Low-energy-consumption non-contact targeted oxygen supply system

Abstract

The invention relates to the technical field of building environments, in particular to a low-energy-consumption non-contact targeted oxygen supply system which comprises an oxygen supply device body, a static pressure box, an oxygen supply hose, a fan, a personalized oxygen supply terminal and an oxygen detection device used for sampling oxygen concentration in a sampling area. The personalized oxygen supply terminal comprises a plurality of air outlet pipes, and the air outlet pipes have adjustable oxygen supply parameters; the system further comprises a joint optimization module which is used for carrying out iterative optimization on the oxygen supply concentration and the oxygen supply parameters according to a Kriging-multi-island genetic algorithm to obtain an optimal oxygen supply parameter combination so as to realize optimal oxygen supply. According to the method, through combined optimization of side type multi-angle jet flow, local encryption grid CFD analysis and the Kriging-multi-island genetic algorithm, the problems that in the prior art, oxygen conveying energy consumption is large, the oxygen supply effect is poor, and testing means are lacked are solved, accurate oxygen supply only for the breathing area is achieved, energy consumption is remarkably reduced, and comfort and safety are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of building environment, and particularly relates to a low-energy-consumption non-contact targeted oxygen supply system. BACKGROUND

[0002] At present, the oxygen supply technology of the plateau sleep environment mainly adopts room whole diffusion oxygen supply or mask / nasal catheter oxygen inhalation. The diffusion oxygen supply needs to continuously transport high-concentration oxygen to the whole room, and the system has large volume and high energy consumption. A large amount of oxygen is lost through the door and window gaps, ventilation system and other non-breathing areas, and the oxygen utilization rate is less than 30%. In addition, the diffusion oxygen supply is prone to cause oxygen enrichment in the lower part of the room, and it is difficult to take into account the local hypoxia caused by the difference in personnel activity range. Although the mask and nasal catheter method can realize individual oxygen supply, it is not comfortable to wear, and it is easy to fall off at night. The user compliance is insufficient, which affects sleep, and there is a risk of oxygen leakage and fire. The above methods do not optimize the flow field characteristics of the breathing area under the supine sleeping position, and the oxygen supply position and angle are difficult to accurately match, resulting in the misalignment of the oxygen-enriched area and the breathing area, and the serious waste of oxygen.

[0003] In addition, the existing equipment lacks real-time optimization capability based on environmental differences. The large diurnal temperature difference and low air humidity in plateau areas cause significant fluctuations in oxygen diffusion rate and human metabolic demand, and the breathing characteristics of different users (such as children and the elderly) differ significantly. The existing equipment cannot adaptively adjust the oxygen supply parameters according to altitude, individual differences and environmental changes, and it is difficult to achieve a comfortable experience of "oxygen without wind". SUMMARY

[0004] In view of the deficiencies of the prior art, the present application aims to provide a low-energy-consumption non-contact targeted oxygen supply system. Through side-type multi-angle jet flow, local encryption grid CFD analysis and Kriging-multiple island genetic algorithm joint optimization, the problems of large oxygen delivery energy consumption, poor oxygen supply effect and lack of test means in the prior art are solved, and only the breathing area is accurately supplied with oxygen, which significantly reduces energy consumption and improves comfort and safety.

[0005] In order to achieve the above-mentioned purpose, the present application provides the following technical scheme: A low-energy-consumption non-contact targeted oxygen supply system, comprising an oxygen supply device body, a static pressure tank, an oxygen supply hose, a fan, a personalized oxygen supply terminal and an oxygen detection device for sampling the oxygen concentration of a sampling area; The personalized oxygen supply terminal comprises a plurality of air outlet pipes, and the air outlet pipes have adjustable oxygen supply parameters. Further comprising a joint optimization module for iteratively optimizing the oxygen supply concentration and the oxygen supply parameters according to the Kriging-multiple island genetic algorithm to obtain the optimal oxygen supply parameter combination and achieve the best oxygen supply.

[0006] Preferably, the system further comprises a three-dimensional modeling process, comprising: S1, establishing a non-structural grid of the room; S2, taking a plurality of air outlet pipes and a human head as an oxygen supply area, locally encrypting the oxygen supply area, and determining the number of grids; S3, setting boundary conditions; S4, selecting a turbulence model; S5, obtaining a three-dimensional simulation model through experimental verification; Wherein, the joint optimization module iteratively optimizes the simulation calculation results of the three-dimensional simulation model.

[0007] Preferably, the oxygen supply parameters include inner diameter D, downward angle β and inner angle θ; the Kriging-multiple island genetic algorithm includes an optimization target, and the optimization target is: MinC target Further comprising a constraint condition: C avg (D, θ, β, C) ≥ 24%vol; wherein, C is the oxygen concentration value; C target is the concentration target value; C avg is the average concentration value.

