Control system and method based on integrated intelligent micro-high-pressure oxygen-enriched cabin

Abstract

The invention discloses a control system and method based on an integrated intelligent micro-high-pressure oxygen-enriched cabin, and the system comprises a data collection module which collects the real-time physiological parameters of a user and the environment data in the cabin, and obtains a multi-modal fusion data set; the decision obtaining module is used for constructing a high-fidelity physiological deduction model to predict the health trend and oxygen therapy response of the user by utilizing an adversarial digital twinborn body and combining a physiological environment dynamic causal mining mechanism based on the multi-modal fusion data, and obtaining a personalized oxygen therapy optimization scheme; the structure of the high-fidelity physiological deduction model comprises a causal antagonism digital twin body improved through a physiological environment dynamic causal mining mechanism, a causal generator and a causal discriminator; the intelligent control module is used for adjusting the pressure, the oxygen concentration and the treatment duration of the micro hyperbaric oxygen chamber in real time on the basis of a personalized oxygen therapy optimization scheme, and intelligent control is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent control, and particularly relates to an integrated intelligent micro-high-pressure oxygen-enriched cabin control system and method. BACKGROUND

[0002] As a key equipment in the field of medical treatment and rehabilitation, the oxygen cabin aims to provide an environment with controlled air pressure and high-concentration oxygen to assist patients in rehabilitation and treatment. In the civil field, it has gradually become a new choice for health care. Micro-pressure oxygen cabins are gradually expanding from traditional medical professional scenes to diversified scenes such as families, health care centers, and sports rehabilitation institutions. For example, in some marathon exhibition events, the micro-pressure oxygen cabin of Haier Biomedical provides pre-race oxygen supplementation services for athletes to help them maintain the best state.

[0003] At present, the micro-pressure oxygen cabin on the market is constantly innovating in hardware design. From the cabin material, there are hard oxygen cabins made of aviation-grade aluminum alloy materials and foldable soft oxygen cabins made of high-molecular flexible materials to meet the needs of portability and stability in different scenes. However, in terms of function implementation, although some oxygen cabins have introduced the concept of intelligence, they have realized basic functions such as touch screen control and data monitoring, but in terms of intelligent control, they still remain at the stage of simple threshold control, that is, only when a certain index exceeds the preset threshold during oxygen supply, the micro-pressure oxygen cabin is adjusted. The overall oxygen supply quality, physiological feedback, and causal relationship are not comprehensively detected and deeply intelligently controlled. Therefore, it is not possible to predict and actively optimize the oxygen therapy scheme in advance according to the real-time changes of multiple physiological parameters of the user combined with the causal rules. Therefore, an integrated intelligent micro-high-pressure oxygen-enriched cabin control system and method are proposed. SUMMARY

[0004] In order to overcome the above-mentioned defects of the prior art, in order to achieve the above-mentioned purposes, the present application proposes the following technical solutions:

[0005] An integrated intelligent micro-high-pressure oxygen-enriched cabin control system comprises:

[0006] A data acquisition module acquires real-time physiological parameters of a user and cabin environment data to obtain a multi-modal fusion data set;

[0007] A decision acquisition module, based on the multi-modal fusion data, uses an adversarial digital twin combined with a physiological environment dynamic causal mining mechanism to construct a high-fidelity physiological deduction model to predict user health trends and oxygen therapy responses and obtain an individualized oxygen therapy optimization scheme;

[0008] The structure of the high-fidelity physiological deduction model comprises an improved causal adversarial digital twin, a causal generator, and a causal discriminator through a physiological environment dynamic causal mining mechanism;

[0009] Intelligent control module: based on the real-time adjustment of the individualized oxygen therapy optimization scheme, the pressure, oxygen concentration and treatment time of the micro-high pressure oxygen cabin are adjusted to realize intelligent control.

[0010] The physiological parameters include heart rate data, blood oxygen data, and blood pressure data, and the cabin environment data includes temperature data, humidity data, oxygen concentration data, and pressure data.

[0011] The physiological environment dynamic causal mining mechanism utilizes a causal inference algorithm to sort out the causal relationship between the user's physiological parameters and the cabin environment from the multi-modal data, and then realizes it in combination with adversarial learning training.

