Multi-dimensional data linkage regulation and control method and system for hypoxia treatment cabin

Abstract

The invention relates to the technical field of non-electrical variable control, in particular to a multi-dimensional data linkage regulation and control method and system for a hypoxia treatment cabin, and the method comprises the steps: collecting the monitoring data of the body of a user in real time when the hypoxia treatment cabin is used, recording the environment data when the abnormal monitoring data disappears after the monitoring data is abnormal, and storing the environment data in a database; the target value and the control parameter of the PID algorithm are updated; a plurality of correlative dimensions are screened from all the dimensions of the environmental data, and then the sampling frequency of the PID algorithm during regulation and control of each dimension in the environmental data is updated, so that the change trends of dimension values output by the PID algorithm under all the correlative dimensions are most relevant. According to the invention, through updating the PID algorithm, the environmental condition in the cabin is ensured to adapt to the body condition of the user, and frequent or continuous abnormal conditions of the user are avoided as far as possible.

Description

Technical Field

[0001] This invention relates to the field of non-electric variable control technology, specifically to a method and system for multi-dimensional data linkage regulation of a hypoxia treatment chamber. Background Technology

[0002] When using a hypoxia therapy chamber, users need to set appropriate environmental data, which includes multiple dimensions such as oxygen concentration, pressure, ambient temperature, humidity, and carbon dioxide concentration. Setting fixed values ​​may cause some users to experience adverse effects, leading to abnormal physical conditions. A common solution is to use a PID controller to dynamically adjust these parameters. However, different dimensions of environmental data have varying impacts on user health monitoring, especially the synergistic effects between different dimensions. Furthermore, individual differences exist among users. Directly using a PID algorithm to control each dimension according to pre-defined rules (e.g., setting different target values ​​for the PID algorithm at different times) cannot guarantee that the changes and synergies between different dimensions are appropriate for the user's physical condition. For example, when a user experiences abnormal physical conditions, it cannot be guaranteed that the changes and synergies between different dimensions will effectively alleviate or eliminate the abnormality. Summary of the Invention

[0003] To address the aforementioned problems, this invention provides a method and system for multi-dimensional data linkage control of a hypoxia treatment chamber.

[0004] The present invention provides a multi-dimensional data linkage control method and system for a hypoxia treatment chamber, which adopts the following technical solution:

[0005] One embodiment of the present invention provides a method for multi-dimensional data linkage control of a hypoxia treatment chamber, the method comprising the following steps:

[0006] When using the hypoxia therapy chamber, real-time monitoring data of the user's body is collected, and environmental data when the abnormal monitoring data disappears after an abnormality occurs is recorded as the reference environmental data of the abnormal monitoring data. Several dimensions of the environmental data include the oxygen concentration, pressure, ambient temperature, humidity, and carbon dioxide concentration of the hypoxia therapy chamber. The dimension value of each dimension of the environmental data is controlled by a PID algorithm.

[0007] The abnormal monitoring data that is generated again is recorded as the first data. At the same time, in the process of PID algorithm regulating environmental data, based on the correlation between the changing trend of each dimension in environmental data and the changing trend of monitoring data, several interrelated dimensions are selected from all dimensions of environmental data.

[0008] The reference environmental data of the anomaly monitoring data with the greatest similarity to the first data is denoted as the target environmental data; the target value of the PID algorithm is set as the target environmental data; the control parameters of the PID algorithm are updated based on the distribution differences between the reference environmental data of several anomaly monitoring data with the greatest similarity to the first data and the target environmental data; then the sampling frequency of the PID algorithm in regulating each dimension of the environmental data is updated so that the changing trend of the dimension value output by the PID algorithm under all interrelated dimensions is most relevant.

[0009] Preferably, the step of selecting several interrelated dimensions from all dimensions of the environmental data based on the correlation between the changing trends of each dimension in the environmental data and the changing trends of the monitoring data includes the following specific steps:

[0010] For each dimension of the environmental data and the correlation index with the monitoring data, the K-Means clustering algorithm is used to cluster all the correlation indexes of all dimensions into two categories. The category with the largest average value of the correlation index is obtained, and all dimensions in this category are recorded as the interrelated dimensions.

[0011] Preferably, the specific steps for updating the control parameters of the PID algorithm using reference environmental data from several anomaly monitoring data that are most similar to the first data and the distribution differences between the reference environmental data and the target environmental data are as follows:

[0012] Reference environmental data of several anomaly monitoring data most similar to the first data are recorded as candidate environmental data. For any dimension of the candidate environmental data, all dimension values ​​of all candidate environmental data in that dimension are recorded as first dimension values, and the dimension values ​​of the target environmental data in that dimension are recorded as target dimension values. All first dimension values ​​greater than the target dimension value are recorded as first-class dimension values, and all first dimension values ​​less than the target dimension value are recorded as second-class dimension values. The overshoot amplitude is obtained by using the first difference between the first-class dimension values ​​and the target dimension values, and the undershoot amplitude is obtained by using the second difference between the target dimension values ​​and the second-class values. The control parameters of the PID algorithm are updated using the overshoot amplitude and the undershoot amplitude.

[0013] Preferably, the sampling frequency of the updated PID algorithm in regulating each dimension of the environmental data is such that the changing trends of the dimensional values ​​output by the PID algorithm under all interrelated dimensions are most relevant. The specific steps include the following:

[0014] Simulate all dimension values ​​of the PID algorithm after the current time when it adjusts each dimension, and record them as the change sequence of each dimension. Then, divide the change sequence into several subsequences.

[0015] For any one of the interrelated dimensions, and for any one dimension and its correlation index with the monitoring data; the dimension with the largest correlation index among all interrelated dimensions is designated as the baseline dimension; any dimension other than the baseline dimension among all interrelated dimensions is designated as dimension A.

[0016] For any subsequence in dimension A, a subsequence with the highest correlation to the subsequence is extracted from the variation sequence of the baseline dimension, and denoted as the matching subsequence of any subsequence in dimension A; the value of the highest correlation is denoted as the matching degree of any subsequence in dimension A; the lag time of the subsequence relative to the matching subsequence is obtained; the average lag time of all subsequences with a matching degree greater than the threshold Q is denoted as the average lag time tA of dimension A; the sampling time of the last dimension value in the last subsequence of the variation sequence of the baseline dimension is denoted as y; the sampling frequency of the PID algorithm when adjusting the baseline dimension is denoted as F0; the sampling frequency of the PID algorithm when adjusting dimension A is updated as F1 = F0 × y / (y + tA).

[0017] Preferably, the specific steps for obtaining the threshold Q are as follows:

[0018] The standard deviation of the lag time of all subsequences in dimension A is denoted as the overall matching error of dimension A; for the linkage index between dimension A and the monitoring data, the threshold Q is positively correlated with the overall matching error and negatively correlated with the linkage index.

[0019] Preferably, the specific steps for obtaining the overshoot magnitude using the first difference between a type-one dimension value and a target dimension value, and obtaining the undershoot magnitude using the second difference between the target dimension value and a type-two dimension value, are as follows:

[0020] For each reference environment data and the first data, the similarity is defined as the similarity corresponding to all dimension values ​​in each reference environment data; the difference between each first-class dimension value and the target dimension value is obtained and denoted as the first difference; the similarity of the first-class dimension values ​​is normalized, and the weighted sum of these first differences is obtained using the normalized similarity to obtain the overshoot amplitude; the difference between the target dimension value and each second-class dimension value is obtained and denoted as the second difference; the similarity of the second-class dimension values ​​is normalized, and the weighted sum of the second differences is obtained using the normalized similarity to obtain the undershoot amplitude.

