A Ventilator Fault Prediction Method and System Based on Big Data

Abstract

The present invention discloses a method and system for predicting ventilator faults based on big data, specifically related to the technical field of fault prediction. It includes using the long short-term memory algorithm to predict the process of the ventilator dynamically adjusting the gas volume. Through the output results of the LSTM model, the prediction results at different time steps are obtained, the prediction results are analyzed, the possible abnormal situations are evaluated, early warning grading is carried out, and it is judged whether to detect the output pipeline during the operation of the ventilator. A fault assessment model is constructed to quantify the results of the possibility of the ventilator having a fault. According to the results of the fault assessment model, the fault assessment coefficient for the ventilator to dynamically match the patient's breathing process is obtained and compared with the threshold of the fault assessment coefficient to reconfirm the early warning level of the ventilator. The present invention helps to take preventive measures in the initial stage of the fault, prevent the harm caused by the accumulation of faults, and improve the reliability of ventilator fault identification.

Description

Technical Field

[0001] The present invention relates to the technical field of fault prediction, and more specifically, to a method and system for predicting ventilator faults based on big data. Background Art

[0002] Currently, a ventilator can dynamically adjust the gas volume to match the patient's breathing frequency. Among them, the process of the ventilator dynamically adjusting the gas volume is to automatically adjust parameters such as the flow rate, pressure, and oxygen concentration of the gas by real-time monitoring the patient's breathing condition to meet the patient's breathing needs. The whole process is a closed-loop process. However, once the ventilator fails, it may lead to an inability to correctly match the patient's breathing needs.

[0003] Traditional ventilator fault prediction methods mainly rely on simple algorithms or single-sensor data, such as pressure, air flow, temperature, etc. for prediction, which may have certain limitations. Due to the complex working environment of the ventilator and the difficulty of fully reflecting the changes in the patient's state through these single parameters in a timely manner, the prediction accuracy is usually not high. Therefore, it is difficult to determine whether it is a ventilator fault or a patient factor when there is a problem with the dynamic adjustment of the ventilator, and it is not easy to give a timely warning of possible faults of the ventilator.

[0004] To solve the above defects, a technical solution is provided now. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a method and system for predicting ventilator faults based on big data to solve the problems raised in the above background art.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A method for predicting ventilator faults based on big data specifically includes the following steps:

[0008] S1: Use the long short-term memory algorithm to predict the process of the ventilator dynamically adjusting the gas volume, and determine to detect abnormal situations in advance when the ventilator may fail;

[0009] S2: Obtain the prediction results at different time steps through the output results of the LSTM model, analyze the prediction results, evaluate possible abnormal situations, perform early warning grading, and determine whether to detect the output pipeline during the operation of the ventilator;

[0010] S3: Collect the flow information and air pressure information in the output pipeline during the operation of the ventilator, construct a fault assessment model based on the comprehensive analysis of the flow information and air pressure information, and quantify the results of the possibility of the ventilator failing;

[0011] S4: Based on the results of the fault assessment model, obtain the fault assessment coefficient for the ventilator to dynamically match the patient's breathing process, and compare it with the fault assessment coefficient threshold to reconfirm the warning level of the ventilator.

[0012] In a preferred embodiment, the process of the long short-term memory algorithm predicting the dynamic adjustment of the gas volume of the ventilator includes:

[0013] Determine the input variables and output variables of the long short-term memory algorithm. Among them, the input variables of the long short-term memory algorithm are the residual oxygen concentration, the residual carbon dioxide concentration, the gas exchange rate, and the tidal volume deviation, and the output variable is the adjustment deviation, that is, the difference between the total gas volume output from the ventilator output pipeline after each dynamic matching of the ventilator and the tidal volume inhaled by the patient;

[0014] Keep the time step of the adjustment deviation consistent with the output time series generated by the smallest time step among the input variables.

[0015] Construct an LSTM model, use LSTM units to process time series data. Each LSTM unit contains a memory unit and a gating mechanism, and pass the output of the LSTM to the fully connected layer;

[0016] According to the output result of the LSTM model, obtain the predicted adjustment deviation, and calculate the average value and standard deviation of the adjustment deviation according to the adjustment deviations of different time steps.

