A method for assessing potential electromagnetic exposure safety uncertainty for pregnant women

Abstract

The application relates to the technical field of computers, and particularly discloses a potential electromagnetic exposure safety uncertainty evaluation method for pregnant women, which comprises the following steps: constructing an uncertainty quantification surrogate model based on a Newton-Raphson improved random forest; constructing a multi-dimensional electromagnetic exposure safety evaluation index system; calculating the probability density function and the exceeding probability of the evaluation index based on the improved random forest uncertainty quantification surrogate model; and quantifying the influence weight of each input variable on the evaluation index in combination with a Sobol method; the application can efficiently calculate the probability density function and the exceeding probability of the evaluation index under the consideration of uncertainty factors by proposing the Newton-Raphson improved random forest uncertainty quantification surrogate model.

Description

Technical Field

[0001] This invention relates to the field of computer technology, specifically a method for assessing the safety uncertainty of potential electromagnetic exposure in pregnant women. Background Technology

[0002] With the widespread application of electromagnetic technology in various electronic devices and energy transmission fields, the impact of the electromagnetic environment on human health has received increasing attention. Among these concerns, the electromagnetic exposure safety assessment of pregnant women and fetuses, as particularly sensitive groups, has become a crucial technical issue that urgently needs to be addressed. If the electromagnetic field strength within a pregnant woman's body exceeds the safety threshold, it may pose a potential threat to the normal development of the fetus. Therefore, establishing a scientific and precise method for assessing electromagnetic exposure safety in pregnant women is of significant practical importance and value for protecting the health of pregnant women and fetuses and for standardizing the design of electromagnetic equipment.

[0003] Currently, existing methods for assessing the safety of electromagnetic exposure in pregnant women mainly revolve around the calculation and analysis of electromagnetic field strength. The core process typically includes the following steps: First, determining the physiological model of the assessment subjects (pregnant woman and fetus), generally using a simplified three-dimensional human model to simulate the anatomical structure of the pregnant woman's abdomen and the fetus; second, selecting assessment indicators, with existing technologies often using the maximum induced field strength within the pregnant woman and fetus as the core assessment indicator, and some methods also considering basic electromagnetic parameters such as electric field strength and magnetic field strength; third, using electromagnetic simulation software (such as FDTD, FEM, etc.) to numerically calculate the electromagnetic field strength within the pregnant woman's body to obtain the calculation results of the core indicator; finally, comparing the calculation results with preset safety thresholds to determine whether the electromagnetic exposure is within a safe range.

[0004] However, in actual scenarios involving electromagnetic exposure assessment of pregnant women, numerous uncertainties exist. These factors can significantly impact the accuracy and reliability of the assessment results, and existing assessment methods exhibit obvious technical deficiencies in handling these uncertainties. As pregnant women and fetuses are particularly sensitive groups, the safety assessment of their electromagnetic exposure has become a critical technical issue that urgently needs to be addressed. If the electromagnetic field strength within a pregnant woman's body exceeds the safety threshold, it may harm the normal development of the fetus. Therefore, establishing a scientific and precise assessment method is of significant practical importance for protecting maternal and infant health and standardizing the design of electromagnetic equipment. Summary of the Invention

[0005] The purpose of this invention is to provide a method for assessing the safety uncertainty of potential electromagnetic exposure in pregnant women, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A method for assessing the safety uncertainty of potential electromagnetic exposure in pregnant women, the method comprising:

[0008] A multi-dimensional electromagnetic exposure safety assessment index system is constructed, which includes: the maximum induced electric field intensity of pregnant women, the maximum induced electric field intensity of fetuses, the stimulation depth of the induced electric field in the pregnant woman's abdomen, and the spatial focusing degree of the induced electric field in the pregnant woman's abdomen.

[0009] An uncertainty quantification proxy model for random forest based on the Newton-Raphson improved model is constructed. The construction process includes: optimizing the hyperparameters of random forest using the Newton-Raphson search rule and trap avoidance operator.

[0010] The probability density function and the probability of exceeding the standard of the evaluation index are calculated based on the random forest uncertainty quantification surrogate model. The influence weight of each input variable on the evaluation index is quantified by combining the Sobol method. The influence weight includes the first-order sensitivity index and the total effect index.

