Oxygen production adjusting method for medical molecular sieve oxygen production equipment

Abstract

The invention discloses an oxygen production adjusting method for medical molecular sieve oxygen production equipment, which comprises the following steps: S1, collecting respiratory flow, pressure and electrocardiosignals, determining an inspiration starting point, and generating a respiratory phase prediction result; s2, applying perturbation excitation in a low-load stage, collecting acoustics and temperature rise signals, and generating a health fingerprint spectrum; s3, constructing a space-time belief network, and fusing the time confidence coefficient and the space confidence coefficient to form a dynamic confidence coefficient; s4, inputting a prediction result and a health map, and generating a scheduling instruction by a dynamic confidence coefficient controller; s5, performing long-term scheduling on the global layer, performing real-time adjustment on the local layer, and adjusting the weight of the interaction layer by using a confidence coefficient; and S6, executing the scheduling instruction, monitoring the oxygen concentration, and switching a safety mode and giving an alarm when the oxygen concentration is lower than a threshold value. Through multi-source prediction and confidence scheduling, accurate synchronization and intelligent oxygen supply of the molecular sieve oxygen generation equipment are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical gas equipment, and particularly relates to an oxygen production adjusting method for a medical molecular sieve oxygen production equipment. BACKGROUND

[0002] At present, the medical molecular sieve oxygen production equipment has been widely applied to clinical oxygen supply, and its basic working principle is to separate nitrogen and oxygen in air by using pressure swing adsorption technology, so as to output high-concentration oxygen. However, most of the existing molecular sieve oxygen production equipment adopts a fixed cycle mode and a valve control strategy, and lacks real-time adaptive ability to the dynamic characteristics of patient breathing. In the process of clinical use, the breathing frequency and breathing phase of patients have obvious individual differences and dynamic fluctuations, and the fixed oxygen supply rhythm is easy to be out of synchronization with the patient's inspiration time, which not only reduces the oxygen utilization efficiency, but also may cause the patient to fail to obtain sufficient oxygen flow at the key inspiration moment, thereby affecting the treatment effect.

[0003] On the other hand, as the core adsorption material of the oxygen production equipment, the health status of the molecular sieve directly determines the stability of the oxygen concentration and the service life of the equipment. The existing technology mainly depends on the running time or simple flow change to estimate the degradation of the molecular sieve, and it is difficult to diagnose the failure modes such as pulverization, channelization or wetting in time. This lagging detection method is easy to cause oxygen supply fluctuation or sudden failure of the equipment in clinical application, increasing the risk of patient use. The uncontrollable service life of the molecular sieve also increases the maintenance cost and replacement frequency, causing waste of medical resources.

[0004] The traditional scheduling method is mainly based on the deterministic control strategy, and lacks the modeling and adaptive ability to uncertain factors. There are noises and prediction errors in the patient breathing signal, and there are data fluctuations in the health diagnosis of the molecular sieve. The existing technology does not take this uncertainty into the scheduling optimization process, which is easy to cause the incoordination between the global scheduling and the local response, and further affects the stability and safety of the oxygen production system.

[0005] Therefore, how to provide an oxygen production adjusting method for a medical molecular sieve oxygen production equipment is a problem to be solved by those skilled in the art. SUMMARY

[0006] One object of the present application is to provide an oxygen production adjusting method for a medical molecular sieve oxygen production equipment. The present application makes full use of the respiratory phase prediction, the molecular sieve health fingerprint diagnosis and the hierarchical confidence space-time scheduling control technology, and details the precise locking of the respiratory phase and the oxygen production cycle, the molecular sieve health fingerprint spectrum diagnosis constructed by acoustics-thermal response, and the hierarchical scheduling optimization based on the dynamic confidence coefficient, so as to realize the intelligent adjustment of the oxygen supply process. The present application has the advantages of high synchronization between oxygen supply and patient breathing, good stability of oxygen concentration output, prolonged service life of molecular sieve, and improved safety and reliability of the overall system.

[0007] The method comprises the following steps: S1, collecting a patient's respiratory flow signal and pressure signal, extracting a breathing-related feature in combination with an electrocardio signal, determining a next inspiration starting point, and generating a breathing phase prediction result; S2, applying a perturbation excitation in a low load stage of a molecular sieve adsorption cycle, collecting a molecular sieve bed outer wall acoustic admittance spectrum and a micro-temperature rise signal, constructing an acoustic-thermal dynamic response dataset, generating a molecular sieve health fingerprint, and calculating a residual adsorption capacity factor and a bed health index; S3, constructing a space-time confidence network model, taking the uncertainty of the breathing phase prediction result as a time confidence input, taking the diagnostic uncertainty of the bed health index and the residual adsorption capacity factor in the molecular sieve health fingerprint as a space confidence input, forming a dynamic confidence coefficient, and dynamically adjusting a scheduling optimization range; S4, inputting the breathing phase prediction result and the molecular sieve health fingerprint into an oxygen production regulation controller, determining a valve opening and closing sequence, an adsorption cycle phase slip amount, and a compressor speed scheduling parameter based on a hierarchical confidence space-time scheduling control method, and generating an oxygen production scheduling instruction; S5, in a global layer of the oxygen production regulation controller, executing a long-term operation mode scheduling according to the bed health index and the residual adsorption capacity factor in the molecular sieve health fingerprint, in a local layer, adjusting a cycle phase and a valve action in real time according to the breathing phase prediction result, and in an interactive layer, adjusting the scheduling weight of the global layer and the local layer according to the dynamic confidence coefficient; S6, executing the oxygen production scheduling instruction, monitoring an outlet oxygen concentration in real time, comparing the real-time concentration with a preset threshold, switching to a safe oxygen supply curve when the outlet oxygen concentration is lower than the threshold or an abnormal operating state is detected, maintaining stable oxygen supply, and outputting an alarm signal.