[0008] Further, the optimization process of the Kriging-multiple island genetic algorithm is as follows: S1: offline sampling stage: using optimal Latin hypercube sampling method, generating sample set within the variable range of preset oxygen concentration C and oxygen supply parameters D, β and θ, and obtaining corresponding concentration target value C target and average oxygen concentration value C avg through three-dimensional simulation model calculation, to obtain a sample library; S2: establishment of proxy model: training Kriging model based on the sample library, using Gaussian kernel function and maximum likelihood estimation, and the model determination coefficient R²≥0.95; S3: online optimization stage: using multiple island genetic algorithm to optimize search on the trained Kriging proxy model to quickly generate the optimal oxygen supply parameter combination.

[0009] Preferably, in S1, the concentration target value C target is calculated according to the following formula: C target =Σ(C i -C avg )² / n Wherein, n is the total number of measuring points; C i is the oxygen concentration of the i-th measuring point, unit %vol; C avg is the average value of oxygen concentration of all measuring points, used to evaluate the overall oxygen supply level of the breathing area, unit %vol; C targetThe concentration target value is used to measure the uniformity or concentration degree of oxygen concentration in the breathing area.

[0010] Preferably, in S1, the variable range of the oxygen supply concentration C is: C∈[30%, 50%]. The variable range of the oxygen supply parameters D, β and θ is: D∈[1, 2] cm, β∈[10°, 30°], and θ∈[5°, 15°].

[0011] Preferably, the sampling area is centered on the facial breathing area and arranged in a three-dimensional grid, and the nodes of the three-dimensional grid are the set measurement points, and the distance between every two measurement points ranges from 1cm×1cm×1cm to 8cm×8cm×8cm.

[0012] Preferably, the oxygen detection device comprises an oxygen detector and a sampling probe located above the facial breathing area of the human body, and the sampling probe is used to collect the oxygen concentration at the measurement points.

[0013] Preferably, the personalized oxygen supply terminal further comprises a multi-degree-of-freedom positioning mechanism for adjusting the oxygen supply parameters of each air outlet pipe.

[0014] Preferably, in S3, the multi-island genetic algorithm sets the number of islands to 10 and the population size of each island to 10 individuals; the evolution is 1000 generations, the crossover probability is preferably set to 0.8-1.0, the mutation probability is preferably set to 0.005-0.02, and the migration interval is 5 generations.

[0015] Compared with the prior art, the present application has the following beneficial effects: (1) The present application obtains Kriging surrogate model training samples by local encryption grid CFD, and combines multi-island genetic algorithm for joint optimization to realize online optimization, and automatically adjusts the inner diameter, angle and oxygen concentration of the air outlet pipe through the multi-degree-of-freedom positioning mechanism, thereby avoiding repeated manual debugging in slow changing scenarios such as sleep, and ensuring the consistency of oxygen supply effect and energy consumption advantage under different altitudes and individual differences.

[0016] (2) The system of the present application realizes the transformation from "whole room oxygen supply" to "targeted oxygen supply" by arranging a side type multi-angle jet flow at the oxygen supply terminal and directing the oxygen-rich airflow to the facial breathing area of the person lying on his back, which significantly improves the utilization rate of oxygen in the target breathing area, and increases the oxygen utilization rate by more than 60% under the premise of ensuring comfort, greatly reducing the energy consumption of oxygen production and transportation process.

[0017] (3) The present application realizes fast, low computational cost approximate evaluation and large-scale parallel search of the oxygen supply parameter space by using Kriging surrogate model and multi-island genetic algorithm in combination, which significantly improves the accuracy of high optimization results. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 is a schematic diagram of a system arrangement of a low-energy non-contact targeted oxygen supply system of the present application; Figure 2 is a schematic diagram of a room structure of a low-energy non-contact targeted oxygen supply system of the present application; Figure 3 is a schematic diagram of a respiratory zone targeted area positioning of a low-energy non-contact targeted oxygen supply system of the present application; Figure 4 is a schematic diagram of a respiratory zone measurement point array arrangement of a low-energy non-contact targeted oxygen supply system of the present application; Figure 5 is a simulation and experimental oxygen concentration comparison verification diagram of a low-energy non-contact targeted oxygen supply system of the present application; Figure 6 is a schematic diagram of a key angle definition of an air outlet of an air outlet pipe of a low-energy non-contact targeted oxygen supply system of the present application; Figure 7 is a schematic diagram of a non-structured grid and grid encryption of a low-energy non-contact targeted oxygen supply system of the present application; Figure 8 is a grid independence verification diagram of a low-energy non-contact targeted oxygen supply system of the present application; Figure 9 is a diagram of the relationship between the average wind speed and the outflow velocity in the respiratory zone of a low-energy non-contact targeted oxygen supply system of the present application; Figure 10 is a diagram of the relationship between the average oxygen concentration and the oxygen supply concentration in the respiratory zone of a low-energy non-contact targeted oxygen supply system of the present application Figure 11 is a turbulence model applicability verification diagram of a low-energy non-contact targeted oxygen supply system of the present application; Figure 12 is a Kriging model goodness-of-fit diagram of a low-energy non-contact targeted oxygen supply system of the present application; Figure 13 is a multi-island genetic algorithm iteration process diagram of a low-energy non-contact targeted oxygen supply system of the present application; Figure 14 is a schematic diagram of an optimization process of a low-energy non-contact targeted oxygen supply system of the present application.