[0011] As a further embodiment of the present invention, the stimulation depth is defined as the longest distance from the beginning of the skin layer along the pregnant woman's abdomen model to the point where the maximum induced electric field intensity in the abdomen decays to 1 / 2.

[0012] As a further embodiment of the present invention, the spatial focus is defined as the area of ​​the stimulation region in the pregnant woman's abdominal region where the induced electric field intensity exceeds half of the maximum electric field intensity, and is calculated using a three-dimensional evaluation formula.

[0013] As a further aspect of the present invention, the step of optimizing the hyperparameters of the random forest using the Newton-Raphson search rule and trap avoidance operator specifically includes:

[0014] The search for a better location in the feasible region is carried out using the Newton-Raphson search rule. The search direction is constructed based on the positional relationship between sample points and the Taylor expansion.

[0015] The trap avoidance operator prevents the algorithm from getting trapped in local optima, and generates a solution with enhanced quality by combining the optimal position and the current vector position.

[0016] As a further embodiment of the present invention, the step of calculating the probability density function and the probability of exceeding the standard based on the improved random forest uncertainty quantification surrogate model specifically includes: inputting the sampled input variable samples into the trained improved random forest uncertainty quantification surrogate model to obtain the corresponding output response;

[0017] Statistically analyze the probability density function and the probability of exceeding the standard for each evaluation indicator.

[0018] As a further aspect of the present invention, the step of quantifying the influence weights of each input variable on the evaluation index using the Sobol method specifically includes:

[0019] Based on variance decomposition, the independent contribution of a single variable to the output variance is calculated as the first-order sensitivity index. The sum of the first-order sensitivity index of a variable and the sensitivity index of the interaction between the variable and other variables is calculated as the total effect index.

[0020] Compared with existing technologies, the beneficial effects of this invention are: it innovatively applies stimulation depth and focus indexes to the assessment of the induction field in the pregnant woman's abdomen, and improves the assessment dimensions of field strength distribution: the "intensity-depth-range" of electric field distribution are three core dimensions, and focus complements the maximum field strength and stimulation depth, comprehensively characterizing the electric field distribution features from three levels: "how much intensity", "how deep the penetration", and "how large the concentration range", making the assessment more scientific.

[0021] A Newton-Raphson Improved Random Forest (NRBO-RF) uncertainty quantification proxy model is proposed, which can efficiently calculate the probability density function and overshoot probability of the evaluation index under the consideration of uncertainty factors, and combine the Sobol method to quantify the influence weight of input variables. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention.

[0023] Figure 1 A block diagram of a safety assessment system for electromagnetic exposure of pregnant women provided in an embodiment of the present invention.

[0024] Figure 2 A flowchart of a proxy model for quantifying the uncertainty of electromagnetic exposure in pregnant women, provided in an embodiment of the present invention.

[0025] Figure 3 (a) is a schematic diagram of a pregnant woman model under electromagnetic exposure; (b) is a cross-sectional view of the electric field strength on the abdomen of the pregnant woman, where the black line is the test line passing through the point of maximum field strength; (c) is the distribution of electric field strength on the black test line; (d) is the overall distribution of the induced electric field of the abdominal model; (e) is the numerical distribution of the field strength value and the cumulative volume of the abdominal model.

[0026] Figure 4 A flowchart of a proxy model for assessing electromagnetic exposure in pregnant women based on NRBO-RF, provided for embodiments of the present invention.

[0027] Figure 5 (a) is a vector distribution diagram of the induced magnetic field in the solution domain of the case of wireless charging for electric vehicles; (b) is a heat map of the induced electric field intensity distribution in the pregnant woman's body in a standing posture; (c) is a heat map of the induced electric field intensity distribution in the fetus in the abdomen.

[0028] Figure 6 Probability density distribution function curves of four indicators of electromagnetic radiation exposure; (a) Figure shows E pre (a) The probability density distribution curve of the index; (b) The figure shows E baby (c) The probability density distribution curve of the indicator; Figure d is shown. 1 / 2 The probability density distribution curve of the index; (d) Figure S 1 / 2 The probability density distribution curve of the indicator.