[0008] Optionally, the breathing phase prediction result is generated by extracting a flow reversal point, a pressure trough point, and a modulation component related to a breathing rhythm in an electrocardio signal in a breathing cycle, fusing the extracted time features, determining a time position of a next inspiration starting point, and outputting the determined time position as the breathing phase prediction result.

[0009] Optionally, S2 specifically comprises: S21, applying a sequentially coded perturbation sequence in a low load stage of a molecular sieve adsorption cycle, the sequentially coded perturbation sequence being composed of a combination of a plurality of frequencies and a plurality of amplitudes, and sequentially acting on an inlet valve within a determined start and end time, and synchronously collecting an acoustic original waveform and a temperature original sequence of a molecular sieve bed outer wall; S22, fixed band and fixed time window segmentation processing is performed on the acoustic original waveform and the temperature original sequence, an acoustic heat feature set of each frequency band is extracted, and a multi-dimensional response feature set is generated; S23, a molecular sieve health fingerprint is established according to the multi-dimensional response feature set in sequence of the perturbation sequence and the frequency band number, and a baseline fingerprint is obtained through full frequency scanning when the equipment is first operated or is reset for maintenance; S24, a residual adsorption capacity factor is calculated by using a heat-acoustic mutual verification capacity evaluation method: The acoustic response energy ratio in the molecular sieve health fingerprint is compared with the baseline fingerprint, and a comparison deviation is recorded; The time lag of the acoustic peak value and the temperature peak value is compared with the baseline fingerprint, and a lag deviation is recorded; The consistency of the peak value decay rate of the multi-frequency band is compared with the baseline fingerprint, and a consistency deviation is recorded; The comparison deviation, the lag deviation and the consistency deviation are aggregated according to a preset segmentation weighting rule, and a residual adsorption capacity factor is output; S25, a bed health index is calculated by using a degradation mode determination and risk aggregation method: The pulverization mode, the channeling mode and the wetting mode are identified on the molecular sieve health fingerprint according to a feature combination rule, and mode risk scores are obtained respectively; The time stability of the molecular sieve health fingerprint is checked under a plurality of continuous and sequential coded perturbation sequences, and a stability score is obtained; A temperature drift compensation score is obtained by performing temperature drift compensation on the temperature original sequence; The mode risk score, the stability score and the temperature drift compensation score are aggregated according to a preset segmentation weighting rule, and a bed health index is output.

[0010] Optionally, the feature combination rule refers to: When the acoustic resonance peak energy ratio continuously decreases, the temperature rise amplitude abnormally fluctuates, and the acoustic peak value decay rate exceeds the stable range, the pulverization mode is determined; When the acoustic resonance peak position shifts to the high frequency end, the acoustic response energy ratio increases, and the time lag between the acoustic peak value and the temperature peak value shortens, in combination with the fact that the bed health index decreases at a rate exceeding twice the baseline average degradation rate, the channeling mode is determined; When the temperature rise amplitude continuously decreases, the time lag between the acoustic peak value and the temperature peak value increases, and the residual adsorption capacity factor decreases at an accelerated rate, the wetting mode is determined.

[0011] Optionally, the S3 specifically comprises: S31, input and alignment window of the space-time confidence network model are established, the breathing phase prediction result and the corresponding actual inspiration starting time mark in the continuous control period are obtained, and the molecular sieve health fingerprint and the baseline fingerprint in the continuous control period are obtained; S32, the space-time confidence network model is constructed, and the space-time confidence network model is composed of a time evidence fusion layer, a space fingerprint inference layer and a space-time arbitration and priority decision layer; S33, the time evidence fusion layer is constructed, the stability of the respiratory flow reversal point, the consistency of the respiratory pressure wave trough phase and the predictability of the ECG advance are evaluated in parallel, the time confidence, the time stability label and the alignment window check result are output according to the fixed length statistical window and the alignment window, and the isolation and rollback mark are executed on the abnormal segment; S34, the space fingerprint inference layer is constructed, the molecular sieve health fingerprint and the baseline fingerprint are differentially compared, the baseline consistency deviation, the decay mode matching degree and the temperature drift compensation residual error are evaluated according to the order encoding perturbation index, the space confidence, the decay mode label and the capacity risk level are output, and the unqualified fingerprint segment is executed to be rejected and replaced; S35, the space-time arbitration and priority decision layer is constructed, two sets of interlocking decision tables of alignment priority channel and service life priority channel are set, adaptive arbitration is carried out according to the time confidence, the space confidence, the alignment window check result and the capacity risk level, the dynamic confidence coefficient is generated, the escort time slot protection and the execution delay compensation parameter are generated, the freeze and release mechanism is enabled for the dynamic confidence coefficient, and the rapid fluctuation is inhibited.

[0012] Optionally, the S4 specifically comprises: S41, receiving the breathing phase prediction result and extracting the time position of the next inspiration starting point, receiving the molecular sieve health fingerprint and extracting the bed health index and residual adsorption capacity factor therein, establishing a time slot allocation matrix at a fixed time resolution, and setting an alignment window aligned with the time position; S42, in the global layer of the oxygen production regulation controller, the allowed range of the cycle phase slip, the allowed range of the valve opening and closing, and the allowed range of the compressor speed are set according to the bed health index and the residual adsorption capacity factor, and the alignment priority scheduling table and the service life priority scheduling table are respectively generated within the allowed range; S43, in the local layer of the oxygen production regulation controller, the target cycle phase slip, the valve opening and closing sequence and the holding time, the compressor speed scheduling parameter and the buffer tank pre-charge of the current period are determined according to the time slot allocation matrix and the alignment window, and the buffer tank is allocated with an escort time slot when the molecular sieve bed is switched; S44, in the interactive layer of the oxygen production regulation controller, performing weight synthesis between the alignment priority scheduling table and the life priority scheduling table according to the dynamic confidence coefficient to form a comprehensive scheduling table of the current period, and adding execution delay compensation time and execution priority to each time slot in the comprehensive scheduling table; S45, encoding the comprehensive scheduling table into an oxygen production scheduling instruction, the oxygen production scheduling instruction containing atomized instruction entries of the cycle phase slip amount, the valve opening and closing sequence and the holding time length, the compressor speed scheduling parameter, the buffer tank pre-charge amount, the escort time slot and the execution delay compensation time, and outputting to a to-be-executed state.