[0012] The causal adversarial digital twin acquisition process is:

[0013] A basic adversarial digital twin is constructed and embedded with a causal correlation graph:

[0014] Based on the causal mining algorithm, historical multi-modal data is analyzed to mine the causal relationship between environmental parameters and physiological parameters, and these causal relationships are encoded as a graph structure and stored in the basic digital twin to obtain a causal adversarial digital twin.

[0015] The causal correlation graph acquisition process is:

[0016] A constraint-based PC algorithm is selected to mine the causal connection relationship between variables from multi-modal fusion data through conditional independence testing, and a causal graph structure is constructed, i.e., testing the conditional independence of variable pair (X, Y) given the variable set Z, if (X Y | Z), then deleting the direct edge between X and Y;

[0017] Multi-modal fusion data is collected, and all data in the multi-modal fusion data is taken as a node (X, Y) to construct a fully connected undirected graph;

[0018] For each pair of nodes (X, Y), given the other node subset Z, use partial correlation analysis to test whether it is independent, if it is independent, delete the edge between X and Y, if it is not independent, keep the edge and mark the condition set Z;

[0019] The window length is set to 1 hour and the time sliding window step is 10 minutes, for the data in each window, causal discovery and effect quantification are performed;

[0020] After each window mines the local causal graph, it is fused with the global causal graph, the stable causal edges are retained by using a graph merging algorithm to update the dynamically changing causal edges, and a dynamically evolving causal correlation graph G(t) is formed.

[0021] The causal generator acquisition process is:

[0022] Based on deep neural network combined with multi-modal fusion data, features are extracted to generate a basic generator for simulating physiological parameter changes;

[0023] A causal constraint block is added after the hidden layer of the basic generator. When generating simulated physiological data, the causal constraint block calls the causal correlation atlas of the digital twin, and forces the generation process to follow the causal logic to obtain a causal generator;

[0024] The parameters of the discriminator are fixed, and the generator simulation data is input into the discriminator. The generator loss function also incorporates causal consistency constraints, and the formula is:

[0025]

[0026] Where L G is the loss value of the generator, z represents the input noise of the generator, G(z) represents the simulated data generated by the generator using noise z, D(G(z) represents the probability that the discriminator judges G(z) to be real data, C(G(z)) represents the probability that the discriminator judges G(z) to comply with causal logic, and λ is the causal consistency loss weight.

[0027] The causal discriminator acquisition process is:

[0028] Based on deep neural network, a basic discriminator is constructed, and causal consistency features are added to the input of the discriminator to obtain a causal discriminator;

[0029] Real data and simulation data are input into the discriminator, and on the basis of traditional cross-entropy, causal consistency loss is incorporated to train the causal discriminator, and the formula is:

[0030]

[0031] Where x is the input data, C(x) is the causal consistency discrimination function, λ is the causal consistency loss weight, P data is the real data distribution, P z is the input noise distribution of the generator, D(x) is the data distribution discrimination probability, G(z) is the generator simulation data, D(x) is the probability that the discriminator judges real data x to be real data, C(G(z) is the probability that the discriminator judges simulation data G(z) to comply with causal logic, and D(G(z) represents the probability that the discriminator judges simulation data G(z) to be real data.

[0032] The causal consistency feature acquisition process is:

[0033] The continuously collected real-time multi-modal fusion data is cut into small fragments according to a fixed time window, and the difference between the first and last data points in the slice is compared with the corresponding data change rate to obtain the actual causal effect feature b1;

[0034] According to prior knowledge, the starting change time of the environmental parameter in the multi-modal fusion data is defined as the first change quantity, and the starting change time of the physiological parameter response is defined as the second change quantity, the data segment is observed, and the difference between the first change quantity and the second change quantity is obtained to obtain a causal time sequence feature b2;

[0035] If the first change quantity-second change quantity is negative, the feature value is set to 0, and if the first change quantity-second change quantity is positive, the feature value is set to 1.

[0036] Based on the actual causal effect feature and the causal time sequence feature, a causal consistency feature vector B=(b1, b2) is obtained.

[0037] A control method based on an integrated intelligent micro-high-pressure oxygen-enriched cabin comprises:

[0038] S1: Collecting real-time physiological parameters of users and cabin environmental data to obtain a multi-modal fusion data set;

[0039] S2: Based on the multi-modal fusion data, a high-fidelity physiological deduction model is constructed by using an adversarial digital twin combined with a physiological environment dynamic causal mining mechanism to predict user health trends and oxygen therapy responses, and an individualized oxygen therapy optimization scheme is obtained;

[0040] S3: Based on the individualized oxygen therapy optimization scheme, the pressure, oxygen concentration and treatment time of the micro-high-pressure oxygen cabin are adjusted in real time to realize intelligent control.