[0029] Figure 7 The Sobol method is used to calculate the first-order sensitivity index S and the total sensitivity index S for four electromagnetic radiation exposure parameters. T Results are shown in the figure; (a) Figure E pre The Sobol method is used to calculate the first-order sensitivity index S and the total sensitivity index S. T Results are shown in the figure; (b) Figure E baby The Sobol method is used to calculate the first-order sensitivity index S and the total sensitivity index S. T Results are shown in Figure d; (c) Figure d is a result of the results. 1 / 2 The Sobol method is used to calculate the first-order sensitivity index S and the total sensitivity index S. T Results are shown in the figure; (d) Figure S 1 / 2 The Sobol method is used to calculate the first-order sensitivity index S and the total sensitivity index S. T Results are shown in the image. Detailed Implementation

[0030] To make the technical problems to be solved, the technical solutions, and the beneficial effects of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present invention and are not intended to limit the present invention.

[0031] In this embodiment of the invention, a method for assessing the safety of potential electromagnetic exposure in pregnant women includes:

[0032] An uncertainty quantification proxy model for random forest based on the Newton-Raphson improved model is constructed. The construction process includes: optimizing the hyperparameters of random forest using the Newton-Raphson search rule and trap avoidance operator.

[0033] A multi-dimensional electromagnetic exposure safety assessment index system is constructed, which includes: the maximum induced electric field intensity of pregnant women, the maximum induced electric field intensity of fetuses, the stimulation depth of the induced electric field in the pregnant woman's abdomen, and the spatial focusing degree of the induced electric field in the pregnant woman's abdomen.

[0034] The probability density function and the probability of exceeding the standard of the evaluation index are calculated based on the random forest uncertainty quantification surrogate model. The influence weight of each input variable on the evaluation index is quantified by combining the Sobol method. The influence weight includes the first-order sensitivity index and the total effect index.

[0035] like Figures 1 to 7 As shown, in this embodiment, a commercially available high-precision anatomical voxel model of a pregnant woman and a nine-month-old fetus is used to focus on exploring the uncertainty distribution of the amplitude, stimulation depth, and focusing of the induced field strength in the body of the pregnant woman and the fetus under electromagnetic field exposure, in order to achieve a comprehensive, accurate, and effective assessment of the safety of electromagnetic exposure for pregnant women.

[0036] A Newton-Raphson-Random Forest (NRBO-RF) proxy model for quantifying the uncertainty of electromagnetic exposure was constructed. This model can be used to calculate the probability density function and the probability of exceeding the standard in the uncertainty quantification of electromagnetic exposure risk in pregnant women, as well as the sensitivity analysis of the Sobol parameter, thereby achieving accurate assessment of the induced electric field in the body of pregnant women.

[0037] It is necessary to introduce uncertainty quantification methods to quantify the influence of relevant parameters on the human body's electromagnetic field exposure dose and thus predict the effect of this uncertainty.

[0038] Regarding the maximum amplitude of the induced electric field, the International Commission on Non-Ionizing Radiation Protection (ICNIRP) clearly stipulated in its 2010 guidelines that, to avoid nerve stimulation to the human body, the induced electric field strength should be used as the limit assessment indicator (when the frequency of the interference source is below 100 kHz). In daily life, numerous low-frequency and ultra-low-frequency electromagnetic interference sources exist in environments such as inside cars and around household appliances, posing potential exposure risks to the human body. Therefore, this study analyzed the maximum induced electric field E of the fetus (Charlie) based on a virtual anatomical human model. baby The maximum induced electric field E of the pregnant woman (Ella) pre Spatial distribution characteristics.

[0039] As a preferred embodiment of the present invention, the stimulation depth is defined as the longest distance from the beginning of the skin layer of the pregnant woman's abdomen model to the point where the maximum induced electric field intensity in the abdomen decays to 1 / 2.

[0040] In this embodiment, the stimulation depth represents the penetrating power of a transient electromagnetic field induced by an external radiation source within human tissue. This study defines the stimulation depth as follows: Figure 3 As shown in (b) and (c), the black test line starts from the skin layer of the abdominal model and extends to the maximum induced electric field intensity (E) on the abdomen. maxThe longest distance at which the decay reaches half its original length is denoted as d. 1 / 2 .