[0013] Optionally, the S5 specifically comprises: S51, in the global layer of the oxygen production regulation controller, reading the bed layer health index and the residual adsorption capacity factor in the molecular sieve health fingerprint, selecting a long-term operation mode, and setting an allowed range of the cycle phase slip amount, an allowed range of the valve opening and closing, an allowed range of the compressor speed, and an allowed range of the buffer tank pre-charge amount; S52, in the local layer of the oxygen production regulation controller, determining a target cycle phase slip amount of the current period according to the breathing phase prediction result, arranging the opening and closing sequence and the holding time length of each execution valve, giving a compressor speed scheduling parameter consistent with the target cycle phase, and a buffer tank pre-charge target of the current period; S53, in the interactive layer of the oxygen production regulation controller, obtaining the dynamic confidence coefficient, and performing weight synthesis on each allowed range set by the global layer and each target value formed by the local layer to generate a comprehensive scheduling entry; S54, in the interactive layer of the oxygen production regulation controller, performing consistency checking on the comprehensive scheduling entry, checking the alignment degree of the cycle phase and the breathing phase, the mutual exclusion relationship between the valve actions, the rate limit of the compressor speed change, and the conflict relationship between the buffer tank pre-charge and the emptying, and making a nearby correction to the entries that do not meet the constraints within the allowed range; S55, after completing the consistency checking, submitting the comprehensive scheduling entry as the output of the long-term operation mode scheduling.

[0014] Optionally, the S6 specifically comprises: S61, executing the oxygen production scheduling instruction, executing according to the cycle phase slip amount, the valve opening and closing sequence and the holding time length, the compressor speed, and the buffer tank target pressure in the oxygen production scheduling instruction, and recording the execution time stamp and the actual feedback of each execution component; S62, obtaining the outlet oxygen concentration, the pipeline pressure, the bed temperature, the valve position feedback and the compressor speed feedback at a fixed sampling period, and calculating the sliding average value of the outlet oxygen concentration according to a fixed time window; S63, compare the sliding average of the outlet oxygen concentration with a preset lower oxygen concentration threshold value, when lower than the preset lower oxygen concentration threshold value and the duration reaches or exceeds a preset duration threshold value, determine a low oxygen event, generate a low oxygen event flag and record the event time; S64, perform an abnormal operation determination in the same sampling period, specifically: detect whether the pipeline pressure exceeds a preset upper limit threshold value, and whether the bed body temperature exceeds a preset upper limit threshold value; compare whether the difference between the valve position command and the valve position feedback exceeds a preset allowable deviation, and whether the difference between the compressor speed command and the compressor speed feedback exceeds a preset allowable deviation; statistically determine whether the number of apnea reaches a preset triggering threshold value, and if any condition is met, generate a fault event flag and record the fault category; S65, when the low oxygen event flag or the fault event flag is set, immediately switch to a safe oxygen supply curve, close the cycle phase shift function, call a safe valve opening and closing sequence, set a safe compressor speed and a safe buffer tank target pressure, and at the same time start a local audible and light alarm and send a remote alarm; S66, continuously monitor the outlet oxygen concentration, the pipeline pressure and the bed body temperature under the safe oxygen supply curve, and when all indicators return to their respective normal working intervals and continuously meet a preset stability duration threshold value, clear the low oxygen event flag and the fault event flag, remove the alarm, and restore to execute the latest oxygen production scheduling instruction.

[0015] The beneficial effects of the present application are: The present application introduces a breathing phase prediction method in the oxygen production process, which can accurately identify the next inspiration starting point before the patient inhales, thereby realizing the dynamic locking of the oxygen production cycle and the patient's breathing rhythm. Compared with the traditional fixed oxygen supply mode, the synchronization of oxygen supply and breathing is improved, so that the patient can obtain sufficient and stable oxygen supply at the key inspiration moment, thereby improving the oxygen utilization efficiency and improving the clinical treatment effect, while reducing the discomfort caused by asynchronous oxygen supply.

[0016] In terms of molecular sieve operating state monitoring, the present application applies a perturbation excitation in the low load stage of the molecular sieve adsorption cycle, combines acoustic-thermal dynamic response characteristics, and constructs a molecular sieve health fingerprint, which can identify different failure modes such as pulverization, channelization and wetting in real time, and quantitatively obtain the bed health index and residual adsorption capacity factor, thereby improving the accuracy and real-time performance of molecular sieve health diagnosis, avoiding the problems of relying on experience estimation and lag detection in the prior art, effectively prolonging the service life of the molecular sieve, and reducing the maintenance frequency and replacement cost.

[0017] At the scheduling and control level, the hierarchical confidence space-time scheduling control method and the dynamic confidence coefficient mechanism proposed by the present application enable the organic combination of long-term operation optimization of the global layer, real-time oxygen supply demand of the local layer and dynamic weight adjustment of the interaction layer. By introducing time confidence and space confidence in the scheduling process, the system can maintain the robustness of scheduling in an uncertain environment, realize dynamic correction and optimization of the oxygen production instruction. This mechanism not only improves the stability and safety of oxygen concentration output, but also enhances the self-adaptive ability of the system in complex use scenarios, improves the safety, reliability and clinical application value of the medical molecular sieve oxygen production equipment. BRIEF DESCRIPTION OF DRAWINGS

[0018] The accompanying drawings are included to provide a further understanding of the present application, and constitute a part of the specification, illustrate the present application and explain the principles of the present application, and do not constitute a limitation of the present application. In the drawings:

[0019] Figure 1 A flowchart of an oxygen production regulation method for a medical molecular sieve oxygen production equipment proposed by the present application; Figure 2 A three-layer structure schematic diagram of a space-time confidence network model of an oxygen production regulation method for a medical molecular sieve oxygen production equipment proposed by the present application. DETAILED DESCRIPTION

[0020] The present application will now be further described in detail in conjunction with the drawings. These drawings are all simplified schematic diagrams, and only illustrate the basic structure of the present application in a schematic manner, and therefore only show the components related to the present application.

