z), B measured (x, y, z) represents the actual measured magnetic field strength, B target Indicates the target magnetic field strength;

[0126] According to the feedback magnetic field deviation, the current adjustment amount ΔI that needs to be compensated is calculated. The compensation current and the deviation are controlled by proportional integral differential PID. Where ΔI(t) represents the compensation current adjustment at time t, K p ,K i ,K d are proportional, integral and differential gain coefficients respectively, ΔB(t) represents the magnetic field deviation at the current moment, represents the integral of the magnetic field deviation, that is, the cumulative deviation, It represents the derivative of the magnetic field deviation, i.e. the rate of change;

[0127] The calculated current adjustment ΔI(t) is applied to the compensation coil, and the adjusted current I(t) is used to compensate for the magnetic field deviation. The current adjustment process is expressed as I(t)=I base +ΔI(t), where I(t) represents the current of the compensation coil at the current moment, I base Indicates the initial setting current of the compensation coil;

[0128] In step S4, the local magnetic field compensation method further includes:

[0129] Calculate the adjusted local magnetic field B based on the adjusted current I(t) comp (x, y, z), the compensation effect is fed back to the intelligent sensing system, and the magnetic field compensation formula is: Among them, B comp (x, y, z) represents the magnetic field strength after compensation, k represents a constant, R represents the radius of the compensation coil, and r represents the distance from the center of the coil to the compensation position;

[0130] Continuously monitor ΔB(x,y,z) and adjust the compensation current according to the latest deviation. The dynamic feedback method is:

[0131] ΔB new =ΔBold -B comp (x,y,z), where ΔB new Indicates the new magnetic field deviation after compensation, ΔB old Indicates the magnetic field deviation before compensation; dynamically adjusts the current of the compensation coil according to the real-time feedback data to achieve fast and accurate compensation of the local magnetic field, thereby improving the overall uniformity and stability of the magnetic field;

[0132] Specifically, the current and spacing of the coils are continuously optimized through deep reinforcement learning algorithms, which reduces a lot of manual debugging and offline simulation time, greatly improves design efficiency, and the established fast response model can make rapid adjustments when the environment changes, thereby reducing the response time to magnetic field inhomogeneity.

[0133] Step S5, finite element simulation and closed-loop verification,

[0134] The optimized coil parameters and compensation strategy are input into the finite element simulation software for overall verification.

[0135] Simulate and analyze the uniformity, stability and resistance to external interference of the magnetic field.

[0136] The simulation results are fed back to the optimization module. If the uniformity index is not met, return to step S3 to repeatedly adjust the optimization parameters for iterative optimization.

[0137] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A uniform magnetic field design system based on intelligent perception, characterized in that: include, The intelligent sensing system, based on the sensor array deployed in the magnetic field area, collects multi-dimensional data including magnetic field strength, direction, temperature and external interference in real time, and uses data fusion technology to integrate and denoise the data of multiple sensors. It is also responsible for transmitting the collected data to the compensation module and optimization module in real time; The sensor array consists of multiple sensors distributed in the magnetic field area, which is used to collect magnetic field parameters at multiple points and monitor environmental parameters; Quantum simulation system, used to simulate and analyze the collected magnetic field data at the microscopic level, dynamically simulate the evolution of the magnetic field in different time and space, and identify magnetic field inhomogeneity and potential interference sources; The compensation module adjusts the current or position of the local compensation coil in real time according to the feedback from the intelligent sensing system; The optimization module uses deep reinforcement learning to optimize the system coil design and compensation strategy. The optimization adjustment is made in real time based on real-time feedback.

2. A design method for a uniform magnetic field design system based on intelligent perception, characterized in that: Based on the uniform magnetic field design system based on intelligent perception as described in claim 1, The following steps are included: Step S1: Intelligent sensing system deployment and data collection. Deploy intelligent perception systems and sensor arrays to build a magnetic field distribution model. The sensor array collects multi-dimensional perception data including magnetic field strength, direction, and ambient temperature in real time. The perception data is processed using multi-sensor fusion technology, and denoised and standardized. Step S2, microscopic analysis of magnetic field based on quantum simulation, The pre-processed sensing data is input into the quantum simulation system for microscopic analysis, simulating the dynamic evolution of the magnetic field, identifying local inhomogeneous areas, thermal noise influence points and potential interference sources in the magnetic field, and outputting local inhomogeneity and interference source distribution areas. Step S3, deep reinforcement learning algorithm optimization, Based on the results of quantum simulation analysis, the coil parameters including current and coil spacing are optimized using deep reinforcement learning algorithms. Output optimized coil parameters, form a preliminary optimization plan, and establish a fast response model; Step S4: local real-time dynamic compensation. Arrange local compensation modules in key areas identified by quantum simulation analysis and connect them to the intelligent perception system. The intelligent perception system feeds back the local deviation data in the key area to the compensation module. According to the feedback data, the compensation module dynamically adjusts the current of the compensation coil to perform local magnetic field compensation; Step S5, finite element simulation and closed-loop verification, The optimized coil parameters and compensation strategy are input into the finite element simulation software for overall verification. Simulate and analyze the uniformity, stability and resistance to external interference of the magnetic field. The simulation results are fed back to the optimization module. If the uniformity index is not met, return to step S3 to repeatedly adjust the optimization parameters for iterative optimization.