[0017] In a preferred embodiment, the warning classification is carried out, including:

[0018] Set the average value threshold of the adjustment deviation and the standard deviation threshold of the adjustment deviation, compare the average value and standard deviation of the adjustment deviation with the average value threshold of the adjustment deviation and the standard deviation threshold of the adjustment deviation, and generate the following situations:

[0019] If the average value threshold of the adjustment deviation is greater than the average value of the adjustment deviation, generate a first warning signal;

[0020] If the average value threshold of the adjustment deviation is less than the average value of the adjustment deviation and the standard deviation of the adjustment deviation is greater than the standard deviation threshold of the adjustment deviation, generate a second warning signal;

[0021] If the average value threshold of the adjustment deviation is less than the average value of the adjustment deviation and the standard deviation of the adjustment deviation is less than the standard deviation threshold of the adjustment deviation, no warning signal is generated.

[0022] In a preferred embodiment, collect the flow information in the output pipeline during the operation of the ventilator, including:

[0023] Determine the moment when the second warning signal is generated, collect the flow information and air pressure information in the output pipeline during the operation of the ventilator, quantify the flow information through the gas matching variation coefficient and the flow waveform difference coefficient, and quantify the air pressure information through the air pressure output deviation coefficient.

[0024] In a preferred embodiment, the acquisition logic of the gas matching variation coefficient is as follows:

[0025] Obtain the total gas concentration output in the ventilator output pipeline at the current time point, and mark the total gas concentration output in the ventilator output pipeline at the current time point as: , obtain the total gas concentration output in the patient output pipeline at the current time point, and mark the total gas concentration output in the patient output pipeline at the current time point as: , where g = 1, 2, 3, ……, C, C is a positive integer, and g is the number of the patient's respiratory cycle within the monitoring time period ;

[0026] Obtain the total gas concentration output in the ventilator output pipeline and the total gas concentration output in the patient output pipeline at the previous time point, and mark them as: and ;

[0027] Calculate the first-order difference of the total gas concentration output in the ventilator output pipeline. The calculation formula is: ; where is the first-order difference of the total gas concentration output in the ventilator output pipeline;

[0028] Calculate the first-order difference of the total gas concentration output in the patient output pipeline. The calculation formula is: ; where is the first-order difference of the total gas concentration output in the patient output pipeline;

[0029] Calculate the first-order difference of the dynamically matched gas. The calculation formula is: ; where is the first-order difference of the dynamically matched gas;

[0030] Calculate the second-order difference of the total gas concentration output in the ventilator output pipeline. The calculation formula is: ; where is the second-order difference of the total gas concentration output in the ventilator output pipeline;

[0031] Calculate the second-order difference of the total gas concentration output in the patient output pipeline. The calculation formula is: ; where is the second-order difference of the total gas concentration output in the patient output pipeline;

[0032] Calculate the second-order difference of the dynamically matched gas. The calculation formula is: ; where is the second-order difference of the dynamically matched gas;

[0033] Calculate the mean value of the second-order difference of the dynamically matched gas within the monitoring time period . The calculation formula is: ; where is the mean value of the second-order difference of the dynamically matched gas within the monitoring time period ;

[0034] Calculate the standard deviation of the second-order difference of the dynamically matched gas within the monitoring time period . The calculation formula is: ;

[0035] Calculate the gas matching coefficient of variation. The calculation formula is: ; where is the gas matching coefficient of variation.