[0021] Preferably, the specific steps for updating the control parameters of the PID algorithm using overshoot and undershoot amplitudes are as follows:

[0022] The control parameters include a proportional coefficient and an integral coefficient. The updated proportional coefficient is positively correlated with the adjustment coefficient w, and the updated integral coefficient is negatively correlated with the adjustment coefficient w. Where q1 represents the overshoot magnitude, a2 represents the undershoot magnitude, and q0 represents the target dimension value; This refers to the sigmod function.

[0023] Preferably, the specific steps for the linkage indicator are as follows:

[0024] Within a preset time period prior to the current moment, obtain the abnormal feature sequence composed of abnormal features of the monitoring data of the same user at all times, and the dimension values ​​of each dimension of the environmental data at all times constitute the dimension sequence of each dimension.

[0025] The absolute value of the Pierre correlation coefficient between the dimensional sequence and the outlier feature sequence at each time step is denoted as the similarity of the changing trend of each dimension at each time step.

[0026] Within a preset time period, for the similarity of the change trends of all dimensions at the same moment, the several dimensions with the greatest similarity of change trends are marked as the relevant dimensions at each moment; for any dimension, the mean of the similarity of all change trends of that dimension at all moments marked as relevant dimensions is recorded as the linkage index of that dimension.

[0027] The method for obtaining the abnormal characteristics of the monitoring data includes:

[0028] The monitoring data consists of several indicators. The indicator data collected for each indicator within a preset time period is input into the LSTM neural network to obtain the abnormal evaluation value of each indicator in the monitoring data. The mean of the abnormal evaluation values ​​of all indicators in the monitoring data is recorded as the abnormal feature.

[0029] Preferably, the abnormal monitoring data includes the following specific steps:

[0030] The monitoring data consists of several indicators. Within a preset time period before the current moment, the indicator data collected for each indicator within the preset time period is input into the LSTM neural network to obtain the abnormal evaluation value of each indicator in the monitoring data. The average of the abnormal evaluation values ​​of all indicators in the monitoring data is recorded as the abnormal feature. The monitoring data with abnormal features greater than a first preset threshold is recorded as abnormal monitoring data.

[0031] Another embodiment of the present invention provides a multi-dimensional data linkage control system for a hypoxia therapy chamber, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the above-mentioned multi-dimensional data linkage control method for a hypoxia therapy chamber.

[0032] The beneficial effects of the technical solution of the present invention are:

[0033] First, this invention records the environmental data when the abnormal monitoring data disappears after an anomaly occurs, and this is recorded as the reference environmental data for the abnormal monitoring data. The reference environmental data of the abnormal monitoring data with the highest similarity to the first data is then recorded as the target environmental data. The target value of the PID algorithm is set as the target environmental data. This process updates the target value of the PID algorithm based on historical environmental data from when users' abnormal conditions subsided while using the hypoxia therapy chamber, initially enabling the PID to adapt the hypoxia therapy chamber environment to the user's physical condition through dynamic adjustment of the environmental data.

[0034] Furthermore, by utilizing reference environmental data from several anomaly monitoring data points with the highest similarity to the first data, and updating the control parameters of the PID algorithm based on the distribution differences with the target environmental data, this process updates the control parameters of the PID algorithm. This ensures that the environmental data of the hypoxia treatment chamber, when regulated by the PID algorithm, approaches and stabilizes at the target environmental data, while also taking into account reference environmental data under other abnormal conditions. This effectively alleviates the user's abnormal physical condition and avoids the problem of frequent or persistent abnormal physical conditions caused by only considering the anomaly monitoring data with the highest similarity, given the differences in physical conditions among different users.

[0035] Furthermore, this invention, based on the correlation between the changing trends of each dimension in the environmental data and the changing trends of the monitoring data, selects several interrelated dimensions from all dimensions of the environmental data. Then, it updates the sampling frequency of the PID algorithm when regulating each dimension in the environmental data, ensuring that the changing trends of the dimensional values ​​output by the PID algorithm under all interrelated dimensions are most relevant. This process considers that after updating the control parameters of the PID control algorithm under each dimension of the environmental data, there may be significant differences in the changes between the dimensional values ​​of different dimensions. Based on this, this invention first obtains the interrelated dimensions that can work together to effectively change the user's physical condition. Then, by changing the sampling frequency, it ensures that even after adjusting the control parameters, the PID regulates the interrelated dimensions as synchronously as possible, ensuring that the user's physical condition can still be changed through the synergy and linkage between the interrelated dimensions, thus further alleviating the user's abnormal physical condition.

[0036] In summary, this invention ensures that the environmental conditions inside the cabin are adapted to the user's physical condition by updating the target value, control parameters, and sampling frequency of the PID algorithm in interrelated dimensions, thereby minimizing the occurrence of frequent or continuous abnormal situations for the user. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 This is a flowchart illustrating the steps of a multi-dimensional data linkage control method for a hypoxia treatment chamber according to an embodiment of the present invention.

[0039] Figure 2 This is a simulation of the oxygen concentration before the PID algorithm control parameters are updated, as provided in one embodiment of the present invention.

[0040] Figure 3 This is a simulation of the oxygen concentration regulation process after updating the PID algorithm control parameters, as provided in one embodiment of the present invention.

[0041] Figure 4 This is an actual process for controlling oxygen concentration provided in one embodiment of the present invention;

[0042] Figure 5 This describes the actual temperature control process after the sampling frequency of the PID algorithm is updated, as provided in one embodiment of the present invention. Detailed Implementation

[0043] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a multi-dimensional data linkage control method and system for a hypoxia treatment chamber proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0044] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0045] The following description, in conjunction with the accompanying drawings, details the specific scheme of the multi-dimensional data linkage control method and system for a hypoxia treatment chamber provided by this invention.

[0046] Example 1:

[0047] Please see Figure 1 The diagram illustrates a flowchart of a multi-dimensional data linkage control method for a hypoxia treatment chamber according to an embodiment of the present invention. The method includes the following steps:

[0048] Step S101: When using the hypoxia therapy chamber, real-time monitoring data of the user's body is collected, and the environmental data of the hypoxia therapy chamber is controlled using a PID algorithm.

[0049] The hypoxia therapy chamber is equipped with a variety of sensors to monitor the user inside. These sensors include: wearable electrocardiograph, earlobe optical sensor, chest and abdominal motion sensor, etc., which are used to monitor the user's electrocardiogram signals, blood oxygen saturation (SpO2), respiratory rate (RR) and other data.

[0050] In this embodiment, the sensor collects data every 0.1 seconds, and records the ECG signal, blood oxygen saturation (SpO2), respiratory rate (RR), and other indicators collected by the same user as the user's monitoring data.

[0051] Simultaneously, before using the hypoxic therapy chamber, the oxygen concentration, pressure, ambient temperature, humidity, and carbon dioxide concentration inside the chamber need to be set to appropriate values. In this embodiment, the oxygen concentration in the hypoxic therapy chamber is set to 15%, the pressure to 110 kPa, the ambient temperature to 24°C, the humidity to 55%, and the carbon dioxide concentration to 0. This embodiment records the data of the five dimensions—oxygen concentration, pressure, ambient temperature, humidity, and carbon dioxide concentration—as environmental data, and the environmental data set above is recorded as the initial environmental data. Other embodiments can add more dimensions, such as light intensity, ozone content, and airflow velocity.