[0041] The present application has the following beneficial effects:

[0042] In the present application, first, through the high-fidelity physiological deduction model, individualized oxygen therapy schemes can be tailored for different users according to their age, gender, underlying diseases, real-time physiological parameters and other individual characteristics, combined with the causal relationship between environmental and physiological parameters in the causal correlation graph. For example, for elderly users with cardiovascular diseases, the model can accurately adjust the oxygen cabin pressure rate and pressure upper limit according to the causal effect of pressure change on the cardiovascular system to avoid cardiovascular risks caused by pressure fluctuations; for users who need to recover physical fitness after exercise, the oxygen concentration and treatment time are optimized according to the causal logic of exercise intensity and body metabolism demand to improve the oxygen therapy effect and significantly improve the user's oxygen therapy experience and effect.

[0043] Secondly, the introduction of the causal constraint block forces the generator to follow the causal logic when simulating physiological parameter changes and adjusting oxygen cabin environmental parameters. For example, during the adjustment of oxygen cabin pressure and oxygen concentration, real-time monitoring of human physiological parameters (heart rate, blood oxygen saturation, etc.) is performed, and once it is found that the actual physiological response does not conform to the causal law (such as the oxygen concentration increases but the blood oxygen saturation does not rise as expected), the parameter change is immediately suspended or adjusted to prevent harm to the human body caused by unreasonable environmental parameter settings, and to provide a safer and more reliable oxygen therapy environment for users.

[0044] Finally, based on real-time monitoring of environmental and physiological parameters, and combined with causal logic judgment, the system achieves intelligent dynamic adjustment of oxygen chamber pressure, oxygen concentration, and treatment duration. For example, during oxygen therapy, if the system detects that the user's heart rate increases due to pressure rise and reaches a threshold that may affect health, it automatically reduces the rate of pressure increase or appropriately lowers the pressure value based on causal relationships. At the same time, it dynamically adjusts the oxygen concentration according to the changing trend of blood oxygen saturation, always keeping the oxygen therapy process in an optimal state without frequent manual intervention, thus improving the level of intelligence in oxygen therapy. Attached Figure Description

[0045] Figure 1 This is a system block diagram of an integrated intelligent micro-high pressure oxygen-enriched chamber control system and method proposed in this invention.

[0046] Figure 2 This is a flowchart illustrating the steps of an integrated intelligent micro-high pressure oxygen-enriched chamber control system and method proposed in this invention. Detailed Implementation

[0047] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0048] Example 1

[0049] like Figure 1 As shown, the present invention proposes an integrated intelligent micro-high pressure oxygen-enriched chamber control system, comprising:

[0050] Data acquisition module: Collects real-time physiological parameters of users and cabin environment data to obtain multimodal fusion dataset;

[0051] Physiological parameter sensor acquisition includes:

[0052] Heart rate sensor: A photoelectric heart rate sensor, such as MAX30102, is selected. The measurement range is 30-240 beats / minute and the accuracy is ±2 beats / minute. Heart rate data is collected by wearable devices such as wristbands or chest straps installed in the oxygen-enriched chamber.

[0053] Blood oxygen sensor: A reflective blood oxygen sensor, such as the MAX30105, is used. It can simultaneously measure blood oxygen saturation and heart rate. The blood oxygen saturation measurement range is 70%-100%, with an accuracy of ±2%. The installation position is the same as that of the heart rate sensor, and it collects blood oxygen data.

[0054] Blood pressure sensor: Select non-invasive blood pressure sensor such as MPX5010GP, which measures the range of 0-10kPa, accuracy of ±0.2kPa to collect blood pressure data;

[0055] Environmental sensor collection includes:

[0056] Temperature sensor: Choose DS18B20 digital temperature sensor, measurement range of -55℃-125℃, accuracy of ±0.5℃, installed in different positions in the cabin, such as top, middle and bottom, to monitor the temperature in the cabin;

[0057] Humidity sensor: SHT30 temperature and humidity sensor is used, humidity measurement range is 0%-100%RH, accuracy is ±2%RH, and it is installed together with the temperature sensor to collect humidity data synchronously;