[0036] In a preferred embodiment, the acquisition logic of the flow waveform difference coefficient is as follows:

[0037] Through the monitoring device in the ventilator output pipeline, obtain the gas flow rate output by the ventilator output pipeline within the monitoring time period , and mark the gas flow rate output by the ventilator output pipeline within the monitoring time period as: , where b = 1, 2, 3,..., B, B is a positive integer, and b is the number of the gas flow rate collected within the monitoring time period ;

[0038] Generate the flow rate waveform of the output gas according to the gas flow rate data output by the ventilator output pipeline within the monitoring time period , and integrate the flow rate waveform of the output gas within the monitoring time period ;

[0039] Obtain the gas reference flow rate waveform of the gas output by the ventilator output pipeline, and integrate the gas reference flow rate waveform of the gas output by the ventilator output pipeline within the monitoring time period ;

[0040] Calculate the flow waveform difference coefficient. The calculation formula is: ; where ~ are the start time and end time of the monitoring time period , is the flow rate waveform of the output gas generated from the gas flow rate data output by the ventilator output pipeline within the monitoring time period , It is the gas reference flow rate waveform of the gas output from the ventilator output pipe.

[0041] In a preferred embodiment, the acquisition logic of the air pressure output deviation coefficient is as follows:

[0042] Obtain the monitoring time period The pressure of the gas in the ventilator output pipeline during each breathing cycle of the patient within the period is obtained, and the average value of the pressure in the ventilator output pipeline during each breathing cycle of the patient is obtained. Mark the average value of the pressure in the ventilator output pipeline during each breathing cycle of the patient as: Obtain the standard deviation of the average value of the pressure in the ventilator output pipeline during each breathing cycle of the patient, and mark the standard deviation of the average value of the pressure in the ventilator output pipeline during each breathing cycle of the patient as: where, , ;

[0043] Obtain the peak value of the pressure of the gas in the ventilator output pipeline during each breathing cycle of the patient within the monitoring time period , and mark the peak value of the pressure of the gas in the ventilator output pipeline during each breathing cycle of the patient as: ;

[0044] Set the pressure range threshold of the gas in the ventilator output pipeline, and mark the pressure range threshold as: ;

[0045] Calculate the air pressure output deviation coefficient, and the calculation formula is:

[0046] ;

[0047] where, is the air pressure output deviation coefficient.

[0048] In a preferred embodiment, a fault assessment model is constructed, including:

[0049] Perform weighted analysis on the gas matching variation coefficient, the flow waveform difference coefficient, and the air pressure output deviation coefficient to construct a fault assessment model and generate a fault assessment coefficient. The expression of the fault assessment coefficient is: where, is the fault assessment coefficient, , , are the proportionality coefficients of the gas matching variation coefficient, the flow waveform difference coefficient, and the air pressure output deviation coefficient respectively, , , are all greater than 0.

[0050] In a preferred embodiment, reconfirming the warning level of the ventilator includes:

[0051] Set a threshold for the fault evaluation coefficient, compare the fault evaluation coefficient with the threshold. If the fault evaluation coefficient is greater than the threshold, change the second warning signal of the ventilator to the first warning signal. If the fault evaluation coefficient is less than the threshold, continue the process of steps S1 to S4 until the second warning signal disappears or changes to the first warning signal.

[0052] In a preferred embodiment, a big data-based ventilator fault prediction system includes a data acquisition module, an LSTM model warning module, a data analysis module, and a fault evaluation module;

[0053] The data acquisition module is used to collect the input variables required by the LSTM model warning module, including the residual oxygen concentration, the residual carbon dioxide concentration, the gas exchange rate, and the tidal volume deviation;

[0054] The LSTM model warning module is used to construct an LSTM model, analyze the prediction results at each time step, and perform warning classification based on the prediction results of the LSTM model;

[0055] The data analysis module is used to analyze the physical data collected in the ventilator output pipeline, including the gas matching variation coefficient, the flow waveform difference coefficient, and the air pressure output deviation coefficient, and send them to the fault evaluation module;

[0056] The fault evaluation module is used to construct a fault evaluation model, generate a fault evaluation coefficient, quantitatively evaluate the possibility of a fault, and update the warning level.