[0052] It should be noted that the hypoxic therapy chamber uses PSA (Pressure Swing Adsorption) technology to separate oxygen or nitrogen from the air, generating high-purity gases (such as 95% O2 or 99% N2). A two-position three-way solenoid valve is then used to switch between oxygen-enriched (oxygen-increasing) and nitrogen-enriched (oxygen-decreasing) gas inputs, thereby setting the oxygen concentration. The pressure is set by controlling the intake rate (air / O2 / N2 mixture) using a solenoid valve. Additionally, a centrifugal fan draws the oxygen from the hypoxic therapy chamber into a molecular sieve adsorption tank, where 13X zeolite material adsorbs CO2, thus altering the CO2 concentration.

[0053] To ensure that the environmental data in the hypoxic therapy chamber remains stable at the initial level during use, this embodiment uses a PID algorithm to regulate the environmental data. Specifically, the hypoxic therapy chamber is equipped with oxygen concentration sensors, pressure sensors, temperature sensors, humidity sensors, and carbon dioxide sensors to monitor the values ​​of oxygen concentration, pressure, ambient temperature, humidity, and carbon dioxide concentration in real time (i.e., the dimensional values ​​of each dimension in the environmental data). Then, five PID controllers are used to regulate each dimension of the environmental data based on the real-time monitored dimensional values ​​(for example, using PID controllers to regulate the opening ratio of the solenoid valves for oxygen and nitrogen, the air intake rate of the solenoid valves, and the speed of the centrifugal fan, etc.), ultimately ensuring that the environmental data remains stable at the initial level.

[0054] It should be noted that in this embodiment, only the dimension value of each dimension in the environmental data changes over time; the number and types of dimensions in the environmental data remain constant.

[0055] Each PID controller implements a PID algorithm, meaning that each dimension of the environmental data is independently controlled by a PID algorithm to adjust the dimension value, thereby making the cabin environmental data approach and dynamically stabilize at the initial environmental data (specifically, dynamically stabilize at each dimension value of the initial environmental data).

[0056] It should be noted that the setting of environmental data (i.e., oxygen concentration, pressure, ambient temperature, humidity, and carbon dioxide concentration) of the hypoxia therapy chamber, the use of PID algorithms to regulate environmental data, and the tuning methods of PID algorithm control parameters (such as proportional coefficient, integral coefficient, and derivative coefficient) are all well-known technologies and will not be described in detail in this embodiment. For example, MATLAB is used to tune the control parameters in this embodiment.

[0057] It should be noted that in this embodiment, environmental data is collected every 0.1 seconds. Furthermore, the sampling frequency of the PID algorithm when adjusting the dimensional values ​​of each dimension in the environmental data is 0.5, meaning that the dimensional value of each dimension is adjusted 0.5 times per second (i.e., once every two seconds). After multiple adjustments to the dimensional values ​​of each dimension, the PID algorithm stabilizes the environmental data of the hypoxia treatment chamber at the initial environmental data (this process is the principle of the PID algorithm, which will not be elaborated in detail in this embodiment). In some embodiments, the sampling frequency can be set to other values; this embodiment does not impose specific limitations. For example, in some embodiments, to save computational load or energy consumption (or to improve the endurance of the hypoxia treatment chamber), the sampling frequency is set to 0.25 (i.e., once every four seconds for each dimension).

[0058] Step S102: Record the environmental data when the abnormal monitoring data disappears after the monitoring data generates an anomaly, and record it as the reference environmental data of the abnormal monitoring data.

[0059] This embodiment takes into account that due to individual differences among users, the initial environmental data within the hypoxia treatment chamber may not be suitable for all users, which could lead to anomalies in the user's health monitoring data. This embodiment uses a 1-second interval as a time period; at each time period, anomaly detection is performed on the monitoring data for each user, acquiring the abnormal characteristics of the monitoring data at each time period, and determining whether the monitoring data at each time period is abnormal based on the abnormal characteristics. If abnormal monitoring data is found, it is recorded as abnormal monitoring data. The presence of abnormal monitoring data indicates that the user's health condition is abnormal.

[0060] For the same user during the historical use of the hypoxia therapy chamber, whenever the user's monitoring data becomes abnormal (i.e., after abnormal monitoring data is generated), and the monitoring data no longer becomes abnormal (i.e., when the abnormal monitoring data disappears), the environmental data at that moment is recorded. This environmental data is recorded as the reference environmental data for the abnormal monitoring data. The abnormal monitoring data and its corresponding reference environmental data are stored in a database (e.g., a MySQL database).

[0061] Reference environmental data refers to valuable environmental data that can prevent the user from experiencing abnormal physical conditions, serving as a reference for setting environmental data in the hypoxia treatment chamber should the user experience abnormal physical conditions again in the future.

[0062] As the hypoxia therapy chamber is used multiple times by different users, the database gradually records multiple reference environment data from different users.

[0063] Step S103: The abnormal monitoring data that is generated again is recorded as the first data. At the same time, in the process of controlling the environmental data by the PID algorithm, based on the correlation between the changing trend of each dimension in the environmental data and the changing trend of the monitoring data, several interrelated dimensions are selected from all dimensions of the environmental data.

[0064] At any given moment, if any user's monitoring data becomes abnormal, this newly generated abnormal monitoring data is recorded as the first data point. At this point, it is necessary to change the environmental data within the hypoxic therapy chamber. Specifically, since this embodiment uses a PID algorithm to adjust the environmental data in real time, keeping the environmental data dynamically stable (e.g., dynamically stable at the initial environmental data), when the environmental data within the hypoxic therapy chamber remains consistently stable, there may be situations where the user's body is not adapted to the chamber's environment, leading to abnormal user health monitoring data. In this case, it is necessary to adjust the environmental data within the hypoxic therapy chamber to stabilize it at other environmental data levels, rather than always stabilizing at the same environmental data level, thereby eliminating the user's abnormal condition. By dynamically adjusting the environmental data within the hypoxic therapy chamber, it is adapted to the user's physical condition.

[0065] Before altering the environmental data within the hypoxic therapy chamber, this implementation considered that changes in user health monitoring data and the generation of abnormal monitoring data are not only related to the user themselves but also affected by changes in environmental data. Environmental data contains multiple dimensions, and changes in monitoring data and the existence of abnormal monitoring data may be due to the correlation and synergy between these multiple dimensions (for example, the correlation and synergy between oxygen concentration, pressure, and carbon dioxide concentration can affect respiratory rate).

[0066] Based on this, in the process of controlling environmental data using the PID algorithm, this embodiment selects several interrelated dimensions from all dimensions of the environmental data according to the difference between the changing trends of each dimension in the environmental data and the changing trends of the monitoring data. Adjusting or changing the dimensional values ​​between these interrelated dimensions may cause changes in the monitoring data and the generation or disappearance of abnormal monitoring data.

[0067] As an example, in the process of PID algorithm control of environmental data, based on the correlation between the changing trends of each dimension of the environmental data and the changing trends of the monitoring data, several interrelated dimensions are selected from all dimensions of the environmental data. The method includes:

[0068] For a preset time period prior to the current moment (e.g., the previous minute including the current moment), acquire the abnormal characteristics of the same user at each moment. The abnormal characteristics at all moments constitute an abnormal characteristic sequence. Perform linear normalization on the abnormal characteristic sequence to remove the dimensions and orders of magnitude of the abnormal characteristic sequence.

[0069] It should be noted that within the preset time, the environmental data inside the hypoxic therapy chamber is regulated by a PID control algorithm. During the PID control process, the environmental data changes dynamically in real time (e.g., the environmental data dynamically approaches and stabilizes at the initial environmental data). In particular, considering that users consume oxygen and produce carbon dioxide in the hypoxic therapy chamber, and that users also have a significant impact on the temperature and humidity inside the hypoxic therapy chamber, the dynamic changes in environmental data during the PID control process are quite significant.