[0041] In a preferred embodiment of the present invention, the spatial focusing degree is defined as the area of ​​the stimulation region in the pregnant woman's abdominal region where the induced electric field intensity exceeds half of the maximum electric field intensity, and is calculated using a three-dimensional evaluation formula.

[0042] In this embodiment, the focusing power is used to characterize the focus of an external radiation source on the pregnant woman's abdominal region. Figure 3 The spatial focusing characteristics of the induced electric field generated in the human tissue region (defined by the black line in (a)) are quantified by the area of ​​the field strength focused in the body. This indicator is defined as the area of ​​the stimulation region in the human tissue where the intensity of the induced electric field exceeds half of the maximum electric field intensity. It directly reflects the ability of the radiation source to spatially concentrate the induced electric field in the human tissue.

[0043] Due to the irregularity of the induced electric field distribution and the three-dimensional structural characteristics of abdominal tissue, the focal area is difficult to measure directly. Therefore, the following three-dimensional evaluation formula is used to approximate the calculation of the focal area:

[0044] ;

[0045] in, Focus, For pregnant women, the abdominal area is larger than E max Half of the cumulative volume, such as Figure 3 As shown in (e).

[0046] This study explores the impact of the uncertainty of the dynamic spatial position of pregnant women and radiation sources in everyday electromagnetic device usage scenarios on the assessment results of electromagnetic safety exposure of pregnant women. To construct a high-precision, high-dimensional data regression prediction model and quantify the uncertainty of electromagnetic safety risks for pregnant women, this invention proposes a random forest regression prediction model based on an improved Newton-Raphson optimizer, aiming to adapt to multi-dimensional input variables and improve the computational efficiency of the surrogate model. This model can comprehensively characterize the induced electromagnetic field characteristics within human tissue from three dimensions: stimulus field strength, depth of action, and field clustering degree. Simultaneously, combined with Sobol global sensitivity analysis, it achieves accurate prediction of multi-dimensional parameters of the induced field within the body under the uncertainty of electromagnetic exposure safety and quantitative analysis of the coupling influence between external input parameters.

[0047] As a preferred embodiment of the present invention, the step of optimizing the hyperparameters of the random forest using the Newton-Raphson search rule and trap avoidance operator specifically includes:

[0048] The search for a better location in the feasible region is carried out using the Newton-Raphson search rule. The search direction is constructed based on the positional relationship between sample points and the Taylor expansion.

[0049] The trap avoidance operator prevents the algorithm from getting trapped in local optima, and generates a solution with enhanced quality by combining the optimal position and the current vector position.

[0050] In this embodiment, Random Forest (RF), as a machine learning algorithm, adopts the ensemble learning idea of ​​"The wisdom of them all outweighs the wisdom of the few." By fusing multiple unrelated decision trees (CART), it achieves an ensemble classifier with better prediction performance. The ensemble classifier then votes on all regression prediction results and designates the category with the most votes as the final output.

[0051] The Random Forest algorithm can detect interactions between features and calculate their importance scores during training. The Gini index (GI) can be used as an evaluation metric for assessing feature importance in Random Forest.

[0052] ;

[0053] Where K represents the total number of categories, Let X be the proportion of samples of class K in node M. Two samples with different class labels are randomly selected from node M. The variable importance score is represented by VIM, and feature X... j The Gini index score is used Indicates that feature X j The importance of node M (i.e., the change in Gini index before and after node M splits) can be expressed as:

[0054] ;

[0055] In the formula GI l and GI r These are the Gini indices of the two new nodes after the split. If feature X... j If the set of nodes appearing in the i-th decision tree is M, then feature X j The importance in the i-th decision tree can be represented as:

[0056] ;

[0057] Assuming there are k trees in a random forest, the importance of feature j can be expressed as:

[0058] ;

[0059] However, the selection of hyperparameters in RF often relies on traditional methods such as grid search and random search. These methods generally have significant drawbacks such as low optimization efficiency and easy getting trapped in local optima. The number of decision trees is the primary core hyperparameter that determines the performance of the RF model: it directly affects the fitting and generalization ability, and its impact on the model is global, smooth, and unidirectional.