[0021] Reference Figure 1 and Figure 2 An oxygen production regulation method for a medical molecular sieve oxygen production equipment, comprising the following steps: S1, collecting patient respiratory flow signal and pressure signal, extracting respiratory related features combined with electrocardiogram signal, determining next inspiration starting point, generating respiratory phase prediction result; S2, applying perturbation excitation at the low load stage of the molecular sieve adsorption cycle, collecting molecular sieve bed outer wall acoustic admittance spectrum and micro-temperature rise signal, constructing acoustic-thermal dynamic response data set, generating molecular sieve health fingerprint spectrum, calculating residual adsorption capacity factor and bed health index; S3, constructing a space-time confidence network model, taking the uncertainty of the respiratory phase prediction result as the time confidence input, taking the diagnostic uncertainty of the bed health index and the residual adsorption capacity factor in the molecular sieve health fingerprint spectrum as the space confidence input, forming a dynamic confidence coefficient, and dynamically adjusting the scheduling optimization range; S4, input the breathing phase prediction result into the oxygen production regulation controller, determine the valve opening and closing sequence, adsorption cycle phase shift amount and compressor speed scheduling parameters based on the hierarchical confidence space-time scheduling control method, and generate oxygen production scheduling instructions; S5, in the global layer of the oxygen production regulation controller, execute long-term operation mode scheduling according to the bed layer health index and residual adsorption capacity factor in the molecular sieve health fingerprint, in the local layer, adjust the cycle phase and valve action in real time according to the breathing phase prediction result, and in the interactive layer, adjust the scheduling weight of the global layer and the local layer according to the dynamic confidence coefficient; S6, execute the oxygen production scheduling instructions, monitor the outlet oxygen concentration in real time, compare the real-time concentration with the preset threshold value, when the outlet oxygen concentration is lower than the threshold value or an abnormal operating state is detected, switch to the safe oxygen supply curve, maintain stable oxygen supply and output an alarm signal.

[0022] In the embodiment, the generation of the breathing phase prediction result refers to extracting the flow reversal point, pressure valley point and modulation component related to the breathing rhythm in the electrocardiogram signal in the breathing cycle, fusing and determining the time position of the next inspiration starting point, and outputting the determined time position as the breathing phase prediction result.

[0023] In the embodiment, S2 specifically includes: S21, apply a sequence-encoding perturbation sequence in the low load stage of the molecular sieve adsorption cycle, the sequence-encoding perturbation sequence is composed of a combination of multiple frequencies and multiple amplitudes, and acts on the inlet valve in a determined start and end time in sequence, and the acoustic original waveform and temperature original sequence of the outer wall of the molecular sieve bed are synchronously collected; S22, segment the acoustic original waveform and temperature original sequence by fixed frequency band and fixed time window, extract the set of acoustic-thermal features of each frequency band, and generate a multi-dimensional response feature set; S23, establish a molecular sieve health fingerprint according to the multi-dimensional response feature set in the order of the perturbation sequence and the frequency band number, and obtain a baseline fingerprint by full-frequency scanning when the device is first operated or maintenance is reset; S24, calculate the residual adsorption capacity factor by using the thermo-acoustic mutual verification capacity evaluation method: Compare the acoustic response energy ratio in the molecular sieve health fingerprint with the baseline fingerprint, and record the comparison deviation; Compare the time lag of the acoustic peak value and the temperature peak value with the baseline fingerprint, and record the lag deviation; Compare the consistency of the peak decay rate of the multi-frequency band with the baseline fingerprint, and record the consistency deviation; According to the preset segmentation weighting rule, aggregate the comparison deviation, lag deviation and consistency deviation, and output the residual adsorption capacity factor; S25, determine the bed health index by using the decline mode and risk aggregation method: On the molecular sieve health fingerprint map, identify the pulverization mode, channelization mode and humidification mode according to the characteristic combination rule, and obtain the mode risk score; Under the continuous multiple sequential encoding perturbation sequence, the time stability of the molecular sieve health fingerprint map is checked, and a stability score is obtained; The temperature original sequence is compensated for temperature drift to obtain a temperature drift compensation score; According to the preset segmented weighting rule, the mode risk score, the stability score and the temperature drift compensation score are aggregated, and the bed health index is output.

[0024] In this embodiment, the characteristic combination rule refers to: When the acoustic resonance peak energy ratio continues to decrease, the temperature rise amplitude abnormally fluctuates, and the acoustic peak attenuation rate exceeds the stable range, it is determined as the pulverization mode; When the acoustic resonance peak position shifts to the high frequency end, the acoustic response energy ratio increases, and the time lag between the acoustic peak value and the temperature peak value shortens, combined with the fact that the bed health index decline rate exceeds twice the baseline average decline rate, it is determined as the channelization mode; When the temperature rise amplitude continues to decrease, the time lag between the acoustic peak value and the temperature peak value increases, and the residual adsorption capacity factor decreases at an accelerated rate, it is determined as the humidification mode.