[0070] Within a preset time period, the dimensional values ​​of each dimension of the environmental data at all times constitute a dimensional sequence for each dimension. Each dimensional sequence is then linearly normalized to remove its dimensions and order of magnitude.

[0071] For any moment within a preset time period, construct a window of length 11 centered on that moment. The dimension values ​​in the dimension sequence of each dimension that are within the window are denoted as L1, and the abnormal features in the abnormal feature sequence that are within the window are denoted as L2. The absolute value of the Pierce correlation coefficient between L1 and L2 is denoted as the similarity of the change trend of each dimension at any moment.

[0072] For the similarity of the change trends of all dimensions at the same time, obtain several dimensions (e.g., 3) with the greatest similarity in change trends, and mark these dimensions as the relevant dimensions for each time.

[0073] In some embodiments, dimensions with a similarity in trend greater than a preset similarity trend threshold (e.g., 0.35) can be labeled as relevant dimensions at each time point.

[0074] For any given dimension, it is marked as a related dimension at certain times within a preset time period. At all times when it is marked as a related dimension, the mean of the similarity of all changing trends of the dimension is recorded as the linkage index between the dimension and the monitoring data.

[0075] For all dimensions of the environmental data, the correlation indicators for all dimensions were obtained. The K-Means clustering algorithm was used to cluster all the correlation indicators into two categories. Each category contains several dimensions and their correlation indicators. The category with the largest average correlation indicator was selected, and all dimensions within that category were denoted as the interrelated dimensions.

[0076] Specifically, in some embodiments, if no related dimensions exist at any time within a preset time period, the interrelated dimensions are not acquired. If the interrelated dimensions contain only one dimension, the interrelated dimensions are also not acquired. If no interrelated dimensions are acquired, the subsequent step S106 of this embodiment is not executed.

[0077] Step S104: The reference environmental data of the anomaly monitoring data most similar to the first data is denoted as the target environmental data; the target value of the PID algorithm is set as the target environmental data.

[0078] The similarity between each anomaly monitoring data point stored in the database (representing historical anomaly monitoring data) and the first data point is calculated. A higher similarity indicates that the user's physical abnormality described by the anomaly monitoring data point is more similar to the user's physical abnormality described by the first data point.

[0079] In the database, the reference environmental data of the anomaly monitoring data that is most similar to the first data (i.e., the most similar) is obtained and denoted as the target environmental data. The target environmental data can be used as the environmental data in the hypoxia therapy chamber at subsequent times. In other words, when using the PID algorithm to adjust the environmental data in the chamber in real time, it is necessary to make the environmental data approach and dynamically stabilize at the target environmental data in order to ensure that the user's abnormal condition is alleviated under the target environmental data, or to ensure that the environmental data in the hypoxia therapy chamber (i.e., the environmental data that approaches and dynamically stabilizes at the target environmental data) can adapt to the user's physical condition.

[0080] In order to dynamically stabilize the environmental data in the hypoxia treatment chamber at the target environmental data, this embodiment sets the target value of the PID algorithm to the target environmental data.

[0081] The target value of a PID algorithm refers to the value that, in PID control, is adjusted to approach and stabilize near a certain value. This target value is defined in the PID control algorithm. In this embodiment, the environmental data has multiple dimensions; therefore, the target value that each dimension aims to approach and stabilize using PID control is the corresponding dimension value in the target environmental data.

[0082] It should be noted that there may be multiple hypoxia treatment chambers in this embodiment, but only one user can use each chamber at a time to isolate different users or protect user privacy. In some embodiments, if there are multiple users in a hypoxia treatment chamber, the above process calculates a target environmental data for each user. In this case, the average of the target environmental data calculated for all users is used as the target value of the PID algorithm.

[0083] Step S105: Update the control parameters of the PID algorithm by using the distribution differences between the reference environmental data and the target environmental data of several abnormal monitoring data that are most similar to the first data.

[0084] This step takes into account that due to individual differences among users, and even the same user may have different physical conditions at different times during the use of the hypoxia therapy chamber, even if the target value of the PID algorithm is set to the target environmental data most similar to the abnormal situation, it may not be able to fully adapt to the user's physical condition. This is especially true when the target environmental data has not yet stabilized at the target environmental data under the control of the PID algorithm, or when the target environmental data dynamically stabilizes around the target environmental data with large fluctuations.

[0085] Therefore, this embodiment needs to take into account reference environmental data under other abnormal conditions stored in the database. Specifically, this embodiment uses the distribution differences between the reference environmental data of several abnormal monitoring data most similar to the first data and the target environmental data to update the control parameters of the PID algorithm. This ensures that when the environmental data of the hypoxia treatment chamber is regulated by the PID algorithm, it approaches and stabilizes at the target environmental data, while also taking into account reference environmental data under other abnormal conditions (i.e., reference environmental data of several abnormal monitoring data most similar to the first data), thus effectively alleviating the user's abnormal physical condition.

[0086] As an example, updating the control parameters of the PID algorithm using the distribution differences between reference environmental data and target environmental data from several anomaly monitoring data most similar to the first data includes the following methods:

[0087] Reference environmental data from several anomaly monitoring data points (e.g., 20) that are most similar to the first data point are recorded as candidate environmental data. Specifically, if there are fewer than 20 candidate environmental data points, this step and subsequent steps will not be performed.

[0088] In some embodiments, reference environment data of all anomaly monitoring data that have a similarity greater than a preset similarity threshold (e.g., 0.6) to the first data can be recorded as candidate environment data. Specifically, if there are insufficient candidate environment data (e.g., less than 20), this step and subsequent steps will not be performed.

[0089] For any given dimension, obtain all dimension values ​​of all candidate environment data in that dimension, denoted as the first dimension value, and obtain the dimension value of the target environment data in that dimension, denoted as the target dimension value.

[0090] It should be noted that since each candidate environment data corresponds to a similarity with the first data, each first dimension value also corresponds to a similarity.

[0091] All first dimension values ​​greater than the target dimension value are denoted as a class dimension value. The difference between each class dimension value and the target dimension value is recorded as the first difference. The similarity of these class dimension values ​​is normalized. In this implementation, the SoftMax formula is used for normalization. The normalized similarity is used to perform a weighted sum of these first differences to obtain the overshoot amplitude.

[0092] All first-dimensional values ​​less than the target dimension value are denoted as second-dimensional values. The difference between the target dimension value and each second-dimensional value is recorded as the second difference. The similarity of these second-dimensional values ​​is normalized. In this implementation, the SoftMax formula is used for normalization. The second difference is weighted and summed using the normalized similarity to obtain the under-adjustment magnitude.

[0093] The control parameters of the PID algorithm are updated using overshoot and undershoot amplitudes. When the overshoot amplitude is larger than the undershoot amplitude, the PID algorithm's response needs to be accelerated (i.e., the PID algorithm needs to quickly adjust the dimension value to the target dimension value), and significant overshoot or oscillation is allowed to ensure that the adjusted dimension traverses as many types of dimension values ​​as possible. When the overshoot amplitude is smaller than the undershoot amplitude, the PID algorithm's response needs to be suppressed (i.e., the PID algorithm needs to slowly adjust the dimension value to the target dimension value), and significant overshoot or oscillation is avoided to ensure that the adjusted dimension traverses as many types of dimension values ​​as possible. This process, by updating the control parameters of the PID algorithm, ensures that the environmental data of the hypoxia treatment chamber, when adjusted by the PID algorithm, approaches and stabilizes at the target environmental data, while also taking into account reference environmental data under other abnormal conditions (i.e., reference environmental data of several abnormal monitoring data most similar to the first data), so that the physical abnormalities of users with individual differences can be effectively alleviated.