[0060] Therefore, this paper adopts the Newton-Raphson Optimization (NRBO) algorithm to effectively overcome the above problems. The NRBO algorithm belongs to a new type of metaheuristic optimization algorithm, which mainly consists of the Newton-Raphson Search Rule (NRSR) and the Trap Avoidance Operator (TAO). Overall, the NRBO algorithm, with its good balance between the ability to explore and develop the solution space, possesses the core advantages of fast convergence speed and strong local optimum avoidance ability, providing an efficient solution path for complex optimization problems. Its process is as follows: Figure 4 As shown.

[0061] The Newton-Raphson Search (NRSR) mechanism is the core of the NRBO algorithm. It aims to find a better location within the feasible region using the Raphson search method. This mechanism does not depend on the specific expression of the objective function; instead, it constructs the search direction through the positional relationships between sample points and the Taylor expansion. The main expression is:

[0062] ;

[0063] Among them, X W X represents the worst individual in the current population. b This represents the best individual in the current population. The step size vector is determined by the difference between the current individual and the optimal solution, x. n The search position is determined by introducing a dynamic adjustment factor; r2 is a randomly selected integer from the population, δ is a parameter representing the interaction between individuals, and the population orientation is determined, allowing the algorithm to construct a new search position in each iteration. and Generate new search locations :

[0064] ;

[0065] .

[0066] The Trap Avoidance Operator (TAO) aims to improve the effectiveness of NRBO in handling practical problems and avoid local optima. TAO employs an improved and enhanced computational formula proposed by Ahmadianfar et al. Using TAO can significantly change... The position, by combining the optimal position x b and current vector position To produce with enhanced quality The solution.

[0067]

[0068] ;

[0069] in, , The result is a uniformly random number between (−1, 1) and (−0.5, 0.5); and It is a random number.

[0070] As a preferred embodiment of the present invention, the step of calculating the probability density function and the probability of exceeding the standard based on the random forest uncertainty quantification surrogate model specifically includes: inputting the sampled input variable samples into the trained random forest uncertainty quantification surrogate model to obtain the corresponding output response;

[0071] Statistically analyze the probability density function and the probability of exceeding the standard for each evaluation indicator.

[0072] As a preferred embodiment of the present invention, the step of quantifying the influence weights of each input variable on the evaluation index using the Sobol method specifically includes:

[0073] Based on variance decomposition, the independent contribution of a single variable to the output variance is calculated as the first-order sensitivity index. The sum of the first-order sensitivity index of a variable and the sensitivity index of the interaction between the variable and other variables is calculated as the total effect index.

[0074] In this embodiment, the Sobol method is a widely applied global sensitivity analysis method based on variance decomposition. It calculates sensitivity indices by quantifying the contribution of single variables or the interactions between multiple variables to the output variance. Considering the numerous variables involved in the uncertainty assessment of electromagnetic safety for pregnant women, quantifying the impact of different input variables on the electromagnetic safety of pregnant women and identifying key parameters that significantly affect the output results can provide important theoretical basis for subsequent electromagnetic safety protection and human electromagnetic safety assessment standards. Therefore, this paper, based on the NRBO-RF surrogate model, combines Sobol to solve for the global sensitivity indices of different variables in the actual scenario of pregnant women.

[0075] Assume the output of the uncertainty quantification model for electromagnetic safety of pregnant women is Y=f(X), where X=(X1,X2,...,X...). n Let Y be the independent input parameter of the model. The total variance D = Var(Y) is decomposed as follows:

[0076] ;

[0077] in, , And so on. The influence of different input variables and interactions between variables on the output variance is derived. The Sobol sensitivity index is defined as:

[0078] ;

[0079] Among them, S i The first-order sensitivity index (main effect) represents the independent contribution of a single variable to the output variance. The sum of the first-order sensitivity index of a variable and the sensitivity indices of its interactions with other variables is defined as the total effect index.

[0080] ;

[0081] In summary, this paper constructs an NRBO-RF-based electromagnetic exposure safety algorithm for pregnant women and combines it with an uncertainty quantification analysis model based on Sobol sensitivity analysis. The specific process is as follows: First, the Monte Carlo (MC) method is used to obtain N... MC Sample Next, the proxy model is invoked to calculate the corresponding output response. Based on these samples, the statistical characteristics of the induced electric field parameters of pregnant women exposed to electromagnetic radiation can be directly solved and analyzed, including the statistical probability of exceeding the ICNIRP standard limits and the Sobol sensitivity analysis of each input parameter to the output parameter. Finally, the accuracy of the calculation results can be verified by comparing with the Monte Carlo method.