[0019] In the figure: 1, oxygen supply device body; 2, static pressure tank; 3, oxygen supply hose; 4, fan; 5, personalized oxygen supply terminal; 6, oxygen detection device; 51, air outlet pipe; 61, probe; 62, oxygen detector. DETAILED DESCRIPTION

[0020] The preferred embodiments of the present application are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to explain and illustrate the present application, and are not used to limit the present application, and the embodiments in the present application and the features in the embodiments can be combined with each other without conflict.

[0021] Embodiment The embodiment provides a low-energy-consumption non-contact targeted oxygen supply system, which comprises an oxygen supply device body 1, a static pressure tank 2, an oxygen supply hose 3, a fan 4, a personalized oxygen supply terminal 5 and an oxygen detection device 6 for sampling the oxygen concentration of a sampling area. The personalized oxygen supply terminal 5 comprises a plurality of air outlet pipes 51, and the air outlet pipes 51 have adjustable oxygen supply parameters. Further comprising a joint optimization module configured to perform iterative optimization on the oxygen supply concentration and the oxygen supply parameters according to a Kriging-multiple island genetic algorithm, so as to obtain an optimal oxygen supply parameter combination and achieve optimal oxygen supply.

[0022] Reference Figure 1 The oxygen supply device body 1 is an oxygen tank, the oxygen supply device body 1 delivers oxygen to the static pressure tank 2 through the oxygen supply hose 3, the fan 4 delivers air to the static pressure tank 2 through the oxygen supply hose 3, the oxygen supply device body 1 and the fan 4 are connected in parallel to deliver air and oxygen to the static pressure tank 2, and the air and the oxygen are mixed in the static pressure tank 2 and then delivered to the air outlet pipes 51 through the hose for air outlet, thereby forming an oxygen delivery pipeline.

[0023] The air outlet pipes 51 are provided in two layers and arranged in pairs, the inner diameter D of the air outlet pipes 51 is 1.25 cm, the downward inclination angle theta is 9.83°, and the inward inclination angle beta is 22.95°, in order to realize positioning of the oxygen supply outlet in three-dimensional space, the multi-degree-of-freedom positioning mechanism in the embodiment adopts a positioning mechanism combined with a three-axis sliding table and a spherical hinge as a preferred implementation, the three-axis sliding table provides linear displacement of X, Y and Z directions by ±50 mm, a micro spherical hinge is arranged at the end to realize fine adjustment and angle compensation of the attitude in the theta and beta directions, and the nozzle can be accurately aligned with the breathing area in combination, so that the oxygen supply targeting is improved. The linear sliding table and the micro spherical hinge adopt linear modules and precision spherical hinge parts commonly used in the market, such as Thorlabs and MISUMI, or other mechanisms, which are not limited herein; the multi-degree-of-freedom positioning mechanism further comprises a 3D printing variable nozzle, and the inner diameter of the air outlet pipe 51 is adjusted by a stepping motor of the 3D printing variable nozzle.

[0024] The flow regulating valve is installed on the oxygen supply device body 1 to control the oxygen output and adjust the oxygen concentration. The fan speed regulator is arranged on the fan 4 to control the air outlet speed. The uniform flow orifice plate is arranged in the static pressure box 2. The uniform flow orifice plate is a single-layer structure and is arranged in the static pressure box 2 and located downstream of the oxygen and air inlets and about 5 cm away from the outlet of the oxygen supply hose 3. The orifice diameter of the uniform flow orifice plate is 2 mm. The orifice diameters are consistent. The orifice spacing is 4 mm. The orifice spacing is arranged in a matrix. The opening rate of the uniform flow orifice plate in the embodiment is 40%. The oxygen and air mixing uniformity is greater than or equal to 95%. The opening rate calculation method is: opening rate = (total orifice area / orifice plate area) x 100%. The low-resistance regulating valve is further installed on the oxygen supply hose 3 to steplessly adjust the outlet flow speed of the air outlet pipe 51 in the range of 0.2 m / s to 1.5 m / s. The oxygen supply hose 3 adopts the corrugated hose with an inner diameter of 25 mm. The flow resistance coefficient is 0.018. The sampling probe 61 of the oxygen detector 62 is used to detect the oxygen concentration above the human face breathing area. The orifice diameter of the sampling probe 61 is 4 mm. The pump suction type is used for sampling. The flow rate is 50 mL / min. The response time is 1-5 s.