[0094] As a preferred example, updating the control parameters of the PID algorithm using overshoot and undershoot includes the following methods:

[0095] The control parameters include proportional coefficient, integral coefficient, and derivative coefficient. In this embodiment, only the proportional coefficient and integral coefficient are updated. The proportional coefficient and integral coefficient before the update are denoted as kp and ki, and the proportional coefficient after the update is denoted as kp1 and ki1. kp1 = (1 + w) × f1 × kp, ki1 = (1 - w) × f2 × ki. Where w represents the adjustment coefficient, and f1 and f2 represent the preset scaling coefficients, respectively. In this embodiment, f1 = 1.5 and f2 = 1.67 are used as examples.

[0096] The formula for calculating the adjustment factor w is as follows: , where q1 represents the overshoot magnitude, a2 represents the undershoot magnitude; q0 represents the target dimension value, used to remove dimensions; This represents the sigmod function, used for normalization.

[0097] As an alternative example, updating the control parameters of the PID algorithm using overshoot and undershoot amplitudes includes the following methods:

[0098] When the overshoot is greater than or equal to the undershoot, increase kp by 40% and decrease ki by 10%. When the overshoot is less than the undershoot, do not update kp and ki.

[0099] In the above example, when the overshoot is larger than the undershoot, the updated proportional coefficient and integral coefficient are made to be greater than and less than the original proportional coefficient and integral coefficient, respectively; or, when the overshoot is smaller than the undershoot, the updated proportional coefficient and integral coefficient are made to be less than and greater than the original proportional coefficient and integral coefficient, respectively. This ensures that when the environmental data of the hypoxia treatment chamber is controlled by the PID algorithm, it approaches and stabilizes at the target environmental data, while also taking into account the reference environmental data under other abnormal conditions (i.e., the reference environmental data of several abnormal monitoring data that are most similar to the first data), so that the physical abnormalities of users with individual differences can be effectively alleviated.

[0100] It should be noted that in this embodiment, each dimension corresponds to a PID control algorithm, and the overshoot and undershoot obtained in each dimension are used to update the control parameters of the PID algorithm corresponding to that dimension.

[0101] Step S106: Update the sampling frequency of the PID algorithm in each dimension of the control environment data so that the changing trend of the dimension value output by the PID algorithm in all interrelated dimensions is most relevant.

[0102] After updating the control parameters of the PID control algorithm for each dimension in the environmental data, the changes in the dimensional values ​​of different dimensions may vary significantly. When the changes in the dimensional values ​​of interrelated dimensions differ greatly, it may be impossible to influence and change the user's physical condition (e.g., alleviate the user's abnormal physical condition at the current moment) through the synergy between different dimensions, which is not conducive to ensuring that the environmental conditions inside the cabin are adapted to the user's physical condition. This implementation solves this problem by updating the sampling frequency of the PID algorithm in each dimension of the control environmental data, so that the changing trends of the dimensional values ​​output by the PID algorithm in all interrelated dimensions are most similar, including:

[0103] Based on the updated control parameters, simulate all dimension values ​​of the PID algorithm when it regulates each dimension at the current time and in the subsequent time (e.g., within the next 2 minutes), and record them as the change sequence of each dimension.

[0104] This embodiment uses MATLAB for simulation. Other embodiments can use tools such as Simulink for simulation. All errors of the PID controller are ignored during simulation.

[0105] Given the control parameters of the PID algorithm and the target value (i.e., the target environment data), the specific simulation process or method is well known. This embodiment does not limit or elaborate on it, but only requires that the same simulation process or method be used throughout the implementation of this embodiment.

[0106] All dimensional values ​​in the change sequence for each dimension are linearly normalized to remove dimensions and orders of magnitude. All subsequent change sequences are normalized change sequences.

[0107] Divide the change sequence of each dimension of interest into several equal subsequences (e.g., 5). Note that if the subsequences are not perfectly equal, delete the last remaining subsequence. Additionally, obtain the standard deviation of all dimension values ​​in each subsequence. If the standard deviations of two adjacent subsequences are both small (e.g., less than 0.1), it indicates that these two subsequences, and subsequent subsequences, may lack a trend; in this case, delete these two subsequences, and all subsequent subsequences.

[0108] For all interrelated dimensions, each interrelated dimension corresponds to a linkage indicator (described in step S103). The dimension with the largest linkage indicator among all interrelated dimensions is denoted as the benchmark dimension.

[0109] The changes in the baseline dimension's value are most correlated with changes in user monitoring data. This embodiment updates the sampling frequency of the PID algorithm for each dimension in the control environment data, ensuring that the changing trends of the PID algorithm's output dimension values ​​in other interrelated dimensions are most correlated with the changing trend of the baseline dimension. Specifically, this includes:

[0110] For any dimension other than the baseline dimension among all interrelated dimensions, it is denoted as dimension A (e.g., the dimension corresponding to oxygen concentration).

[0111] For any subsequence in dimension A, the subsequence with the highest correlation to that subsequence is selected from the variation sequence of the baseline dimension. This subsequence is denoted as the matching subsequence of any subsequence in dimension A. The value of this maximum correlation is denoted as the matching degree of any subsequence in dimension A.

[0112] Obtain the first mean of the sampling times for all dimension values ​​in the subsequence, and the second mean of the sampling times for all dimension values ​​in the matching subsequence. The difference between the first mean and the second mean is denoted as the lag time of any subsequence in dimension A. The sampling time of the dimension value refers to the time taken by the PID algorithm to adjust the dimension value of each dimension.

[0113] The standard deviation of the lag time of all subsequences in dimension A is denoted as the overall matching error of dimension A.

[0114] The matching degree of all subsequences in dimension A is linearly normalized. For the normalized matching degree, all subsequences with a matching degree greater than the threshold Q are obtained. The mean lag time of these subsequences is denoted as the average lag time tA of dimension A.

[0115] The sampling time of the last dimension value in the last subsequence of the change sequence of the baseline dimension is denoted as y. This embodiment requires changing the sampling frequency of dimension A so that the PID algorithm synchronously adjusts the dimension values ​​of dimension A and the baseline dimension at time y and time y+tA respectively, ensuring that the change trends of dimension A and the baseline dimension are as relevant as possible between the current time and time y. This requires satisfying F0×y=F1×(y+tA), or F1=F0×y / (y+tA); where F0 represents the sampling frequency of the PID corresponding to the baseline dimension, and F1 represents the sampling frequency of dimension A after the sampling frequency is changed.

[0116] When adjusting the dimension value of dimension A, the PID algorithm sets its sampling frequency to F1=F0×y / (y+tA), which can ensure that the dimension value of dimension A and the dimension value of the baseline dimension maintain the same or similar trend of change as much as possible. This allows the user's body monitoring data to be effectively changed under the linkage adjustment of dimension A and the baseline dimension (that is, effectively eliminating the current abnormal monitoring data).

[0117] It should be noted that when the PID algorithm adjusts the dimension values ​​of dimensions other than the baseline dimension in the interrelated dimensions, the sampling frequency of the other dimensions will be updated, resulting in inconsistent sampling frequencies of all interrelated dimensions. Consequently, the change sequences of all interrelated dimensions will have inconsistent lengths. In this case, this embodiment obtains the shortest change sequence and truncates the ends of the other change sequences, so that the change sequences of all interrelated dimensions are equal to the length of the shortest change sequence.