[0057] The technical effects and advantages of the present invention:

[0058] The present invention determines the possible fault moments of the ventilator during dynamic adjustment of the gas output volume through the long short-term memory algorithm, and evaluates the possibility of faults in the process of the ventilator dynamically matching the patient's breathing by actually collecting the real-time data in the output pipeline during the operation of the ventilator, which helps to take preventive measures in the initial stage of the fault, prevent the harm caused by the accumulation of faults, and improve the reliability of ventilator fault identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] For the convenience of those skilled in the art to understand, the present invention will be further described below with reference to the accompanying drawings;

[0060] Figure 1 It is a flowchart of a big data-based ventilator fault prediction method of the present invention;

[0061] Figure 2Schematic structural diagram of a ventilator fault prediction system based on big data according to the present invention. Detailed implementation manners

[0062] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0063] Embodiment 1

[0064] Figure 1 A flowchart of a ventilator fault prediction method based on big data according to the present invention is given, which specifically includes the following steps:

[0065] S1: Use the long short-term memory algorithm to predict the process of the ventilator dynamically adjusting the gas volume, and determine to detect abnormal conditions in advance when the ventilator may have a fault;

[0066] S2: Obtain the prediction results at different time steps through the output results of the LSTM model, analyze the prediction results, evaluate the possible abnormal conditions, perform early warning grading, and determine whether to detect the output pipeline during the operation of the ventilator;

[0067] S3: Collect the flow information and air pressure information in the output pipeline during the operation of the ventilator, construct a fault evaluation model based on the comprehensive analysis of the flow information and air pressure information, and quantify the result of the possibility of the ventilator having a fault;

[0068] S4: Obtain the fault evaluation coefficient for the ventilator to dynamically match the patient's breathing process according to the result of the fault evaluation model, and compare it with the fault evaluation coefficient threshold to reconfirm the early warning level of the ventilator.

[0069] The ventilator maintains the same frequency as the patient's inhalation by dynamically adjusting the exhaled air. Therefore, there may be a certain deviation during each dynamic adjustment process, which may cause the deviation to gradually accumulate. Through the long short-term memory algorithm, the deviation of the dynamic adjustment can be predicted by analyzing the time series.

[0070] Determine the input variables and output variables of the long short-term memory algorithm. Among them, the input variables of the long short-term memory algorithm are the residual oxygen concentration, the residual carbon dioxide concentration, the gas exchange rate, and the tidal volume deviation, and the output variable is the adjustment deviation, that is, the difference between the total gas volume output from the output pipeline of the ventilator and the tidal volume inhaled by the patient after each dynamic matching of the ventilator;

[0071] The acquisition logic of the residual oxygen concentration is as follows: The residual oxygen concentration is obtained by subtracting the oxygen concentration remaining after the patient inhales the gas from the oxygen concentration output in the ventilator output pipeline. Among them, the oxygen concentration remaining after the patient inhales the gas is determined by the minimum value of the oxygen concentration when the gas is output twice adjacent to each other in the ventilator output pipeline. The residual oxygen concentration is marked as: , where n = 1, 2, 3, ……, N, N is a positive integer, and n is the time step of the residual oxygen concentration in the long short-term memory algorithm.

[0072] The acquisition logic of the residual carbon dioxide concentration is as follows: The residual carbon dioxide concentration is obtained by subtracting the carbon dioxide concentration remaining after the patient inhales the gas from the carbon dioxide concentration output in the ventilator output pipeline. Among them, the carbon dioxide concentration remaining after the patient inhales the gas is determined by the minimum value of the carbon dioxide concentration when the gas is output twice adjacent to each other in the ventilator output pipeline. The residual carbon dioxide concentration is marked as: , where n = 1, 2, 3, ……, N, N is a positive integer, and n is the time step of the residual carbon dioxide concentration in the long short-term memory algorithm.

[0073] The acquisition logic of the gas exchange rate is as follows: Obtain the total gas concentration output in the ventilator output pipeline during the monitoring time period . Mark the total gas concentration output in the ventilator output pipeline during the monitoring time period as: . Obtain the total gas concentration output in the patient output pipeline during the monitoring time period . Mark the total gas concentration output in the patient output pipeline during the monitoring time period as: . The calculation formula for the average gas exchange rate is: , where i = 1, 2, 3, ……, I, I is a positive integer, i is the number for collecting the total gas concentration output in the ventilator output pipeline and the total gas concentration output in the patient output pipeline during the monitoring time period , and i is the time step of the average gas exchange rate in the long short-term memory algorithm.