[0025] In this embodiment, the S3 specifically includes: S31, establish the input and alignment window of the space-time belief network model, obtain the breathing phase prediction result and the corresponding actual inspiration start time mark in the continuous control period, and obtain the molecular sieve health fingerprint map and the baseline fingerprint map in the continuous control period; S32, construct the space-time belief network model, which is composed of a time evidence fusion layer, a space fingerprint inference layer, and a space-time arbitration and priority decision layer; S33, construct the time evidence fusion layer, and perform parallel evaluation on the breathing flow reversal point stability, the breathing pressure wave trough phase consistency and the electrocardio lead time predictability. Output the time confidence, the time stability label and the alignment window checking result according to the fixed length statistical window and the alignment window, and perform isolation and rollback marking on the abnormal segment; S34, construct the space fingerprint inference layer, and perform differential comparison on the molecular sieve health fingerprint map and the baseline fingerprint map. According to the sequential encoding perturbation index, evaluate the baseline consistency deviation, the decline mode matching degree and the temperature drift compensation residual error. Output the space confidence, the decline mode label and the capacity risk level, and perform rejection and replacement marking on the unqualified fingerprint segment; S35, construct the time and space arbitration and priority decision layer, set two sets of interlocking decision tables of alignment priority channel and life priority channel, perform adaptive arbitration according to time confidence, space confidence, alignment window checking result and capacity risk level, generate dynamic confidence coefficient, and generate escort time slot protection and execution delay compensation parameters, enable freezing and cold release mechanism for the dynamic confidence coefficient, and inhibit rapid fluctuations.

[0026] In the embodiment, the S4 specifically includes: S41, receive the breathing phase prediction result and extract the time position of the next inspiration starting point, receive the molecular sieve health fingerprint and extract the bed health index and residual adsorption capacity factor therein, establish a time slot allocation matrix at a fixed time resolution, and set an alignment window aligned with the time position; S42, at the global layer of the oxygen production regulation controller, set the allowed range of the cycle phase shift, the allowed range of the valve opening and closing, and the allowed range of the compressor speed according to the bed health index and the residual adsorption capacity factor, and generate the alignment priority scheduling table and the life priority scheduling table within the allowed range respectively; S43, at the local layer of the oxygen production regulation controller, determine the target cycle phase shift, the valve opening and closing sequence and the holding time, the compressor speed scheduling parameter and the buffer tank pre-charge amount of the current period according to the time slot allocation matrix and the alignment window, and allocate the escort time slot for the buffer tank when the molecular sieve bed is switched; S44, at the interaction layer of the oxygen production regulation controller, weight synthesis between the alignment priority scheduling table and the life priority scheduling table according to the dynamic confidence coefficient, form the comprehensive scheduling table of the current period, and add the execution delay compensation time and the execution priority to each time slot in the comprehensive scheduling table; S45, encode the comprehensive scheduling table into oxygen production scheduling instructions, the oxygen production scheduling instructions include atomized instruction entries of cycle phase shift, valve opening and closing sequence and holding time, compressor speed scheduling parameter, buffer tank pre-charge amount, escort time slot and execution delay compensation time, and output to the to-be-executed state.

[0027] In the embodiment, the S5 specifically includes: S51, at the global layer of the oxygen production regulation controller, read the bed health index and the residual adsorption capacity factor in the molecular sieve health fingerprint, select the long-term operation mode, set the allowed range of the cycle phase shift, the allowed range of the valve opening and closing, the allowed range of the compressor speed and the allowed range of the buffer tank pre-charge amount; S52, at the local layer of the oxygen production regulation controller, determine the target cycle phase shift of the current period according to the breathing phase prediction result, arrange the opening and closing sequence and the holding time of each execution valve, give the compressor speed scheduling parameter consistent with the target cycle phase and the buffer tank pre-charge amount target of the current period; S53, in the interaction layer of the oxygen production regulation controller, obtain the dynamic confidence coefficient, and perform weight synthesis on the allowed range set by the global layer and the target values formed by the local layer to generate a comprehensive scheduling entry; S54, in the interaction layer of the oxygen production regulation controller, perform consistency checking on the comprehensive scheduling entry, check the alignment degree of the cycle phase and the breathing phase, the mutual exclusion relationship between the valve actions, the rate limit of the compressor speed change, and the conflict relationship of the buffer tank pre-charging and emptying, and make a nearby correction to the entries that do not meet the constraints within the allowed range; S55, after completing the consistency checking, submit the comprehensive scheduling entry as the output of the long-term operation mode scheduling.

[0028] In this embodiment, S6 specifically includes: S61, execute the oxygen production scheduling instruction, execute according to the cycle phase slip amount, valve opening and closing sequence and holding time, compressor speed and buffer tank target pressure specified in the oxygen production scheduling instruction, and record the execution timestamp and the actual feedback of each execution component; S62, obtain the outlet oxygen concentration, pipeline pressure, bed temperature, valve position feedback and compressor speed feedback at a fixed sampling period, and calculate the sliding average of the outlet oxygen concentration according to a fixed time window; S63, compare the sliding average of the outlet oxygen concentration with the preset lower limit threshold of the oxygen concentration, and when it is lower than the preset lower limit threshold of the oxygen concentration and the duration reaches or exceeds the preset duration threshold, determine that it is a low oxygen event, generate a low oxygen event flag and record the event time; S64, execute abnormal operation judgment in the same sampling period, specifically: Detect whether the pipeline pressure exceeds the preset upper limit threshold, and detect whether the bed temperature exceeds the preset upper limit threshold; Compare whether the difference between the valve position command and the valve position feedback exceeds the preset allowed deviation, and compare whether the difference between the compressor speed command and the compressor speed feedback exceeds the preset allowed deviation; Statistically determine whether the number of apnea reaches the preset trigger threshold, and generate a fault event flag and record the fault category if any condition is met; S65, when the low oxygen event flag or the fault event flag is set, immediately switch to the safe oxygen supply curve, close the cycle phase slip function, call the safe valve opening and closing sequence, set the safe compressor speed and the safe buffer tank target pressure, and at the same time start the local sound and light alarm and send the remote alarm; S66, continuously monitor the outlet oxygen concentration, pipeline pressure and bed temperature under the safe oxygen supply curve, and when all indicators return to their respective normal working intervals and continuously meet the preset stable time threshold, clear the low oxygen event flag and the fault event flag, remove the alarm, and restore to execute the latest oxygen production scheduling instruction.