[0118] Furthermore, the threshold Q is obtained from the overall matching error of dimension A and the linkage index of dimension A. In this embodiment, the threshold Q is positively correlated with the overall matching error and negatively correlated with the linkage index. Specifically, the smaller the overall matching error, the more uniformly the different local subsequences in the change sequence of dimension A can match the change sequence of the baseline dimension, or in other words, the change sequence of dimension A can maintain the same or similar change trend as the change sequence of the baseline dimension. In this case, more or even all subsequences need to be used to obtain the average lag time tA (corresponding to a smaller threshold Q). Furthermore, the larger the linkage index of dimension A, the more necessary it is to require the change sequence of dimension A to maintain the same or similar change trend as the change sequence of the baseline dimension to ensure that dimension A and the baseline dimension coordinate to change the user monitoring data (corresponding to a smaller threshold Q). For example, when the threshold Q is less than or equal to 0, it means that all subsequences are used to obtain the average lag time tA.

[0119] Relatively speaking, the larger the overall matching error, the more it indicates that the subsequences in different parts of the change sequence of dimension A cannot be uniformly matched with the change sequence of the benchmark dimension, or that the change sequence of dimension A cannot maintain the same or similar change trend as the change sequence of the benchmark dimension as a whole. If the linkage index of dimension A is small, it means that when dimension A and the benchmark dimension work together to change user monitoring data, the synergistic effect is not obvious. In this case, it is less necessary to require that the change sequence of dimension A cannot maintain the same or similar change trend as the change sequence of the benchmark dimension as a whole. In this case, it is only necessary to ensure that the subsequences with a higher degree of matching in the change sequence of dimension A maintain the same or similar change trend locally with the change sequence of the benchmark dimension (corresponding to the case where the threshold Q is larger).

[0120] As an example, the threshold Q is obtained from the overall matching error of dimension A and the linkage index of dimension A, and the calculation formula is: Q = w1 × H1 - w2 × H2; where H1 represents the overall matching error of dimension A and H2 represents the linkage index of dimension A. w1 and w2 represent preset hyperparameters, respectively. In this embodiment, w1 = 1.7 and w2 = 1.1. In other embodiments, w1 and w2 can be set to other values, and this embodiment does not impose specific limitations. Specifically, when there is no subsequence with a matching degree greater than the threshold Q, the lag time of the subsequence with the highest matching degree is recorded as the average lag time tA of dimension A.

[0121] As another example, the threshold Q is obtained from the overall matching error of dimension A and the linkage index of dimension A, including the following calculation formula: .

[0122] Therefore, in this embodiment, whenever any user generates abnormal monitoring data, the PID algorithm is updated using historical abnormal monitoring data (i.e., each abnormal monitoring data stored in the database). This involves updating the PID algorithm's target value, control parameters, and sampling frequency in the relevant dimensions (the sampling frequency in dimensions other than the relevant dimensions is not updated). Furthermore, by repeating this embodiment, the abnormal monitoring data and its reference environment data for users are continuously stored in the database. Simultaneously, the PID algorithm is continuously updated based on the abnormal monitoring data and its reference environment data in the database.

[0123] This concludes the example.

[0124] Example 2:

[0125] As an example, anomaly detection is performed on the monitoring data of each user at each time point to obtain the anomaly characteristics of the monitoring data at each time point. Based on these anomaly characteristics, it is determined whether the monitoring data at each time point is abnormal. The methods include:

[0126] For the electrocardiogram (ECG) signals in the monitoring data, all ECG signals collected within a certain period of time (e.g., within 1 minute) before each moment are collected to form an ECG signal sequence. This ECG signal sequence is then input into an LSTM neural network, and the LSTM neural network outputs the abnormality assessment value of the ECG signal sequence, which is called the abnormality assessment value of the ECG signal at each moment.

[0127] The training method for this LSTM neural network is as follows: ECG signals of each user are recorded in real time. All recorded ECG signals are divided into equal-length time series, each lasting one minute. Time series shorter than one minute are deleted. Cardiovascular professionals assess the abnormality level of each time series, with the abnormality level ranging from {0, 0.1, 0.2, ..., 0.9}. Each abnormality level represents a label; a higher abnormality level indicates poorer cardiovascular function as described by the time series. All time series and their corresponding labels (i.e., abnormality levels) are used as a dataset. Using this dataset, the LSTM neural network is trained using the mean squared error loss function and the stochastic gradient descent algorithm.

[0128] LSTM neural networks and their training methods are well known, and will not be described in detail in this embodiment.

[0129] In other embodiments, the method can be used to identify whether there is an abnormality in the electrocardiogram (ECG) signal sequence. When an abnormality exists, the abnormality assessment value of the ECG signal sequence is set to 1; when no abnormality exists, the abnormality assessment value of the ECG signal sequence is set to 0. The abnormal ECG signal identification method and device disclosed in CN111832537B can be used to identify whether there is an abnormality in the ECG signal sequence.

[0130] Furthermore, for oxygen saturation (SpO2) and respiratory rate (RR) in the monitoring data, oxygen saturation (SpO2) and respiratory rate (RR) sequences are obtained from the oxygen saturation (SpO2) and respiratory rate (RR) collected within a period of time before each time step (e.g., within 0.5 minutes), respectively. Two other LSTM networks are retrained, and the oxygen saturation sequence and respiratory rate sequence are input into the two LSTM networks respectively to obtain the abnormal assessment values ​​of oxygen saturation and respiratory rate at each time step.

[0131] The training methods for these two LSTM networks are the same as those for the LSTM networks mentioned above.

[0132] In other embodiments, the average oxygen saturation F collected over a period of time (e.g., within 0.5 minutes) prior to each moment is obtained. An upper limit F1 and a lower limit F2 for oxygen saturation are set (e.g., F1=92%, F2=88%). When F is less than F2, (F2-F) / F1 is recorded as an abnormal assessment value for oxygen saturation; when F is greater than F1, (F-F1) / F is recorded as an abnormal assessment value for oxygen saturation. When F is greater than or equal to F2 and less than or equal to F1, the abnormal assessment value for oxygen saturation is set to 0.

[0133] Obtain the mean respiratory rate P collected over a period of time prior to each moment (e.g., within 0.5 minutes). Set an upper limit P1 and a lower limit P2 for the respiratory rate (e.g., P1 = 20 breaths / min, P2 = 12 breaths / min). When P is less than P2, record (P2-P) / P1 as the abnormal assessment value of the respiratory rate; when P is greater than P1, record (P-P1) / P as the abnormal assessment value of the respiratory rate. When P is greater than or equal to P2 and less than or equal to P1, the abnormal assessment value of the respiratory rate is set to 0.

[0134] Furthermore, in this implementation, the average value of the abnormal assessment values ​​of the three indicators—electrocardiogram signal, blood oxygen saturation, and respiratory rate—is recorded as the abnormal characteristics of the monitoring data at each time point.

[0135] In other embodiments, the abnormal assessment values ​​of ECG signal, blood oxygen saturation and respiratory rate are weighted and summed using three weights (e.g. 0.5, 0.3, 0.2) to obtain the abnormal characteristics of the monitoring data at each time point.

[0136] When the abnormal characteristics of the monitoring data at each time point are greater than the first preset threshold th1, it is determined that the monitoring data at each time point is abnormal and the monitoring data is recorded as abnormal monitoring data. When the abnormal characteristics are less than or equal to the first preset threshold th1, it is determined that the monitoring data at each time point is not abnormal.

[0137] This embodiment uses th1=0.2 as an example for description. Other embodiments can be set to other values. This embodiment does not impose specific limitations.

[0138] In some embodiments, as described above, for the same user, the abnormal features obtained at all times are Gaussian filtered using a Gaussian filter kernel of length 7. The Gaussian filtered abnormal features are then used as the abnormal features of the monitoring data at each time point in these embodiments.