[0082] The pseudocode for the NRBO-RF uncertainty quantification model is as follows.

[0083] Input: Monte Carlo (MC) sampling input matrix ;

[0084] Output: The output response of the training samples obtained through finite element simulation software. , thus obtaining training sample D.

[0085] Data initialization:

[0086] The training samples D are obtained using the Bootstrap sampling method;

[0087] Construct a decision tree;

[0088] Construct an NRBO optimizer to optimize the number of decision trees for RF hyperparameters;

[0089] while (IT < );

[0090] The Raphson search method searches for a better location within the feasible area.

[0091] Determine the solution ;

[0092] / / Trap Avoidance Operator (TAO);

[0093] if rand < the deciding factor;

[0094] Find the solution ;

[0095] else determine the solution ;

[0096] end while;

[0097] The final prediction result is determined by combining evaluation indicators and voting.

[0098] Calculate using trained NRBO-RF Corresponding output response ;

[0099] calculate probability distribution function And calculate the statistical probability of exceeding the standard limit;

[0100] Calculate the Sobol first-order sensitivity index S and the total effect index S between variables. T .

[0101] In this embodiment, to verify the superiority of the proposed NRBO-RF uncertainty quantification alternative model, we compared it with mainstream models, including Monte Carlo (MC), Physical Information Neural Network (PINN), Transformer, Radial Basis Function (RBF), Gated Recurrent Unit (GRU), and Random Forest (RF). Key metrics included the coefficient of determination (R²) of the training set, mean squared error (MSE), mean absolute error (MAE), and computational time cost. To ensure fairness and reliability of the results, the sample size for all models was uniformly set to 2000. As shown in Table 1, NRBO-RF outperformed all other models in all performance metrics (except for time cost). In terms of computation time, the MC method is extremely time-consuming in electromagnetic exposure quantification; in contrast, the time cost of NRBO-RF is only 1.87% of that of the MC method. Therefore, NRBO-RF combines high computational efficiency, low cost, and excellent fitting accuracy, fully demonstrating its efficiency and practicality in electromagnetic exposure assessment.

[0102] Table 1 Comparison of performance metrics of proxy models

[0103]

[0104] This invention applies to the assessment of electromagnetic exposure in pregnant women using electric vehicle wireless power transfer (EV-WPT). A simulation model of electromagnetic exposure in a standing position for pregnant women was built using the electromagnetic numerical simulation software Sim4life, including two sub-models: a 9-month-old fetus (Charlie) and a pregnant woman (Ella). The model consists of transmitting and receiving coils, a ferrite layer, and an aluminum shielding layer. The coils have a double-sided structure with 10 turns, a turn spacing of 8 mm, a coil cross-sectional radius of 3.5 mm, and a single side length of approximately 455 mm and a width of 685 mm. The radiation source power is set to 22 kW, and the current source for the transmitting coil is set to 50 A. The ferrite layer is composed of high-permeability ferrite magnetic blocks. Considering the fragility of large-area ferrite materials, it is designed as a small unit of 83 × 83 × 4 mm (length × width × thickness). The ferrite layer is divided into two layers: the lower layer, closer to the coil, contains 32 small units, and the upper layer contains 16 small units. The aluminum shielding layer is identical in size to the coil, with a thickness of 2 cm. The electric vehicle model is set to 4.5 × 2 × 1.5 m in size, and the main material is aluminum. To ensure the accuracy of the simulation results, the materials of components such as windows are set to insulating materials such as glass.

[0105] With the vehicle body, the pregnant woman model, and the coil group in fixed spatial positions (i.e., the fixed working condition under normal operation), the distribution of the induced magnetic field vector in the solution domain is shown in Figure 5(a). The shielding effect of the vehicle chassis and body causes the induced electric field distribution to exhibit non-uniform characteristics, as shown in Table 2, which illustrates the distribution of the pregnant woman's standing body E.pre The electric field strength E in the fetus is 2.08 V / m. Although the soft tissues of the pregnant woman's body, the uterus, and amniotic fluid provide effective electromagnetic shielding for the fetus, the induced electric field strength E in the fetus is relatively low. baby Compared to E pre Significant attenuation was observed, but the stimulation depth index exceeded the 25mm distance between the pregnant woman and the fetus, indicating that the fetus was affected by abdominal E in all three scenarios. max / 2 field strength influence, there is a risk of electromagnetic exposure.