[0025] The processing result of the joint optimization module is received by the MCU. The multi-degree-of-freedom positioning mechanism is controlled to adjust the inner diameter D, the inner inclination angle β and the downward inclination angle θ of the air outlet pipe 51. The flow regulating valve of the oxygen supply device body 1 is controlled to adjust the oxygen supply concentration C to meet the slow change requirement of the sleep environment. Referring to Figure 6 It should be noted that D is the inner diameter of the air outlet pipe 51. β is the inner inclination angle. It is the included angle between the axis of the air outlet pipe 51 and the vertical direction. θ is the downward inclination angle. It is the included angle between the axis of the air outlet pipe 51 and the horizontal direction.

[0026] Referring to Figures 2-4 The test is carried out on the subjects in the indoor environment with an altitude of 3650 m in Lhasa. The subjects have pulse type blood oxygen meters on their fingers. The target point array of the breathing area is arranged in the z=0.36 m plane above the face of the supine human body. The center of the face breathing area is taken as the reference point. The lateral measuring points are arranged along the plane. The array has a total of 20 measuring points. The measuring points are distributed in a three-dimensional equidistant grid. The three-axis spacing is 5 cm (i.e. 5 cm x 5 cm x 5 cm). The horizontal projection distance between the array center and the oxygen supply air outlet is about 0.30 m. The shortest horizontal distance between the array close to the wall surface of the room and the wall surface is about 0.20 m.

[0027] Referring to Figures 4-5 As shown in the figure, the measuring points are sequentially arranged according to the numbers 1-20. The measuring point position coordinates, numbers and relative geometric relationships are marked in the figure. The overall pressure drop of the system is less than or equal to 80 Pa. When the outlet flow speed of the air outlet pipe 51 is 0.5 m / s, the average wind speed of the breathing area is less than or equal to 0.16 m / s. The oxygen sampling device 6 samples the 20 measuring points to measure the oxygen concentration.

[0028] Figure 9 The average wind speed of the target area under different air outlet speeds and oxygen supply concentrations; Figure 10 The average oxygen concentration of the target area under different air outlet speeds and oxygen supply concentrations; Comprehensive Figure 9 And Figure 10 The data show that the oxygen concentration in the personnel breathing area is mainly affected by the oxygen supply concentration, and the wind speed is mainly related to the air outlet speed of the oxygen supply port.

[0029] In this embodiment, the system further comprises a three-dimensional modeling process, which adopts a local encryption grid CFD analysis, including: S1, establishing a non-structured grid of the room; Adopting unstructured tetrahedral mesh; S2, taking a plurality of air outlet pipes 51 and a human head as an oxygen supply area, locally encrypting the oxygen supply area, and determining the number of grids; Referring to Figure 7 , the grid density is 5-8 times that of other areas of the room; In order to reduce the influence of grid division on the numerical simulation results, the grid independence verification of the oxygen supply room model is carried out.

[0030] Referring to Figure 8 , grid independence verification: through the calculation of 5 sets of different density grids 697K-2874K, finally select 2283K grid, both ensure the accuracy and control the calculation cost.

[0031] By solving the models with five different grid numbers 697K, 1251K, 1888K, 2283K and 2874K, and comparing the concentration targeting values calculated under the five grid numbers, only the grid number is changed during the grid independence verification, and other simulation conditions do not change, the iteration step number is 8000 steps, and the calculation formula is as follows: From Figure 8 , it can be seen that when the grid number increases from 2283K to 2874K, continuing to encrypt the grid will not have a great influence on the concentration targeting value, and under the premise of ensuring the calculation accuracy, considering the saving of calculation time and calculation cost, the final grid number is determined to be 2283K.

[0032] Among them, the maximum grid size is 3.258415e×10-4m³, the minimum grid size is 2.397623e×10-10m³, the maximum grid area is 1.149430×10-8m2, and the minimum grid area is 1.149430×10-8m2.