[0139] As an example, whenever a user's monitoring data becomes abnormal, and then the abnormality disappears, the environmental data at that moment is recorded. This environmental data is recorded as the reference environmental data for the abnormal monitoring data, specifically including:

[0140] Acquire all moments when monitoring data anomalies occur. Several consecutive moments among these moments are recorded as a group of times. For example, the first, second, and third moments are consecutively distributed, as are the fifth and sixth moments, but the third and fifth moments are not consecutively distributed.

[0141] The first time point in each time group is denoted as t1, and the last time point in the same time group is denoted as t2. t1 and t2 represent: after the monitoring data at time t1 becomes abnormal, the monitoring data no longer becomes abnormal at time t2; the environmental data at time t2 is recorded as the reference environmental data for the abnormal monitoring data at time t1.

[0142] As an example, the method for obtaining the similarity between each anomaly monitoring data (representing historical anomaly monitoring data) stored in the database and the first data includes:

[0143] For any monitoring data in the database and the first data, the abnormal assessment values ​​of the three indicators of electrical signal, blood oxygen saturation and respiratory rate in the monitoring data center constitute a three-dimensional vector, which is regarded as the monitoring data.

[0144] The cosine similarity between any anomaly monitoring data in the database and the first data is denoted as the similarity.

[0145] As an example, for any subsequence in dimension A, a subsequence with the highest correlation to that subsequence is extracted from the changing sequence of the baseline dimension. This subsequence is denoted as the matching subsequence of any subsequence in dimension A. The methods include:

[0146] Obtain the length x of the subsequence, and establish a window of length x. This window slides across the changing sequence of the baseline dimension with a step size of 1. After each slide (including before the first slide), obtain the sequence consisting of all dimension values ​​within the window, denoted as Lx. Calculate the absolute value of the Pierre Correlation Coefficient between the subsequence and Lx, denoted as the correlation between the subsequence and Lx. After the window has finished sliding, obtain the Lx with the highest correlation and use it as the matching subsequence.

[0147] Specifically, if the correlation between a matching subsequence and another subsequence is low (e.g., below 0.3), then that subsequence is deleted. Once all subsequences in several dimensions A have been deleted, the sampling frequency of dimension A is no longer updated.

[0148] In other examples, multiple (e.g., 3) Lx sequences with the highest relevance are obtained and used as candidate matching subsequences for that subsequence.

[0149] Similarly, all candidate matching subsequences for all subsequences in dimension A are obtained. The KM matching algorithm is then used to match all subsequences with all candidate matching subsequences, resulting in several matching pairs. Each matching pair contains one subsequence and one candidate matching subsequence. The KM matching algorithm maximizes the sum of the correlations between the subsequences and candidate matching subsequences in all pairs. In this embodiment, the candidate matching subsequence in each matching pair is used as the matching subsequence of the subsequence in the matching pair.

[0150] This example can avoid errors in matching subsequences when a subsequence is highly correlated with multiple Lx sequences.

[0151] It should also be noted that this embodiment is applied to the same indication. Users with different indications need to implement this embodiment independently. The indications include hereditary mitochondrial diseases, multiple sclerosis, myocardial ischemia-reperfusion injury, etc.

[0152] Example 3:

[0153] This embodiment provides a multi-dimensional data linkage control system for a hypoxia therapy chamber. The system includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements steps S101 to S106 in the above embodiment. The memory and processor are installed inside the hypoxia therapy chamber.

[0154] Example 4:

[0155] like Figure 2 This illustrates a simulation result of the PID control process for the oxygen concentration dimension before the control parameters are updated in one embodiment described in step S105. The target oxygen concentration value at the current moment changes from 15% set in step S101 to 18% (i.e., the oxygen concentration dimension value in the target environmental data); the control parameters before the update, namely the proportional coefficient, integral coefficient, and derivative coefficient, are 0.8, 2.1, and 1.18, respectively.

[0156] like Figure 3 It shows the simulation results of the PID control process of the oxygen concentration dimension after the control parameters are updated at the current time in one embodiment described in step S105; wherein the updated control parameters, namely the proportional coefficient, integral coefficient and derivative coefficient, are 2.04, 1.05 and 1.18. Figure 3 compared to Figure 2 The response is faster (stabilizing to the target value more quickly) and there is significant overshoot.

[0157] like Figure 4 It shows the curve of the oxygen concentration dimension change during PID control in actual use of the hypoxic therapy chamber, extracted at the sampling frequency of oxygen concentration in step S106 (sampling frequency is 2, oxygen concentration is the baseline dimension). It should be noted that the standard deviation of the oxygen concentration sensor error in this embodiment is 0.2. Due to the existence of error, the oxygen concentration cannot be directly compared to... Figure 3 Unlike simulations that eventually stabilize at the target value, this one dynamically changes around the target value. Figure 5 As shown, this diagram illustrates the temperature dimension change curve during PID control in actual use of the hypoxic therapy chamber after updating the sampling frequency of the temperature dimension in step S106 (the sampling frequency is now 1.1). It should be noted that the standard deviation of the temperature sensor error in this embodiment is 0.1. Figure 4 , 5 The oxygen concentration and temperature shown are interrelated dimensions, with oxygen concentration serving as the baseline dimension, and the temperature dimension updated at a different sampling frequency. Figure 4 , 5 The curves showed a clear correlation.

[0158] for Figure 4 , 5 In this regard, more accurate oxygen concentration sensors and temperature sensors can be used to reduce errors (i.e., reduce the standard deviation of the errors), which can help reduce... Figure 4 , 5 The fluctuation range is relatively large.

[0159] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for multi-dimensional data linkage regulation of a hypoxic therapy chamber, characterized in that, The method comprises the following steps: The monitoring data of the user's body is collected in real time when using the hypoxic treatment cabin, and the environmental data when the abnormal monitoring data disappears after the abnormal monitoring data is generated is recorded as the reference environmental data of the abnormal monitoring data; the environmental data includes the oxygen concentration, pressure, environmental temperature, humidity, and carbon dioxide concentration of the hypoxic treatment cabin; the dimension value of each dimension of the environmental data is regulated by the PID algorithm; The current abnormal monitoring data generated again is recorded as first data, and in the process of regulating the environmental data by the PID algorithm, a plurality of interrelated dimensions are selected from all dimensions of the environmental data according to the correlation between the change trend of each dimension of the environmental data and the change trend of the monitoring data; The reference environmental data of the abnormal monitoring data with the greatest similarity to the first data is recorded as target environmental data; the target value of the PID algorithm is set as the target environmental data; the control parameters of the PID algorithm are updated according to the distribution difference between the target environmental data and the reference environmental data of the abnormal monitoring data with the greatest similarity to the first data; and then the sampling frequency of the PID algorithm when regulating each dimension of the environmental data is updated, so that the change trend of the dimension value output by the PID algorithm under all interrelated dimensions is most relevant.

2. The multi-dimensional data linkage regulation method for a hypoxic treatment chamber according to claim 1, wherein, The specific steps of selecting a plurality of interrelated dimensions from all dimensions of the environmental data according to the correlation between the change trend of each dimension of the environmental data and the change trend of the monitoring data comprise the following steps: For the linkage index of each dimension of the environmental data and the monitoring data; all dimensions of the linkage index are clustered into two categories by using a K-Means clustering algorithm, a category with the greatest average value of the linkage index is obtained, and all dimensions in the category are recorded as interrelated dimensions.