[0074] It should be noted that the monitoring time period is a specific time length, which is used to reflect the dynamic adjustment process between the ventilator performance and the patient's breathing condition within a relatively long time range, help identify long-term patterns and potential faults, provide a more comprehensive analysis perspective, and the monitoring time period is set by the staff in the professional field.

[0075] The acquisition logic of the tidal volume deviation is as follows: Obtain the inspiratory tidal volume of the patient during the monitoring time . Obtain the expiratory tidal volume of the patient during the monitoring time . During the monitoring time The inspiratory tidal volume of the patient and the monitoring time The ratio of the expiratory tidal volume of the patient to the inspiratory tidal volume is used as the tidal volume deviation, and the tidal volume deviation is marked as: , where m = 1, 2, 3, ……, M, M is a positive integer, and M is the time step of the tidal volume deviation in the long short-term memory algorithm.

[0076] It should be noted that the monitoring time and the monitoring time can be the same, which is specifically set by the staff in the professional field.

[0077] The greater the residual oxygen concentration and the residual carbon dioxide concentration, it indicates that each time the ventilator outputs gas, the patient may not fully absorb it, and there is a mismatch between the patient's breathing demand and the device settings. When the tidal volume deviation and the gas exchange rate are 1, it indicates that the dynamic regulation of the ventilator is better. When the tidal volume deviation and the gas exchange rate are less than 1 or greater than 1, it indicates that there may be phenomena such as air leakage in the ventilator or there are certain faults.

[0078] The adjustment deviation is expressed as: , where k = 1, 2, 3, ……, K, K is a positive integer, , that is, the time step of the adjustment deviation is consistent with the output time series generated by the smallest time step in the input variables, indicating the adjustment deviation at the Kth time step distance from the current moment.

[0079] After normalization, the training sample set is obtained, that is ; among them, determine the smallest time step in the input variables, and replace all the time steps of the input variables with the smallest time step, and fill the time series with inconsistent smallest time steps by means of backward filling, forward filling or linear interpolation method, etc.

[0080] Construct an LSTM model, use LSTM units to process time series data, each LSTM unit contains a memory unit and a gating mechanism, which are used to process the time dependence in the data, and transmit the output of the LSTM to the fully connected layer for further processing and prediction.

[0081] Define the loss function, use the mean square error to determine the difference between the prediction result and the actual value, and the expression of the mean square error is: ; among them, is the actual adjustment deviation;

[0082] The acquisition logic of the actual adjustment deviation is: determine the total amount of gas output from the ventilator output pipeline by using monitoring devices such as sensors, and determine the inspiratory tidal volume of the patient by using monitoring devices such as sensors, and subtract the inspiratory tidal volume of the patient from the total amount of gas output from the ventilator output pipeline as the actual adjustment deviation.

[0083] According to the output result of the LSTM model, the predicted adjustment deviation is obtained. Based on the adjustment deviations at different time steps, the average value and standard deviation of the adjustment deviation are calculated, and the average value and standard deviation of the adjustment deviation are respectively marked as: and , where , ;

[0084] Set the average value threshold of the adjustment deviation and the standard deviation threshold of the adjustment deviation. Compare the average value and standard deviation of the adjustment deviation with the average value threshold and the standard deviation threshold of the adjustment deviation to generate the following situations:

[0085] If the average value threshold of the adjustment deviation is greater than the average value of the adjustment deviation, a first warning signal is generated;

[0086] If the average value threshold of the adjustment deviation is less than the average value of the adjustment deviation and the standard deviation of the adjustment deviation is greater than the standard deviation threshold of the adjustment deviation, a second warning signal is generated;

[0087] If the average value threshold of the adjustment deviation is less than the average value of the adjustment deviation and the standard deviation of the adjustment deviation is less than the standard deviation threshold of the adjustment deviation, no warning signal is generated.