[0033] S3, setting boundary conditions; Outlet: velocity-inlet, velocity is set according to experiment; Side wall: pressure-outlet; Other wall: wall, no slip; Human head: temperature 307.6K, heat flux density 7.4W / m².

[0034] S4, selecting a turbulence model; Referring to Figure 11 , the standard k-ε model, the RNG k-ε model, the Realizable k-ε model and the Standard k- -ω model are selected for numerical simulation calculation, and by comparing the oxygen concentration values of each turbulence at the measuring point with the experimental values, the Realizable k-ε model is closest to the experimental values, so the Realizable k-ε model is selected for subsequent numerical simulation work; The Realizable k-ε model is selected, which has the highest accuracy in jet simulation; the grid is greater than or equal to 2.2M, the residual is less than or equal to 10 -3 ; all variable residual curves are stable; the iteration step is greater than or equal to 8000 steps.

[0035] S5, obtaining a three-dimensional simulation model through experimental verification; Through experimental verification, the simulation result has the smallest deviation from the measured value, and a three-dimensional simulation model is obtained.

[0036] The joint optimization module iteratively optimizes the simulation calculation results of the three-dimensional simulation model.

[0037] The joint optimization module uses the calculation results of the three-dimensional simulation model as evaluation input to iteratively optimize the oxygen supply concentration and oxygen supply parameters; in the offline stage, an agent model is constructed based on the three-dimensional simulation results, and in the online stage, the agent is preferentially used for rapid evaluation and the three-dimensional simulation is called as needed for verification and sampling.

[0038] In this embodiment, the Kriging-multiple island genetic algorithm includes an optimization objective, and the optimization objective is MinC target , and further includes a constraint condition: C avg (D, θ, β, C)≥24%vol; wherein C is the oxygen supply concentration; C target is the concentration target value; and C avg is the average concentration value.

[0039] In this embodiment, the optimization process of the Kriging-multiple island genetic algorithm is as follows: S1: offline sampling stage: using optimal Latin hypercube sampling method, generating sample set in the variable range of preset oxygen supply concentration C and oxygen supply parameters D, β, θ, calculating the corresponding concentration targeting value C through three-dimensional simulation model target with the average oxygen concentration C avg , obtaining a sample library; In order to quickly screen out the key factors affecting the oxygen supply performance of the breathing area and determine the reasonable value range of each factor, the orthogonal experimental design is adopted as the preliminary test scheme in this embodiment.

[0040] The specific steps are as follows: 1. Determine the test factors and levels Factor selection: based on process adjustability and engineering experience, four main factors are selected: air outlet pipe diameter D, inner inclination angle β, downward inclination angle θ and oxygen supply concentration C; Level setting: in order to cover the engineering feasible range and form a three-level orthogonal scheme, each factor is set to three levels, and the level value is determined according to the equipment realizable range and comfort / safety constraints.

[0041] 2. Test scheme and implementation L9(3 4 ) orthogonal table is used for test (9 tests cover the orthogonal table of 4 factors and 3 levels), if higher resolution is needed, larger scale orthogonal table can be used; Each orthogonal combination is solved by three-dimensional local encryption CFD, for each test result, the average oxygen concentration C avg and the concentration targeting value C target are calculated, and the system operating parameters and boundary conditions are recorded.

[0042] 3. Data analysis method Range analysis and variance analysis are used to analyze the test data to identify the factors and their preferred levels that significantly affect C avg and C target .

[0043] 4. Result application (boundary and preferred level determination) According to the results of orthogonal test analysis, the sensitive interval and preferred level of each factor are determined, and these intervals are used as the variable range of subsequent optimal Latin hypercube sampling, obtaining the finally used D∈[1,2]cm, β∈[10°,30°], θ∈[5°,15°], C∈[30%,50%], which is the recommended variable range based on orthogonal test and engineering constraints.

[0044] In this embodiment, the orthogonal test results show that the air outlet pipe diameter D and the inner inclination angle β have the greatest impact on C target , the downward inclination angle θ is second, and the oxygen supply concentration C has the least impact on C avgThere is a significant impact, so in the optimal Latin hypercube sampling, more sampling density should be given in the D and β directions or priority should be given to coverage.

[0045] Table 1 orthogonal experimental design table Referring to Table 1: The oxygen supply parameter variable range is determined by the orthogonal test to determine the oxygen supply parameter boundary value.

[0046] Referring to Table 2: Optimal Latin hypercube 50 samples → CFD calculation → sample library.