3. The design method of a uniform magnetic field design system based on intelligent perception according to claim 2, characterized in that: The method of constructing the magnetic field distribution model in step S1 is as follows: The data collected by the sensor are spatially interpolated and integrated. Assume that the sensor array consists of N sensors, and the data collected by each sensor i is the magnetic field intensity B i , direction θ i and ambient temperature T i , Where B(x,y,z) represents the magnetic field intensity at the position (x,y,z) in the three-dimensional coordinate system, and w i represents the weight factor, that is, the contribution of each sensor to the location. The weight is determined based on the distance between the sensor and the location and its measurement accuracy. i represents the magnetic field strength measurement value of the i-th sensor; Weight Among them, α represents the attenuation factor, d i represents the distance from the i-th sensor to the position (x, y, z), Where (x i ,y i ,z i ) is the position coordinate of sensor i, and represents the contribution of the i-th and j-th sensors to the (x, y, z) position, and the weight increases with the distance d i and d j Increase and decrease rapidly; Perform weighted fusion. in, represents the magnetic field strength after fusion, β i represents the fusion weight, weight Among them, σ i represents the standard deviation of the measurement noise of sensor i; Use Kalman filtering to smooth the data and perform normalization. The data smoothing method is: in, represents the estimated magnetic field strength at time k, represents the estimated magnetic field strength at time k-1, K k Represents the Kalman gain, which indicates the degree of correction of the estimated value by the current measurement. Among them, P k-1 represents the covariance of the prediction error, and R represents the covariance of the measurement noise; standardization: Where B′ represents the standardized magnetic field intensity, μ B represents the average value of magnetic field strength, σ B Represents the standard deviation of the magnetic field strength.

4. The design method of a uniform magnetic field design system based on intelligent perception according to claim 3 is characterized in that: In step S2, the quantum simulation system performs microscopic analysis of the magnetic field as follows: The preprocessed perception data Input the quantum simulation system for initialization, where N is the number of sensors. During initialization, the data is mapped to the quantum state space; Where ψ0 represents the initial quantum state, that is, the initial distribution of the magnetic field, N represents the number of sensors, represents the phase factor, φ i Dependent on the sensor position and magnetic field direction, |B i ,θ i ,T i > represents the data collected by sensor i; The dynamic evolution of the magnetic field is simulated by using the time-dependent Schrödinger equation to simulate the dynamic evolution of the magnetic field. The evolution of the quantum state is controlled by the Hamiltonian H(t) to simulate the time variation and disturbance effects of the magnetic field. The evolution formula is: Among them, ψ(t) represents the quantum state at time t, represents the reduced Planck constant, represents partial derivative; H(t)=H0+H disturb (t), represents the total Hamiltonian, which is composed of the initial system Hamiltonian H0 and the perturbation term H disturb (t) is composed of the disturbance term H disturb (t) represents the dynamic influence of thermal noise and external interference, Where M represents the number of disturbance sources, γ k (t) represents the disturbance intensity, which varies with time, σ k represents the Pauli matrix, which represents the local perturbation form of the magnetic field.

5. The design method of a uniform magnetic field design system based on intelligent perception according to claim 4, characterized in that: The microscopic analysis method of the magnetic field by the quantum simulation system in step S2 also includes: The inhomogeneity, thermal noise and interference sources are identified. The probability density function P(x, y, z) of the magnetic field distribution is calculated based on the local evolution of the observed state to identify the inhomogeneous areas and interference sources. The probability density P(x, y, z) = |<x,y,z|ψ(t)> | 2 , where P(x, y, z) represents the probability of magnetic field distribution at position (x, y, z), and |x, y, z> represents the position state in spatial coordinates; Inhomogeneity identification ΔB(x, y, z) = |B target -∫ V P(x, y, z)B(x, y, z)dV|, where ΔB(x, y, z) represents the deviation from the target magnetic field at the position (x, y, z), and B target represents the target magnetic field strength, ∫ V · Indicates that the average value of the local magnetic field distribution is calculated by integrating the volume V.