[0106] Table 2. Numerical distribution of four exposure indicators for pregnant women in three scenarios.

[0107]

[0108] This invention employs an uncertainty quantification method based on NRBO-RF to assess the electromagnetic exposure safety of pregnant women and their fetuses. First, characteristic parameters of different input variables are collected and analyzed. For the standing exposure scenario of pregnant women, uncertainty input variables are designed, as shown in Table 3. Here, U represents a uniform distribution of input parameters, including upper and lower bounds; N represents a normal distribution, including mean and standard deviation. The dimensions of the input variables are dynamically adjusted according to the exposure scenario, resulting in 11 dimensions. The study uses the maximum induced electric field intensity (E) within the pregnant woman's body as the starting point. pre ), the maximum induced electric field intensity of the fetus in the abdomen (E) baby ), depth of stimulation (d) 1 / 2 ) and focus (S 1 / 2 The core output indicator is used to analyze the range of numerical fluctuations under the influence of uncertain factors.

[0109] Table 3 Uncertainty Parameters

[0110]

[0111] This study compares the NRBO-RF model with the MC model, using the probability density function (PDF) of the four output indicators under different exposure scenarios to quantitatively evaluate the predictive performance of each model. Figure 6 The distribution trend of the probability density function shows that the NRBO-RF model and the MC method have a high degree of agreement in terms of calculation and fitting accuracy. The two have good consistency. This result fully verifies that the improved NRBO-RF method has achieved a substantial improvement in data fitting performance and effectively solves the problem of insufficient accuracy of the traditional RF model in the quantification of electromagnetic exposure uncertainty.

[0112] Regarding the field strength amplitude index, this study compares and analyzes it with the limits in the ICNIRP 2010 electromagnetic exposure protection guidelines: Figure 6In (a), the probability density area exceeding the limit is highlighted in pink. The results show that, in the uncertain scenario, the probability of a pregnant woman's electromagnetic radiation exposure dose exceeding the safety limit in the standing position scenario reaches 19.93%. Figure 6 (b) further demonstrates E baby The study found that the uncertainty probability density distribution characteristics of the indicator, in E pre Under the premise that the indicator exceeds the standard, E baby The distribution pattern and E pre The results were highly consistent, all concentrated in the high-value range. For pregnant women in the rear area of ​​the vehicle, although the coil's own shielding structure and the vehicle chassis can provide some electromagnetic shielding, this study comprehensively incorporated the multi-source uncertainties present in the actual radiation scenarios of pregnant women, and can more accurately reveal the potential risk of excessive electromagnetic exposure doses for pregnant women. Figure 6 (c) Further demonstrates the depth of stimulation of the in vivo sensory field (d) 1 / 2 The uncertainty probability density distribution law of the stimulus is investigated. In the numerical model of pregnant women constructed in this study, the average anatomical depth from the pregnant woman's abdominal skin to the fetal skin is 25 mm; when the stimulus depth d > 25 mm, it means that the intensity is E. max A induced electric field of 2 / 2 (the maximum induced electric field in the pregnant woman's abdomen) can penetrate the fetal skin barrier and directly act on fetal tissues. The probability of this risk event occurring in any exposure scenario is 100%, indicating that the risk of fetal tissue being stimulated by the electric field is almost complete. Furthermore, regarding the reference value for electric field exposure (E... pre ) under conditions exceeding the standard d 1 / 2 Analysis of the probability distribution characteristics (highlighted in pink) shows that the stimulation depth distribution of the in vivo induced electric field does not exhibit a high-value clustering characteristic, but rather shows significant dispersion. Correspondingly, Figure 6 (d) The uncertainty probability distribution of the focusing intensity also exhibits a dispersed pattern. This phenomenon reveals that relying solely on the electric field strength as a single indicator cannot accurately characterize the actual distribution law of the field strength within the body, thus verifying the introduction of d in this study. 1 / 2 With S 1 / 2 The necessity of the two supplementary indicators. The inclusion of the above indicators makes up for the inherent deficiency of traditional assessment methods that only measure the risk of fetal electromagnetic exposure based on electric field strength, and makes the assessment system more in line with the physiological effects in actual exposure scenarios.