[0047] Table 2 sampling table In order to efficiently construct a representative training sample library in a multi-dimensional continuous design space, the embodiment adopts optimal Latin hypercube sampling to generate 50 samples, and the response quantity C avg and C target , the specific implementation steps are as follows: 1. Variable space and sample number selection Variable space: Based on the results of the orthogonal test and engineering constraints, the variable range is determined as: D∈[1.0,2.0]cm, β∈[10°,30°], θ∈[5°,15°], C∈[30%,50%].

[0048] Sample number selection: The sample number should be sufficient to cover the representative distribution of the four-dimensional space, and the CFD calculation cost should be considered, and the embodiment preferably selects 50 samples, as shown in Table 2, which can control the amount of calculation while maintaining the model generalization ability; the sample number range can be written as 30-100 to adapt to different computing resources.

[0049] 2. Optimal Latin hypercube generation method Generation tool and random seed: Use Matlab to generate Latin Hypercube initial matrix, and set fixed random seed to ensure repeatability.

[0050] Matlab can use Matlab rng(2025), and Python can also be used for generation, using pyDOE / Scipy.

[0051] Optimality criterion: Perform multiple random iterations on the initial optimal Latin hypercube sampling based on the maximum minimum distance criterion or uniformity criterion, and select the optimal matrix, or use simulated annealing / particle swarm and other metaheuristic algorithms to improve the optimal Latin hypercube sampling to obtain the optimal point distribution.

[0052] The embodiment adopts the maximum minimum distance criterion strategy and performs 2000 times of random rearrangement to obtain the optimal Latin hypercube.

[0053] Continuous variable mapping: map the 50-unit interval of each dimension in the optimal Latin hypercube sampling to the actual numerical range of the corresponding variable (uniform distribution mapping); if a logarithmic or other nonlinear transformation is used for a certain variable, perform the corresponding transformation before mapping.

[0054] 3. Sample quality inspection Minimum distance inspection: calculate the minimum Euclidean distance between any two points in the sample set to ensure that it is not less than the preset threshold.

[0055] Projection uniformity inspection: check the projection in each dimension to ensure that each of the 50 intervals in each dimension is occupied by a sample, and draw a 2D projection scatter plot for manual review to determine whether there is aggregation.

[0056] If the sample does not meet the uniformity or minimum distance standard, increase the number of iterations or regenerate the sample until the quality criteria are met.

[0057] 4. Linkage execution process with CFD Batch processing strategy: split the 50 samples into 5 batches for parallel jobs, 10 samples per batch, and submit CFD solutions in parallel on a high-performance computing cluster or multi-core workstation to shorten the overall time consumption.

[0058] CFD input / output convention: the CFD input file corresponding to each sample is recorded with a standardized file name and parameters, and after the CFD solution is completed, the output script automatically extracts the concentration of the target measurement point in the breathing zone and calculates the C avg , C target , and writes back to the sample library.

[0059] Obtain the oxygen concentration distribution of the target measurement point in the breathing zone through local encrypted grid CFD analysis, calculate the oxygen concentration targeting value C target ; based on the oxygen concentration distribution of the three-dimensional measurement point array in the breathing zone, calculate the average oxygen concentration C avg through weighted average, and obtain the sample library.

[0060] Oxygen concentration targeting value C target is calculated according to the following formula: C target =Σ(C i -C avg )² / n n is the total number of measurement points; Ci is the oxygen concentration of the i-th measurement point; C avg is the average oxygen concentration of all measurement points, used to evaluate the overall oxygen supply level of the breathing zone, with the unit %vol; C target is the concentration targeting value, used to measure the uniformity or concentration of oxygen concentration in the breathing zone.

[0061] Data validation: convergence and quality check for each CFD solution, if a sample does not converge or has abnormal results, it is recorded and marked for re-computation or rejection, and then additional samples are taken until the number of sample library meets the predetermined value.

[0062] 5. Subsequent model training and sampling strategy Agent training: train the Kriging model using the above sample library as the training set, and calculate R² and RMSE through cross-validation.

[0063] If the Kriging performance is lower than the threshold, for example, R²<0.90, trigger adaptive sampling based on prediction variance: in the positions with the highest prediction variance of the agent, this embodiment uses 5 each, generate new local samples and submit CFD calculation, integrate the new samples into the training set and retrain until the performance requirements are met or the sample upper limit is reached.

[0064] S2: Agent model establishment: train the Kriging model based on the sample library, use Gaussian kernel function and maximum likelihood estimation, and the model determination coefficient R²≥0.95; The model is set as follows: The regression term uses a 0-order polynomial (ordinary Kriging); the correlation function uses a Gaussian kernel function: Hyperparameter estimation: maximum likelihood method, automatically completed by Matlab toolbox fitrgp; Accuracy verification: leave-ten cross-validation, determination coefficient; Reference Figure 12 , R²=1−SS res / SS tot =0.952, which meets the engineering accuracy requirements.