[0029] Example 1 To verify the feasibility of the application in implementation, the application is applied to the respiratory medicine department of a people's hospital. A 68-year-old patient with chronic obstructive pulmonary disease receives long-term oxygen therapy due to hypoxemia. The patient's respiratory rate in a quiet state is 16 times per minute, and the inhalation time accounts for about 40% of the entire respiratory cycle. The traditional molecular sieve oxygen generating equipment uses fixed valve switching and fixed cycle mode when supplying oxygen, which cannot dynamically match the patient's breathing rhythm. When the patient uses the traditional equipment at night, the blood oxygen saturation often fluctuates between 91% and 93%, and there is a delay in oxygen flow at the beginning of inhalation, causing the patient to experience mild respiratory discomfort.

[0030] In this scenario, the oxygen generating and adjusting method proposed by the application is applied. The device collects the respiratory flow signal and pressure signal at the patient's nasal catheter end, and synchronously records the electrocardiogram signal for extracting the respiratory related features. By predicting the flow change of the next five consecutive respiratory cycles, the device can lock the next inhalation start point in advance, so that the oxygen generating cycle starts in advance before the patient inhales. The monitoring results show that the time difference between the patient's inhalation start point and the oxygen flow increase point is shortened from an average of 280 milliseconds of the traditional equipment to 90 milliseconds, improving the oxygen supply synchronization.

[0031] In the operation monitoring link of the molecular sieve, the device applies a slight pressure disturbance during the low load stage of the adsorption cycle, collects the acoustic admittance spectrum and micro-temperature rise signal of the bed outer wall, and constructs the molecular sieve health fingerprint. By analyzing the feature combination, three degradation modes of pulverization, channelization and wetting can be identified, and the residual adsorption capacity factor and bed health index can be calculated. In the use cycle of this patient, the residual adsorption capacity factor of the molecular sieve is 0.87 (full value is 1.0), and the bed health index is 0.91, both of which are within the safe range. Unlike the traditional equipment which relies on the use time to estimate the service life, the fingerprint diagnosis method of the application can identify the mild pulverization trend at an early stage, reminding the hospital to replace it in advance in subsequent maintenance, reducing the risk of clinical oxygen supply process.

[0032] In terms of control scheduling, the application uses a space-time confidence network model, taking the uncertainty in respiratory phase prediction as the time confidence input and the uncertainty in molecular sieve diagnosis as the space confidence input, and fusing them to form a dynamic confidence coefficient. When the patient turns over at night, causing the respiratory signal to fluctuate, the time confidence decreases, and the controller automatically increases the weight of the molecular sieve state to avoid oxygen supply fluctuations caused by unstable prediction. In a week of continuous monitoring, the patient's blood oxygen saturation at night is stable between 93% and 95%, which is about 60% less than the traditional equipment when the blood oxygen saturation is below 92%.

[0033] Finally, in the safety control link, the application monitors the outlet oxygen concentration in real time. When the concentration is lower than the preset threshold of 90%, the device immediately switches to the safe oxygen supply curve and starts the audible and visual alarm. In a short pipeline compression event, the outlet oxygen concentration dropped to 89%, and the device completed the switch within 6 seconds, maintaining the patient's blood oxygen saturation above 92% and avoiding a hypoxia event. In contrast to the traditional device, the alarm delay in a similar event was more than 20 seconds, and the patient's blood oxygen dropped to 87%, which is significantly less safe clinically.

[0034] Table 1 Comparison test data of the application method and the traditional device

[0035] As can be seen from the data in Table 1, the application method is significantly better than the traditional device in multiple key indicators. In terms of respiratory synchrony, the traditional device has a delay of about 280 milliseconds between the start of inspiration and the start of oxygen supply, while the application method shortens this delay to 90 milliseconds, a reduction of 68%, enabling the oxygen release to more accurately match the patient's inspiration timing. The oxygen flow at the inspiration moment is increased from 0.75 L / min to 1.15 L / min, an increase of about 53%, and the patient receives more adequate oxygen supply at the peak of inspiration, fundamentally solving the problem of asynchronous oxygen supply.

[0036] In terms of blood oxygen stability, the patient's blood oxygen saturation fluctuation range under the traditional device is 91%-93%, while the application method improves the range to 93%-95%, with smaller fluctuations and higher stability. The number of events with blood oxygen saturation below 92% per night is reduced from 7 to 3, a reduction of about 60%, indicating that the method can effectively prevent the occurrence of hypoxia events, thereby improving the continuity and safety of clinical treatment.

[0037] In terms of molecular sieve health monitoring and safety protection, the application method can calculate the molecular sieve residual adsorption capacity factor and bed health index in real time, reaching 0.87 and 0.91 respectively, while the traditional device cannot achieve real-time diagnosis. This means that the device can identify potential failure modes of the molecular sieve at an early stage, thereby extending the service life and reducing maintenance costs. In abnormal event handling, the application method can complete the switch to the safe oxygen supply mode within 6 seconds when the pipeline is abnormal, while the traditional device needs 22 seconds, with a response speed improvement of 73%, and the patient's blood oxygen minimum value is increased from 87% to 92%, reducing the clinical risk. Overall, the application method not only improves the oxygen supply synchrony and stability, but also enhances the safety and maintainability, with high clinical application value and promotion prospects.