[0113] On the other hand, in order to quantify the influence of input variables on the four indicators, the Sobol method is used to calculate the first-order sensitivity index S of each variable and the overall sensitivity index S. T 2000 sampled values ​​were obtained through Sobol Sequence sampling and then input into the NRBO-RF surrogate model. The results are as follows. Figure 7As shown, the spatial displacement of pregnant women has the greatest impact on the assessment results, especially their displacement along the Y-axis. The sensitivity analysis results not only provide a theoretical basis for the scientific delineation of the electromagnetic safety activity area for pregnant women, helping them effectively avoid the risk of high-field-strength exposure; but also provide key data support for the optimization and improvement of electromagnetic exposure protection standards. This overcomes the limitations of traditional standards that focus on fixed-condition assessments, and promotes relevant standards towards a more refined assessment that incorporates dynamic spatial changes, thereby building a comprehensive electromagnetic safety protection system for pregnant women and other sensitive groups.

[0114] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for assessing the safety uncertainty of potential electromagnetic exposure in pregnant women, characterized in that, The method includes: An uncertainty quantification proxy model for random forest based on the Newton-Raphson improved model is constructed. The construction process includes: optimizing the hyperparameters of random forest using the Newton-Raphson search rule and trap avoidance operator. A multi-dimensional electromagnetic exposure safety assessment index system is constructed, which includes: the maximum induced electric field intensity of pregnant women, the maximum induced electric field intensity of fetuses, the stimulation depth of the induced electric field in the pregnant woman's abdomen, and the spatial focusing degree of the induced electric field in the pregnant woman's abdomen. The probability density function and the probability of exceeding the standard of the evaluation index are calculated based on the improved random forest uncertainty surrogate model. The influence weight of each input variable on the evaluation index is quantified by combining the Sobol method. The influence weight includes the first-order sensitivity index and the total effect index.

2. The method for assessing the safety uncertainty of potential electromagnetic exposure in pregnant women according to claim 1, characterized in that, The stimulation depth is defined as the longest distance from the start of the skin layer along the pregnant woman's abdomen to the point where the maximum induced electric field intensity in the abdomen decays to 1 / 2.

3. The method for assessing the safety uncertainty of potential electromagnetic exposure in pregnant women according to claim 1 or 2, characterized in that, The spatial focus is defined as the area of ​​the stimulation region within the pregnant woman's abdominal region where the induced electric field intensity exceeds half of the maximum electric field intensity, calculated using a three-dimensional evaluation formula.

4. The method for assessing the safety uncertainty of potential electromagnetic exposure in pregnant women according to claim 1, characterized in that, The steps for optimizing the hyperparameters of the random forest using the Newton-Raphson search rule and trap avoidance operator specifically include: The search for a better location in the feasible region is carried out using the Newton-Raphson search rule. The search direction is constructed based on the positional relationship between sample points and the Taylor expansion. The trap avoidance operator prevents the algorithm from getting trapped in local optima, and generates a solution with enhanced quality by combining the optimal position and the current vector position.

5. The method for assessing the safety uncertainty of potential electromagnetic exposure in pregnant women according to claim 1, characterized in that, The steps for calculating the probability density function and over-limit probability of the evaluation index based on the improved random forest uncertainty quantification surrogate model specifically include: inputting the sampled input variables into the trained improved random forest uncertainty quantification surrogate model to obtain the corresponding output response; Statistically analyze the probability density function and the probability of exceeding the standard for each evaluation indicator.

6. The method for assessing the safety uncertainty of potential electromagnetic exposure in pregnant women according to claim 1, characterized in that, The step of quantifying the influence weights of each input variable on the evaluation index using the Sobol method specifically includes: Based on variance decomposition, the independent contribution of a single variable to the output variance is calculated as the first-order sensitivity index. The sum of the first-order sensitivity index of a variable and the sensitivity index of the interaction between the variable and other variables is calculated as the total effect index.

Images