[0065] S3: Online optimization phase: as shown in Figure 13 and Figure 14 , use the multi-island genetic algorithm to search for the optimal oxygen supply parameter combination on the trained Kriging agent model.

[0066] In the online phase, the joint optimization module first runs the multi-island genetic algorithm on the verified Kriging agent model to perform fast parallel search; for high-potential candidate solutions generated by the multi-island genetic algorithm on the agent, the system selects some of them according to the preset priority or trigger rule and submits them to the three-dimensional local encryption CFD or field sensor for high-fidelity verification; the verification results are used to update the sample library and retrain the Kriging agent, thereby forming a closed-loop iterative optimization process of agent-realistic simulation Table 3: Multiple Island Genetic Algorithm Parameter Setting Table Table 3: Multiple Island Genetic Algorithm Parameter Setting Table It should be noted that the values in Table 3 are preferred embodiments, and do not constitute a limitation of the present application.

[0067] The online phase runs the Multiple Island Genetic Algorithm (MIGA) on the trained and validated Kriging surrogate model to perform a constrained optimization search.

[0068] Optimization objective: Minimize C target (D, θ, β, C) Constraints: C avg (D, θ, β, C) ≥ 24% vol, average wind speed in breathing zone ≤ 0.16 m / s, overall system pressure drop ≤ 80 Pa.

[0069] Through the optimization of the Multiple Island Genetic Algorithm, the optimal oxygen supply concentration C and the oxygen supply parameters D, β, θ are determined, as shown in Table 4: Table 4: Optimal Parameter Table The optimal solution in Table 4 is an example value obtained by searching on the Kriging surrogate through the Multiple Island Genetic Algorithm and passing the local CFD verification under the model, grid and boundary conditions. It is an embodiment and does not constitute a necessary limitation of the claims of the present application.

[0070] The obtained optimal parameter combination is transmitted to the MCU through the communication interface RS-485, and the flow control valve of the oxygen supply device body 1 is controlled by the MCU to control the oxygen output, adjust the oxygen supply concentration C, and adjust the oxygen flow feedback through the oxygen monitoring device 6 to form accurate control of the oxygen supply concentration C. The angle θ, β is adjusted by the three-axis sliding table ball hinge combined mechanism, and the inner diameter D of the air outlet pipe 51 can be directly replaced by a 3D printed shell with the optimal solution inner diameter.

[0071] The technical characteristics of the low-energy non-contact targeted oxygen supply system are as follows: 1. Compared with the traditional diffusion or mask oxygen supply mode, the system utilizes a side multi-angle jet to accurately guide the oxygen-rich airflow to the breathing area of the person in a supine position, changes "whole room oxygen supply" to "targeted oxygen supply", and increases the oxygen utilization rate by more than 60%, thereby significantly reducing the energy consumption of oxygen production and transportation; 2. The Kriging surrogate model and the multi-island genetic algorithm are integrated in the device to realize online optimization of the inner diameter, angle and oxygen supply concentration of the air outlet pipe, avoid repeated manual debugging, and ensure the consistency of the oxygen supply effect under different altitudes and individual differences; 3. The wind speed and oxygen concentration detection are built-in, the breathing area environment is monitored in real time, the fan and the flow valve are linked and controlled to maintain the "oxygen without wind" comfortable interval (wind speed ≤ 0.16 m / s, oxygen concentration ≥ 24%), and the blowing feeling and energy consumption are reduced; 4. The system adopts a modular 3D printed shell and a quick-connection hose, and the room structure does not need to be modified on site, the installation is completed in 15 minutes, the maintenance period is extended to more than 1 year, and the operation and maintenance cost is reduced; 5. Experimental verification shows that under the altitude of 3650m, the blood oxygen saturation of the subjects can be increased by an average of 11.3%, the heart rate is reduced by 10bpm, and the risk of altitude reaction is significantly reduced. In summary, the targeted oxygen supply system can accurately, low-energy and comfortably solve the problem of low oxygen in high altitude sleep, the control method takes into account individual adaptation and scene matching, avoids the energy waste and wearing discomfort of the traditional oxygen supply mode, has low operation and management cost, and has low control difficulty. If the technology is widely popularized and applied, the sleep quality and health level of underground space, plateau residents, tourists and stationed soldiers can be improved, the oxygen production and operation cost can be greatly reduced, and remarkable economic benefits and social benefits can be achieved.

[0072] Although the present application has been disclosed as above with examples, it is not intended to limit the present application, and any person skilled in the art can make some changes or modifications to the equivalent embodiments of the above disclosed technical contents without departing from the scope of the technical solutions of the present application. Any simple modification, equivalent change and modification made to the above examples according to the technical essence of the present application still belong to the scope of the technical solutions of the present application.