[0058] Oxygen concentration sensor: Choose ZO-03 oxygen sensor, measurement range is 0%-25%, accuracy is ±0.5%, installed in the gas circulation of the cabin to ensure accurate measurement of oxygen concentration and obtain oxygen concentration data;

[0059] Pressure sensor: MPX2050 pressure sensor is used, measurement range is -50kPa-50kPa, accuracy is ±1kPa, installed inside the cabin body to monitor the pressure data in the cabin;

[0060] Communicate with each sensor through I 2C, SPI or UART interface, read sensor data according to the set sampling frequency, the sampling frequency of physiological parameters is set to 10 times per second to capture subtle changes in parameters, the sampling frequency of environmental data is 1 time per minute, preprocess the collected data, including removing noise and filling missing values, for noise data, use moving average filtering method for processing, the formula is:

[0061] Where, y i is the i-th data after filtering, x j is the original data, n is the size of the moving window, here n is 5, for missing values, use linear interpolation method to fill in, that is, estimate the size of missing values according to the valid data before and after the missing values;

[0062] The preprocessed data is transmitted to the scheme acquisition module through wireless communication.

[0063] Decision acquisition module: Based on multi-modal fusion data, use adversarial digital twin combined with physiological environment dynamic causal mining mechanism to construct high-fidelity physiological deduction model to predict user health trend and oxygen therapy response, and obtain individualized oxygen therapy optimization scheme;

[0064] The adversarial digital twin fuses adversarial learning and digital twin technology, maps the state of the physical entity (user, oxygen-enriched cabin) through a virtual model and simulation, and in the classical framework, introduces a physiological environment dynamic causal mining mechanism;

[0065] Specifically, the physiological environment dynamic causal mining mechanism uses a causal inference algorithm to sort out the causal relationship between user physiological parameters (such as heart rate, blood oxygen) and cabin environment (temperature, oxygen concentration) from multi-modal data, and then combines adversarial learning training to enable the digital twin model to not only replicate data distribution when simulating physiological state changes, but also follow the causal logic of real physiological-environment interaction, improving the reliability of health trend and oxygen therapy response prediction, and accurately simulating user physiological evolution under different oxygen therapy conditions. The process is as follows:

[0066] The structure of the high-fidelity physiological deduction model includes the causal adversarial digital twin improved by the physiological environment dynamic causal mining mechanism, the causal generator, and the causal discriminator;

[0067] The construction process of the causal adversarial digital twin is as follows:

[0068] First, build a basic adversarial digital twin, construct a user digital twin based on user basic information (age, gender, etc.), physiological parameter data, and construct a basic digital twin based on oxygen-enriched cabin design parameters and environmental data history records, replicating the initial state and basic characteristics of the entity;

[0069] Based on the adversarial digital twin, embed the causal correlation graph:

[0070] Based on the causal mining algorithm (PC algorithm), analyze historical multi-modal data and mine the causal relationship between environmental parameters and physiological parameters (such as oxygen concentration directly leading to an increase in blood oxygen saturation, and heart rate changes affecting blood pressure fluctuations), and encode these causal relationships into a graph structure and store them in the digital twin to obtain a causal adversarial digital twin;

[0071] The construction process of the causal correlation graph is as follows:

[0072] Select the constraint-based PC algorithm (Peter-Clark Algorithm) to mine causal connection relationships between variables from multi-modal fusion data through conditional independence testing (such as Pearson correlation testing and partial correlation testing), construct a causal graph structure, i.e. test the conditional independence of variable pair (X, Y) given the variable set Z, and if (X Y | Z), delete the direct edge between X and Y;

[0073] Collect multi-modal fusion data, and use all multi-modal fusion data as nodes (X, Y) to construct a fully connected undirected graph;

[0074] For each pair of nodes (X, Y), test whether they are independent given other nodes subset Z using partial correlation analysis, if independent, remove the edge between X and Y, if not, keep the edge and mark the condition set Z;

[0075] Set a time sliding window (window length set to 1 hour, step 10 minutes), for the data in each window, perform causal discovery and effect quantification;

[0076] Specifically, since the physiological-environmental relationship changes dynamically with the progress of oxygen therapy (for example, the effect of stress on heart rate may change after the user adapts to oxygen therapy), the sliding window can capture this dynamic nature;