3. The multi-dimensional data linkage regulation method for a low-oxygen therapy cabin according to claim 1, characterized in that, The specific steps of updating the control parameters of the PID algorithm according to the distribution difference between the target environmental data and the reference environmental data of the abnormal monitoring data with the greatest similarity to the first data comprise the following steps: The reference environmental data of the abnormal monitoring data with the greatest similarity to the first data is recorded as candidate environmental data; for any one dimension of the candidate environmental data, all dimension values of the candidate environmental data in the dimension are recorded as first dimension values, and the dimension value of the target environmental data in the dimension is recorded as a target dimension value; all first dimension values greater than the target dimension value are recorded as a category of dimension values, and all first dimension values less than the target dimension value are recorded as a category of dimension values; an overshoot amplitude is obtained by using a first difference between the category of dimension values and the target dimension value, and an undershoot amplitude is obtained by using a second difference between the target dimension value and the category of dimension values; the control parameters of the PID algorithm are updated by using the overshoot amplitude and the undershoot amplitude.

4. The multi-dimensional data linkage regulation method for a hypoxic treatment chamber according to claim 1, wherein, The specific steps of updating the sampling frequency of the PID algorithm when regulating each dimension of the environmental data, so that the change trend of the dimension value output by the PID algorithm under all interrelated dimensions is most relevant, comprise the following steps: All dimension values of each dimension when the PID algorithm regulates each dimension after the current time are simulated and recorded as a change sequence of each dimension, and the change sequence is equally divided into a plurality of sub-sequences. For any one of all the interrelated dimensions, and for any one dimension and the linkage index of the monitoring data; the dimension with the largest linkage index in all the interrelated dimensions is recorded as the reference dimension; for any one dimension other than the reference dimension in all the interrelated dimensions, it is recorded as dimension A; For any one sub-sequence in dimension A, a sub-sequence in the change sequence of the reference dimension is intercepted when the correlation of the sub-sequence is the largest, and it is recorded as the matching sub-sequence of any one sub-sequence in dimension A; the maximum value of the correlation is recorded as the matching degree of any one sub-sequence in dimension A; The lag time of the sub-sequence relative to the matching sub-sequence is obtained; the average lag time of all sub-sequences with a matching degree greater than a threshold Q is recorded as the average lag time tA of dimension A; the sampling time of the last dimension value in the last sub-sequence in the change sequence of the reference dimension is recorded as y, and the sampling frequency of the PID algorithm when regulating the reference dimension is recorded as F0, and the sampling frequency of the PID algorithm when regulating dimension A is updated as F1=F0×y / (y+tA).

5. The multi-dimensional data linkage regulation method for a hypoxic treatment chamber according to claim 4, wherein, The specific steps for obtaining the threshold Q are as follows: The standard deviation of the lag time of all sub-sequences in dimension A is recorded as the overall matching error of dimension A; for the linkage index of dimension A and the monitoring data; the threshold Q is positively correlated with the overall matching error and negatively correlated with the linkage index.

6. The multi-dimensional data linkage regulation method for a low-oxygen therapy cabin according to claim 3, characterized in that, The specific steps for obtaining the overshoot amplitude using the first difference between a class of dimension values and the target dimension value and obtaining the undershoot amplitude using the second difference between the target dimension value and a second class of dimension values are as follows: For the similarity of each reference environment data to the first data, the similarity is used as the similarity of all dimension values in each reference environment data; the difference between each first class of dimension values and the target dimension value is obtained and recorded as the first difference; The similarity of the first class of dimension values is normalized, and the normalized similarity is used to weight and sum the first differences to obtain the overshoot amplitude; The difference between the target dimension value and each second class of dimension values is obtained and recorded as the second difference; the similarity of the second class of dimension values is normalized, and the normalized similarity is used to weight and sum the second differences to obtain the undershoot amplitude. The specific steps for updating the control parameters of the PID algorithm using the overshoot amplitude and the undershoot amplitude according to claim 1 are as follows: The control parameters include a proportional coefficient and an integral coefficient, the updated proportional coefficient is positively correlated with the adjustment coefficient w, the updated integral coefficient is negatively correlated with the adjustment coefficient w, and the adjustment coefficient w, wherein q1 represents the overshoot amplitude, a2 represents the undershoot amplitude; q0 represents the target dimension value; represents the sigmod function. The specific steps for the linkage index according to claim 2, 4 or 5 are as follows: Within a preset time before the current time, an abnormal feature sequence composed of abnormal features of the monitoring data of the same user at all times is obtained, and the dimension values of each dimension of the environment data at all times form a dimension sequence of each dimension; The absolute value of the Pearson correlation coefficient between the dimension sequence of each dimension and the abnormal feature sequence at each time point is denoted as the change trend similarity of each dimension at each time point. In the preset time, for the change trend similarities of all dimensions at the same time point, a plurality of dimensions with the maximum change trend similarity are marked as relevant dimensions at each time point; for any dimension, the average of all change trend similarities of the dimension at all time points marked as relevant dimensions is denoted as the linkage index of the dimension. The method for obtaining the abnormal feature of the monitoring data comprises the following steps: The monitoring data is composed of a plurality of indexes, and the index data collected by each index in the preset time is input into an LSTM neural network to obtain the abnormal evaluation value of each index in the monitoring data, and the average of the abnormal evaluation values of all indexes in the monitoring data is denoted as the abnormal feature.

7. The multi-dimensional data linkage regulation method for a hypoxic treatment chamber according to claim 1, wherein, The specific steps of updating the control parameters of the PID algorithm by using the over-amplitude and under-amplitude comprise the following steps: The control parameter includes a proportional coefficient and an integral coefficient, the updated proportional coefficient is positively correlated with the adjustment coefficient w, the updated integral coefficient is negatively correlated with the adjustment coefficient w, and the adjustment coefficient w is determined according to the following formula: wherein q1 represents an overshoot amplitude, a2 represents an undershoot amplitude; q0 represents a target dimension value; and represents a sigmod function.

8. The multi-dimensional data linkage regulation method of a hypoxic treatment cabin according to claim 2, 4 or 5, characterized in that, The specific steps of the linkage index comprise the following steps: In the preset time before the current time, an abnormal feature sequence composed of the abnormal features of the monitoring data of the same user at all time points and a dimension sequence composed of the dimension values of each dimension of the environmental data at all time points are obtained; The absolute value of the Pearson correlation coefficient between the dimension sequence of each dimension and the abnormal feature sequence at each time point is denoted as the change trend similarity of each dimension at each time point. In the preset time, for the change trend similarities of all dimensions at the same time point, a plurality of dimensions with the maximum change trend similarity are marked as relevant dimensions at each time point; for any dimension, the average of all change trend similarities of the dimension at all time points marked as relevant dimensions is denoted as the linkage index of the dimension. The method for obtaining the abnormal feature of the monitoring data comprises the following steps: The monitoring data is composed of a plurality of indexes, and the index data collected by each index in the preset time is input into an LSTM neural network to obtain the abnormal evaluation value of each index in the monitoring data, and the average of the abnormal evaluation values of all indexes in the monitoring data is denoted as the abnormal feature.

9. The multi-dimensional data linkage regulation method for a low-oxygen therapy cabin according to claim 1, characterized in that, The specific steps of the abnormal monitoring data comprise the following steps: The monitoring data is composed of a plurality of indexes, and the index data collected by each index in the preset time is input into an LSTM neural network to obtain the abnormal evaluation value of each index in the monitoring data, and the average of the abnormal evaluation values of all indexes in the monitoring data is denoted as the abnormal feature.

10. A low-oxygen therapy cabin multidimensional data linkage regulation system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The processor implements the steps of the multi-dimensional data linkage regulation method of the hypoxic treatment cabin according to any one of claims 1-9 when executing the computer program.

Images