Claims

1. A low-energy, non-contact, targeted oxygen supply system, characterized in that: It includes an oxygen supply device body (1), a static pressure box (2), an oxygen supply hose (3), a fan (4), a personalized oxygen supply terminal (5), and an oxygen detection device (6) for sampling the oxygen concentration in the sampling area. The personalized oxygen supply terminal (5) includes multiple air outlet pipes (51), and the air outlet pipes (51) have adjustable oxygen supply parameters; It also includes a joint optimization module, which iteratively optimizes the oxygen supply concentration and oxygen supply parameters according to the Kriging-multi-island genetic algorithm to obtain the optimal combination of oxygen supply parameters in order to achieve the best oxygen supply.

2. The low-energy non-contact targeted oxygen supply system according to claim 1, characterized in that: The system also includes a 3D modeling process, including: S1. Create an unstructured grid for the room; S2. Take multiple air outlet pipes (51) and the human head as oxygen supply areas, locally densify the oxygen supply areas, and determine the number of grids. S3. Set boundary conditions; S4. Select the turbulence model; S5. A three-dimensional simulation model was obtained through experimental verification; The joint optimization module iteratively optimizes the simulation results of the three-dimensional simulation model.

3. The low-energy non-contact targeted oxygen supply system according to claim 1, characterized in that: The oxygen supply parameters include inner diameter D, downward tilt angle β, and inward tilt angle θ; the Kriging-multi-island genetic algorithm includes an optimization objective, which is: MinC target It also includes constraints: C avg (D,θ,β,C)≥24%vol; where C is the oxygen supply concentration; C target For concentration target value; C avg This represents the average concentration value.

4. The low-energy non-contact targeted oxygen supply system according to claim 3, characterized in that: The optimization process of the Kriging-multi-island genetic algorithm is as follows: S1: Offline sampling stage: The optimal Latin hypercube sampling method is used to generate a sample set within the range of preset oxygen supply concentration C and oxygen supply parameters D, β, and θ. The corresponding concentration target value C is then calculated using a three-dimensional simulation model. target With average oxygen concentration value C avg This yields the sample library; S2: Proxy model establishment: The Kriging model is trained based on the sample database, using Gaussian kernel function and maximum likelihood estimation. The model determination coefficient R² ≥ 0.

95. S3: Online optimization phase: The multi-island genetic algorithm is used to perform optimization search on the trained Kriging surrogate model to quickly generate the optimal combination of oxygen supply parameters.

5. The low-energy non-contact targeted oxygen supply system according to claim 4, characterized in that: In S1, the concentration target value C target Calculate using the following formula: C target =Σ(C i -C avg )² / n; In the formula, n is the total number of measuring points; C i The oxygen concentration at the i-th measuring point, in %vol; C avg The average oxygen concentration at all measuring points is used to evaluate the overall oxygen supply level in the breathing zone, expressed as %vol; C target The concentration target value is used to measure the uniformity or concentration of oxygen concentration in the breathing zone.

6. The low-energy non-contact targeted oxygen supply system according to claim 4, characterized in that: In S1, the range of the oxygen supply concentration C is: C∈[30%, 50%]; The variable ranges of oxygen supply parameters D, β, and θ are: D∈[1,2]cm, β∈[10°,30°], θ∈[5°,15°].

7. The low-energy non-contact targeted oxygen supply system according to claim 1, characterized in that: The sampling area is arranged in a three-dimensional grid with the facial breathing area as the center. The nodes of the three-dimensional grid are the set measurement points, and the distance between each two measurement points ranges from 1cm×1cm×1cm to 8cm×8cm×8cm.

8. The low-energy non-contact targeted oxygen supply system according to claim 7, characterized in that: The oxygen detection device (6) includes an oxygen detector (62) and a sampling probe (61) located above the breathing area of ​​the human face, the sampling probe (61) being used to collect oxygen concentration at the measuring point.

9. A low-energy non-contact targeted oxygen supply system according to claim 1, characterized in that: The personalized oxygen supply terminal (5) also includes a multi-degree-of-freedom positioning mechanism for adjusting the oxygen supply parameters of each air outlet pipe (51).

10. A low-energy non-contact targeted oxygen supply system according to claim 1, characterized in that: In S3, the multi-island genetic algorithm sets the number of islands to 10, with a population size of 10 individuals per island; it evolves for 1000 generations, with the crossover probability preferably set to 0.8-1.0, the mutation probability preferably set to 0.005-0.02, and the migration interval to 5 generations.

Images