[0077] After mining the local causal graph for each window, it is fused with the global causal graph (initially constructed from historical full data), and the graph merging algorithm is used to retain stable causal edges and update dynamically changing causal edges (such as the effect strength of stress → heart rate decaying over time), forming a dynamically evolving causal correlation graph G(t);

[0078] The causal generator construction and training process is as follows:

[0079] Through a deep neural network containing an input layer, a hidden layer (convolutional layer + fully connected layer), and an output layer, the multi-modal fusion data is received, features are extracted, and simulated physiological parameter changes are obtained to obtain a basic generator;

[0080] A causal constraint block is added after the hidden layer of the basic generator. The causal constraint block calls the causal correlation graph of the digital twin when generating simulated physiological data, and forces the generation process to follow the causal logic to obtain a causal generator;

[0081] For example, if the causal graph shows that "sudden increase in stress will cause short-term increase in heart rate", the generator must embed this causal correlation when simulating heart rate data under stress change scenarios. The causal constraint block makes the output simulated physiological parameters not only conform to the data distribution, but also fit the real physiological causal mechanism, improving the realism of simulation;

[0082] Causal generator training:

[0083] Fix the discriminator parameters, input the generator simulation data into the discriminator, and the generator loss function also incorporates the causal consistency constraint, the formula is:

[0084]

[0085] Where, L G is the loss value of the generator (the smaller the better, the generator can deceive the discriminator and conform to the causal logic), represents the input noise z of the generator (subject to noise distribution P zG(z) represents the simulated data generated by the generator with noise z (such as simulated physiological parameters, environmental parameters), D(G(z) represents the probability that the discriminator judges G(z) to be real data, C(G(z)) represents the probability that the discriminator judges G(z) to conform to the causal logic, and λ is the causal consistency loss weight (set to 0.1-0.5, balancing data distribution and causal logic judgment);

[0086] The causal discriminator construction and training process is:

[0087] Based on a deep neural network, the real physiological data is distinguished from the simulated data generated by the generator, and the output data is the probability that the data is real, and the basic discriminator is constructed;

[0088] The causal consistency feature is added to the input of the discriminator, and the causal correlation feature (such as checking whether the data has a reasonable causal driving relationship according to the causal graph) is extracted from the real data and the simulated data, which is used as a basis for discrimination together with the traditional data distribution feature (such as statistical distribution, time sequence feature), so that the discriminator not only evaluates the authenticity of the data, but also judges whether it conforms to the physiological-environmental causal logic, and guides the generator to generate more reasonable simulated data to obtain the causal discriminator;

[0089] The causal consistency feature acquisition process is:

[0090] The continuously collected real-time multi-modal fusion data (i.e. physiological-environmental data) is cut into small pieces according to a fixed time window (such as 5 minutes), and each piece contains physiological parameters (such as heart rate, blood oxygen saturation) and environmental parameters (such as oxygen concentration, cabin pressure) at several time points, which facilitates the verification of the causal relationship in the local time period;

[0091] The difference between the first and last (first and last data points) data points in the slice (such as the difference between the first and last oxygen concentrations) is multiplied by the corresponding data change rate to obtain the actual causal effect feature b1;

[0092] According to prior knowledge, the starting change time of the environmental parameter in the multi-modal fusion data (such as the oxygen concentration becoming high) is defined as the first change amount, and the starting change time of the physiological parameter response (such as the blood oxygen following the change) is defined as the second change amount, and the difference between the first change amount and the second change amount (the first change amount minus the second change amount) is obtained to obtain the causal time sequence feature b2;

[0093] If the first change amount minus the second change amount is negative (physiological change first, environmental change second, which is unreasonable), it is directly marked as "time sequence contradiction", and the feature value is set to 0, and the first change amount minus the second change amount is positive (environmental change first, physiological change second, which is within a reasonable range), and the feature value is set to 1, indicating that the causal time sequence is reasonable;

[0094] The actual causal effect feature and the causal timing feature are spliced together to form a causal consistency feature vector B=(b1, b2);

[0095] Causal discriminator training:

[0096] The real data and the simulation data are respectively input into the discriminator. The discriminator extracts the distribution features of the traditional multi-modal fusion data on one hand, and extracts the causal consistency features (according to the causal graph, checking whether the causal correlation in the data is reasonable, such as whether the oxygen concentration change causes the blood oxygen change in a reasonable timing and amplitude) on the other hand. The causal discriminator is trained based on the traditional cross-entropy and the causal consistency loss, and the formula is represented as:

[0097]

[0098] Wherein, x is the input data, C(x) is the causal consistency discrimination function (the probability that the output data conforms to the causal logic), λ is the causal consistency loss weight (set to 0.1-0.5 to balance the data distribution and the causal logic discrimination), P data is the real data distribution, P z is the generator input noise distribution;

[0099] D(x) is the data distribution discrimination probability, G(z) is the generator simulation data, D(x) is the probability that the discriminator judges that the real data x is true data (the closer to 1, the more it is considered to be true);

[0100] C(G(z) is the probability that the discriminator judges that the simulation data G(z) conforms to the causal logic (the generator hopes that this value is high, and the discriminator hopes that it is low);

[0101] D(G(z) represents the probability that the discriminator judges that the simulation data G(z) is true data (the generator hopes that this value is high, and the discriminator hopes that it is low);

[0102] The causal discriminator updates the discriminator parameters according to the extended loss function, so that it can distinguish the authenticity of the data distribution and identify the rationality of the causal logic, and improve the discrimination ability;

[0103] From the construction of the twin to the causal enhancement design of the generator and the discriminator, the causal logic constraint is integrated into the whole process, which solves the problem of generating heavy data distribution and light physiological mechanism, and finally constructs a high-fidelity physiological deduction model;

[0104] With the constructed high-fidelity physiological deduction model, input the current physiological parameters (such as heart rate, blood oxygen saturation) of the user, the environmental parameters (such as the current oxygen concentration, pressure) of the oxygen-enriched cabin and the basic information (age, gender, etc.), the model is based on the causal correlation graph and the trained causal generator and causal discriminator to deduce the dynamic changes (such as how the blood oxygen changes with the adjustment of the oxygen concentration, the response of the heart rate to the change of the pressure) of the physiological state of the user under different oxygen therapy conditions (adjusting the oxygen concentration, pressure, treatment time and other parameters), and predict the health trend and the oxygen therapy response.

[0105] According to the above combination results, a personalized oxygen therapy optimization scheme is generated, such as:

[0106] If the blood oxygen saturation of the user is low and the health goal is to quickly improve the hypoxic state, the model deduces that "the blood oxygen is more efficiently improved when the oxygen concentration is 22%-24%", and recommends the oxygen concentration interval and matches the appropriate treatment time (such as initial 30 minutes of observation response);

[0107] If it is an elderly user, according to the causal logic of "the change rate of pressure affects comfort" in the model, a lower pressure change rate (such as a pressure adjustment gradient ≤0.1kPa / minute) is recommended, and the oxygen concentration and treatment time are adjusted to ensure comfort and efficacy.

[0108] Intelligent control module: based on the real-time adjustment of the pressure, oxygen concentration and treatment time of the micro-high pressure oxygen cabin according to the personalized oxygen therapy optimization scheme, intelligent control is realized;

[0109] After the scheme is generated, the personalized oxygen therapy scheme (such as the recommended oxygen concentration range (O 2-range =[22%, 24%], the pressure change rate P rate ≤0.1kPa / minute, and the target treatment time T target =60 minutes) output by the high-fidelity model is disassembled into a control parameter set, which includes:

[0110] Static parameters: target oxygen concentration interval, pressure safety threshold;

[0111] Dynamic parameters: pressure adjustment gradient, oxygen concentration adjustment step, time trigger condition;

[0112] The parameter set is imported into the micro-high pressure oxygen cabin control system to complete the initialization configuration of the control logic and the execution mechanism (such as the oxygen concentration adjusting valve, the pressure adjusting valve and the timing module).

[0113] Embodiment two

[0114] As shown in Figure 2 , the present application proposes an integrated intelligent micro-high pressure oxygen-enriched cabin control method, which includes:

[0115] S1: Collect real-time physiological parameters of users and cabin environment data to obtain a multi-modal fusion data set;

[0116] S2: Based on the multi-modal fusion data, a high-fidelity physiological deduction model is constructed by using an adversarial digital twin combined with a physiological environment dynamic causal mining mechanism to predict user health trends and oxygen therapy responses, and obtain an individualized oxygen therapy optimization scheme;

[0117] S3: Based on the individualized oxygen therapy optimization scheme, the pressure, oxygen concentration and treatment duration of the micro-hyperbaric oxygen chamber are adjusted in real time to achieve intelligent control.