[0038] The above merely describes preferred specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art, according to the technical solution and inventive concept of the present application, makes equivalent replacement or change within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A method for regulating oxygen generation in a medical molecular sieve oxygen generator, characterized in that, Includes the following steps: S1. Collect the patient's respiratory flow and pressure signals, extract respiratory-related features by combining electrocardiogram signals, determine the starting point of the next inspiratory phase, and generate respiratory phase prediction results; S2. Apply perturbation excitation during the low-load stage of the molecular sieve adsorption cycle, collect the acoustic admittance spectrum and micro-temperature rise signal of the outer wall of the molecular sieve bed, construct an acoustic-thermal dynamic response dataset, generate a molecular sieve health fingerprint spectrum, and calculate the residual adsorption capacity factor and bed health index. S3. Construct a spatiotemporal confidence network model, using the uncertainty of the respiratory phase prediction results as the temporal confidence input, and the diagnostic uncertainty of the bed health index and residual adsorption capacity factor in the molecular sieve health fingerprint as the spatial confidence input, to form a dynamic confidence coefficient and dynamically adjust the scheduling optimization range. S4. Input the respiratory phase prediction results and molecular sieve health fingerprint spectrum into the oxygen production regulator controller, and determine the valve opening and closing sequence, adsorption cycle phase shift amount and compressor speed scheduling parameters based on the hierarchical confidence time-space temperature control method to generate oxygen production scheduling instructions. S5. At the global layer of the oxygen generation regulator, the long-term operation mode is scheduled based on the bed health index and residual adsorption capacity factor in the molecular sieve health fingerprint spectrum. At the local layer, the circulation phase and valve action are adjusted in real time based on the respiratory phase prediction results. At the interaction layer, the scheduling weights of the global layer and the local layer are adjusted based on the dynamic confidence coefficient. S6. Execute oxygen production scheduling instructions, monitor the outlet oxygen concentration in real time, compare the real-time concentration with the preset threshold, and when the outlet oxygen concentration is lower than the threshold or an abnormal operating state is detected, switch to the safe oxygen supply curve, maintain stable oxygen supply and output an alarm signal.

2. The oxygen generation regulation method for a medical molecular sieve oxygen generator according to claim 1, characterized in that, The generation of respiratory phase prediction results refers to extracting the flow reversal point, pressure trough point, and modulation components related to respiratory rhythm in the respiratory cycle, fusing and judging the extracted time features, determining the time position of the next inspiratory start point, and outputting the determined time position as the respiratory phase prediction result.

3. The oxygen generation regulation method for a medical molecular sieve oxygen generator according to claim 1, characterized in that, S2 specifically includes: S21. During the low-load stage of the molecular sieve adsorption cycle, a sequentially encoded perturbation sequence is applied. The sequentially encoded perturbation sequence is composed of a preset combination of multiple frequencies and multiple amplitudes, and is applied to the inlet valve in sequence within a determined start and end time. The acoustic waveform and temperature sequence of the outer wall of the molecular sieve bed are collected simultaneously. S22. Perform fixed frequency band and fixed time window segmentation on the original acoustic waveform and temperature sequence, extract the acoustic and thermal feature sets of each frequency band, and generate a multidimensional response feature set. S23. Based on the multidimensional response feature set, establish a molecular sieve health fingerprint spectrum according to the perturbation sequence order and frequency band number, and obtain the baseline fingerprint spectrum by full-frequency scanning when the equipment is first run or during maintenance reset. S24. Calculate the residual adsorption capacity factor using the thermo-acoustic cross-validation capacity assessment method: The acoustic response energy ratio in the molecular sieve health fingerprint spectrum was compared with the baseline fingerprint spectrum, and the comparison deviation was recorded. The time lag between acoustic peak and temperature peak is compared with the baseline fingerprint spectrum, and the lag deviation is recorded. The peak attenuation rate consistency across multiple frequency bands is compared with the baseline fingerprint, and the consistency deviation is recorded. According to the preset segmented weighting rules, the comparison bias, hysteresis bias and consistency bias are aggregated to output the residual adsorption capacity factor. S25. Calculate the bed health index using the decline pattern determination and risk aggregation method: Based on the feature combination rules, the pulverization mode, channelization mode and humidification mode are identified on the molecular sieve health fingerprint spectrum, and the mode risk scores are obtained respectively. The time stability of the molecular sieve health fingerprint was checked under multiple consecutive sequentially encoded perturbation sequences to obtain the stability score; Temperature drift compensation is performed on the original temperature sequence to obtain a temperature drift compensation score. According to the preset segmented weighting rules, the model risk score, stability score and temperature drift compensation score are aggregated to output the bed health index.

4. The oxygen generation regulation method for a medical molecular sieve oxygen generator according to claim 1, characterized in that, The feature combination rule refers to: When the acoustic resonant energy ratio continues to decrease, the temperature rise amplitude fluctuates abnormally, and the acoustic peak decay rate exceeds the stable range, it is determined to be in pulverization mode. When the position of the acoustic resonance peak shifts to the high-frequency end, the acoustic response energy ratio increases, and the time lag between the acoustic peak and the temperature peak shortens, combined with the situation where the rate of decline of the bed health index exceeds twice the baseline average decline rate, it is determined to be a channelization mode. When the rate of temperature increase continues to decrease, the time lag between the acoustic peak and the temperature peak increases, and the rate of decrease of the residual adsorption capacity factor accelerates, it is determined to be a humidification mode.

5. The oxygen generation regulation method for a medical molecular sieve oxygen generator according to claim 1, characterized in that, S3 specifically includes: S31. Establish the input and alignment window of the spatiotemporal confidence network model, obtain the respiratory phase prediction results and the corresponding actual inspiratory start time markers within the continuous control cycle, and obtain the molecular sieve health fingerprint spectrum and baseline fingerprint spectrum within the continuous control cycle. S32. Construct a spatiotemporal confidence network model, which consists of a temporal evidence fusion layer, a spatial fingerprint inference layer, and a spatiotemporal arbitration and priority decision layer. S33. Construct a time evidence fusion layer to evaluate in parallel the stability of respiratory flow reversal point, the phase consistency of respiratory pressure trough, and the predictability of ECG advance. Output time confidence, time stability label, and alignment window verification results according to fixed-length statistical windows and alignment windows, and perform isolation and rollback marking on abnormal segments. S34. Construct a spatial fingerprint inference layer, perform differential comparison between the molecular sieve health fingerprint spectrum and the baseline fingerprint spectrum, encode perturbation index in sequence to evaluate baseline consistency deviation, decay mode matching degree and temperature drift compensation residual, output spatial confidence, decay mode label and capacity risk level, and perform rejection and replacement marking on unqualified fingerprint segments. S35. Construct a spatiotemporal arbitration and priority decision layer, set up two sets of interlocked decision tables for alignment priority channel and lifetime priority channel, and perform adaptive arbitration based on time confidence, spatial confidence, alignment window verification results and capacity risk level to generate dynamic confidence coefficients, and generate escort time slot protection and execution delay compensation parameters. Enable freezing and releasing mechanisms for dynamic confidence coefficients to suppress rapid fluctuations.