6. The design method of a uniform magnetic field design system based on intelligent perception according to claim 5, characterized in that: In step S3, based on the results of quantum simulation analysis, the coil parameters are optimized using deep reinforcement learning as follows: The quantum simulation results are used as the environment state input for reinforcement learning. The initial parameters are set, including the current I0 and the coil spacing d0, and the initial magnetic field uniformity index is calculated. The state space is defined as: s t ={B(x, y, z), ΔB(x, y, z), I, d}, where, s t represents the state vector, i.e. the system state at the current time t, B(x, y, z) represents the magnetic field strength at the position (x, y, z), ΔB(x, y, z) represents the magnetic field inhomogeneity deviation at the position (x, y, z), I represents the current coil current, and d represents the current coil spacing; Using the policy network π(a t |s t )Select action a t , optimize the coil parameters, maximize the reward function R, and the strategy update formula is a t =π θ (s t ), where a t Indicates the action at the current moment, that is, the adjustment value of the current I and the coil spacing d, π θ represents the policy network, with parameter θ; The reward function is R = -α∑ (x，y，z) |ΔB(x,y,z)|-β(I 2 +d 2 ), where R represents the reward value, the goal is to maximize R, α and β represent weight parameters, which control the influence of uniformity and parameter adjustment respectively, and |ΔB(x, y, z)| represents the absolute value of non-uniformity; According to the current strategy π θ , execute the action to adjust the current and coil spacing, and obtain the new state s t+1 and reward R, the state transition formula is s t+1 =f(s t , a t ), where s t+1 represents the new state after executing the action, and f represents the state transition function.

7. The design method of a uniform magnetic field design system based on intelligent perception according to claim 6, characterized in that: In step S3, the method of optimizing coil parameters by using deep reinforcement learning also includes: Using the gradient boosting strategy network parameter θ, the strategy gradient update formula is Among them, θ t+1 represents the updated policy parameter, η represents the learning rate, represents the expected gradient of reward; A fast response model is established. When the environment changes, the coil parameters are adjusted. The fast response model is: in, represents the estimated action of fast response, and g represents the fast response network.

8. The design method of a uniform magnetic field design system based on intelligent perception according to claim 7, characterized in that: In step S4, the key area is the local inhomogeneity and interference source distribution area.

9. The design method of a uniform magnetic field design system based on intelligent perception according to claim 8, characterized in that: In step S4, the current of the compensation coil is dynamically adjusted according to the feedback data to perform local magnetic field compensation in the following manner: The intelligent sensing system feeds back the data to the compensation control module based on the local magnetic field deviation ΔB(x, y, z) collected in real time. The deviation formula is ΔB(x, y, z) = B measured (x, y, z)-B target , where ΔB(x, y, z) represents the magnetic field deviation at position (x, y, z), and B measured (x, y, z) represents the actual measured magnetic field strength, B target Indicates the target magnetic field strength; According to the feedback magnetic field deviation, the current adjustment amount ΔI that needs to be compensated is calculated. The compensation current and the deviation are controlled by proportional integral differential PID. Where ΔI(t) represents the compensation current adjustment at time t, K p , K i , K d are proportional, integral and differential gain coefficients respectively, ΔB(t) represents the magnetic field deviation at the current moment, represents the integral of the magnetic field deviation, that is, the cumulative deviation, It represents the derivative of the magnetic field deviation, i.e. the rate of change; The calculated current adjustment ΔI(t) is applied to the compensation coil, and the adjusted current I(t) is used to compensate for the magnetic field deviation. The current adjustment process is expressed as I(t)=I base +ΔI(t), where I(t) represents the current of the compensation coil at the current moment, I base Indicates the initial setting current of the compensation coil.

10. The design method of a uniform magnetic field design system based on intelligent perception according to claim 9, characterized in that: In step S4, the local magnetic field compensation method further includes: Calculate the adjusted local magnetic field B based on the adjusted current I(t) comp (x, y, z), the compensation effect is fed back to the intelligent sensing system, and the magnetic field compensation formula is: Among them, B comp (x, y, z) represents the magnetic field intensity after compensation, k represents a constant, R represents the radius of the compensation coil, and r represents the distance from the center of the coil to the compensation position; Continuously monitor ΔB(x, y, z) and adjust the compensation current according to the latest deviation. The dynamic feedback method is: ΔB new =ΔB old -B comp (x, y, z), where ΔB new Indicates the new magnetic field deviation after compensation, ΔB old Indicates the magnetic field deviation before compensation.

Images