[0118] In the application, several formulas involved are dimensionless values for numerical calculation, and the establishment of the formula is obtained by software simulation of a large number of collected data to obtain a formula closest to the real situation. Some coefficients or weights in the formula are set by the person skilled in the art according to the actual situation, so this will not be described here.

[0119] The above embodiments can be realized wholly or partially by software, hardware, firmware or any other combination. When realized by software, the above embodiments can be realized wholly or partially in the form of a computer program product. Those skilled in the art can realize that the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solutions.

[0120] Although the embodiments of the present application have been shown and described, those skilled in the art can understand that various changes, modifications, replacements and variations can be made to the embodiments without departing from the principles and spirits of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. An integrated intelligent micro-high pressure oxygen-enriched cabin control system based on, characterized by, The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model.

2. The integrated intelligent micro-high pressure oxygen-enriched cabin control system according to claim 1, characterized in that, The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model.

3. The integrated intelligent micro-high pressure oxygen-enriched cabin control system according to claim 2, characterized in that, The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model.

4. The integrated intelligent micro-high pressure oxygen-enriched cabin control system according to claim 3, characterized in that, The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model.

5. The integrated intelligent micro-high pressure oxygen-enriched cabin control system according to claim 4, characterized in that, The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model.

6. The integrated intelligent micro-high pressure oxygen-enriched cabin control system according to claim 5, characterized in that, The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. wherein L G is the loss value of the generator, represents the input noise z of the generator, G(z) represents the simulated data generated by the generator using the noise z, D(G(z) represents the probability that the discriminator judges G(z) to be real data, C(G(z)) represents the probability that the discriminator judges G(z) to be consistent with the causal logic, and λ is the causal consistency loss weight.

7. The integrated intelligent micro-high pressure oxygen-enriched cabin control system according to claim 6, characterized in that, The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device based on a high-fidelity physiological deduction model. The application relates to a physiological parameter prediction method and device The basic discriminator is constructed based on a deep neural network, and a causal consistency feature is added to the input of the discriminator to obtain a causal discriminator; The real data and the simulation data are input into the discriminator, and a causal consistency loss is integrated into a traditional cross-entropy to train the causal discriminator, and the formula is as follows: where x is input data, C(x) is a causal consistency discrimination function, λ is a causal consistency loss weight, P data is a real data distribution, P z is a generator input noise distribution, D(x) is a data distribution discrimination probability, G(z) is a generator simulated data, D(x) is a probability that the discriminator judges that the real data x is true data, C(G(z) is a probability that the discriminator judges that the simulated data G(z) conforms to causal logic, and D(G(z) represents a probability that the discriminator judges that the simulated data G(z) is true data.

8. The integrated intelligent micro-high pressure oxygen-enriched cabin control system according to claim 7, characterized in that, The causal consistency feature acquisition process is as follows: The continuously collected real-time multi-modal fusion data is cut into small pieces according to a fixed time window, and the difference between the first and last data points in the slice and the corresponding data change rate are subjected to ratio operation to obtain an actual causal effect feature b1; According to prior knowledge, the starting change time of the environmental parameter in the multi-modal fusion data is defined as a first change amount, the starting change time of the physiological parameter response is defined as a second change amount, the difference between the first change amount and the second change amount is obtained by observing the data segment to obtain a causal timing feature b2; If the first change amount-second change amount is negative, the feature value is set to 0, and if the first change amount-second change amount is positive, the feature value is set to 1; The causal consistency feature vector B=(b1, b2) is obtained based on the actual causal effect feature and the causal timing feature.

9. A control method based on integrated intelligent micro-high pressure oxygen-enriched cabin, according to any one of claims 1-8, wherein, It comprises: S1: Collecting real-time physiological parameters and cabin environmental data of a user to obtain a multi-modal fusion data set; S2: Based on the multi-modal fusion data, a high-fidelity physiological deduction model is constructed by using an adversarial digital twin combined with a physiological environment dynamic causal mining mechanism to predict the user's health trend and oxygen therapy response, and an individualized oxygen therapy optimization scheme is obtained; S3: Based on the individualized oxygen therapy optimization scheme, the pressure, oxygen concentration and treatment time of the micro-hyperbaric oxygen chamber are adjusted in real time to realize intelligent control.

Images