6. The oxygen generation regulation method for a medical molecular sieve oxygen generator according to claim 1, characterized in that, S4 specifically includes: S41. Receive the respiratory phase prediction result and extract the time position of the next inhalation start point, receive the molecular sieve health fingerprint spectrum and extract the bed health index and residual adsorption capacity factor therein, establish a time slot allocation matrix with a fixed time resolution, and set an alignment window aligned with the time position. S42. In the global layer of the oxygen generation regulator, the allowable range of the cycle phase slip, the allowable range of valve opening and closing, and the allowable range of compressor speed are set according to the bed health index and residual adsorption capacity factor, and an alignment priority scheduling table and a lifespan priority scheduling table are generated within the allowable ranges respectively. S43. In the local layer of the oxygen generation regulator, the target cycle phase shift, valve opening and closing sequence and holding time, compressor speed scheduling parameters and buffer tank pre-charge amount are determined according to the time slot allocation matrix and alignment window. Escort time slots are allocated to the buffer tank when the molecular sieve bed is switched. S44. In the interaction layer of the oxygen production regulator, the alignment priority scheduling table and the lifespan priority scheduling table are weighted and synthesized according to the dynamic confidence coefficient to form the comprehensive scheduling table for this cycle, and the execution delay compensation time and execution priority are added to each time slot in the comprehensive scheduling table. S45. The comprehensive scheduling table is encoded into an oxygen production scheduling instruction. The oxygen production scheduling instruction includes atomic instruction entries for cycle phase shift, valve opening and closing sequence and holding duration, compressor speed scheduling parameters, buffer tank pre-charge amount, escort time slot and execution delay compensation time, and is output to the pending execution state.

7. The oxygen generation regulation method for a medical molecular sieve oxygen generator according to claim 1, characterized in that, S5 specifically includes: S51. In the global layer of the oxygen generation regulator controller, read the bed health index and residual adsorption capacity factor in the molecular sieve health fingerprint spectrum, select the long-term operation mode, and set the allowable range of the cyclic phase slip, the allowable range of valve opening and closing, the allowable range of compressor speed and the allowable range of buffer tank pre-charge. S52. In the local layer of the oxygen generation regulator controller, the target cycle phase shift amount for this cycle is determined based on the respiratory phase prediction results. The opening and closing sequence and holding time of each actuator valve are arranged, and the compressor speed scheduling parameters consistent with the target cycle phase and the buffer tank pre-charge amount target for this cycle are given. S53. In the interaction layer of the oxygen production regulator, obtain the dynamic confidence coefficient, perform weighted synthesis of each allowable range set in the global layer and each target value formed in the local layer, and generate a comprehensive scheduling entry. S54. In the interaction layer of the oxygen generation regulator controller, perform consistency checks on the comprehensive scheduling items, check the alignment degree between the cycle phase and the breathing phase, the mutual exclusion relationship between valve actions, the rate limit of compressor speed change, and the conflict relationship between buffer tank pre-charging and venting, and make corrections to items that do not meet the constraints within each allowable range. S55. After completing the consistency verification, submit the integrated scheduling entries as the output of the long-term operation mode scheduling.

8. The oxygen generation regulation method for a medical molecular sieve oxygen generator according to claim 1, characterized in that, S6 specifically includes: S61. Execute the oxygen production scheduling command according to the cycle phase slip amount, valve opening and closing sequence and holding time, compressor speed and buffer tank target pressure specified in the oxygen production scheduling command, and record the execution timestamp and the actual feedback of each execution component. S62. Obtain outlet oxygen concentration, pipeline pressure, bed temperature, valve position feedback and compressor speed feedback at a fixed sampling period, and calculate the sliding average value of outlet oxygen concentration according to a fixed time window; S63. Compare the sliding average value of the outlet oxygen concentration with the preset lower limit threshold. When the oxygen concentration is lower than the preset lower limit threshold and the duration reaches or exceeds the preset duration threshold, it is determined to be a low oxygen event, a low oxygen event flag is generated and the event time is recorded. S64. Perform abnormal operation judgment within the same sampling period, specifically as follows: Check whether the pipeline pressure exceeds the preset upper limit threshold and whether the bed temperature exceeds the preset upper limit threshold; Compare whether the difference between the valve position command and the valve position feedback exceeds the preset allowable deviation; compare whether the difference between the compressor speed command and the compressor speed feedback exceeds the preset allowable deviation. The system counts whether the number of apneas reaches a preset trigger threshold. If any condition is met, a fault event flag is generated and the fault category is recorded. S65. When the low oxygen event flag or fault event flag is set, immediately switch to the safe oxygen supply curve, turn off the circulation phase slip function, call the safety valve opening and closing sequence, set the safety compressor speed and the safety buffer tank target pressure, and at the same time start the local audible and visual alarm and send a remote alarm. S66. Under the safe oxygen supply curve, continuously monitor the outlet oxygen concentration, pipeline pressure and bed temperature. When all indicators have returned to their respective normal working range and continuously meet the preset stable duration threshold, clear the low oxygen event flag and fault event flag, deactivate the alarm, and resume the execution of the latest oxygen production scheduling command.

Images