[0088] When the ventilator generates a first warning signal, the staff is immediately notified to stop using the ventilator and transfer the patient's ventilator;

[0089] When the ventilator generates a second warning signal, a fault assessment model is constructed. By collecting real-time data, a fault assessment coefficient is generated to make a more specific judgment on the ventilator.

[0090] It should be noted that when the first warning signal is generated, it indicates that an abnormality has occurred in the dynamic adjustment of the gas by the ventilator, and the adjustment of the ventilator cannot effectively meet the patient's breathing needs, which may lead to insufficient or excessive oxygen supply. When the second warning signal is generated, it indicates that a more complex or potential fault may exist in the ventilator, but it is also possible that the patient occasionally has abnormal breathing needs and suddenly breathes more rapidly, which may lead to a mismatch between the adjustment amount and the actual demand, thus triggering the warning signal.

[0091] Determine the moment when the second warning signal is generated, collect the flow information and air pressure information in the output pipeline during the operation of the ventilator. Quantify the flow information through the gas matching variation coefficient and the flow waveform difference coefficient, and quantify the air pressure information through the air pressure output deviation coefficient.

[0092] The acquisition logic of the gas matching variation coefficient is as follows: Obtain the total gas concentration output in the ventilator output pipeline at the current time point, and mark the total gas concentration output in the ventilator output pipeline at the current time point as: Obtain the total gas concentration output in the patient output pipeline at the current time point, and mark the total gas concentration output in the patient output pipeline at the current time point as: , where g = 1, 2, 3, ……, C, C is a positive integer, and g is the number of the patient's respiratory cycle within the monitoring time period ;

[0093] It should be noted that within the monitoring time period include multiple breaths of the patient. Each inhalation of the patient corresponds to each gas output of the ventilator. Therefore, and may not be at the same time point, but and correspond to each breath of the patient. The monitoring time period is set by the staff in the professional field.

[0094] Obtain the total gas concentration output in the ventilator output pipeline and the total gas concentration output in the patient output pipeline at the previous time point, and mark them as: and ;

[0095] Calculate the first-order difference of the total gas concentration output in the ventilator output pipeline. The calculation formula is: ; where is the first-order difference of the total gas concentration output in the ventilator output pipeline;

[0096] Calculate the first-order difference of the total gas concentration output in the patient output pipeline. The calculation formula is: ; where is the first-order difference of the total gas concentration output in the patient output pipeline;

[0097] Calculate the first-order difference of the dynamically matched gas. The calculation formula is: ; where is the first-order difference of the dynamically matched gas;

[0098] Calculate the second-order difference of the total gas concentration output in the ventilator output pipeline. The calculation formula is: ; where is the second-order difference of the total gas concentration output in the ventilator output pipeline;

[0099] Calculate the second-order difference of the total gas concentration output in the patient output pipeline. The calculation formula is: ; where The second-order difference of the total gas concentration output in the output pipeline for the patient;

[0100] Calculate the second-order difference of the dynamically matched gas, and the calculation formula is: ; where is the second-order difference of the dynamically matched gas;

[0101] Calculate the mean value of the second-order difference of the dynamically matched gas within the monitoring time period The calculation formula is: ; where is the mean value of the second-order difference of the dynamically matched gas within the monitoring time period ;

[0102] Calculate the standard deviation of the second-order difference of the dynamically matched gas within the monitoring time period The calculation formula is: ;

[0103] Calculate the gas matching coefficient of variation, and the calculation formula is: ; where is the gas matching coefficient of variation.

[0104] As can be seen from the formula, the larger the gas matching coefficient of variation, the worse the breathing effect of the ventilator in dynamically matching the patient.

[0105] The acquisition logic of the flow waveform difference coefficient is as follows: Through the monitoring device in the output pipeline of the ventilator, obtain the flow rate of the gas output from the output pipe of the ventilator within the monitoring time period , and mark the flow rate of the gas output from the output pipe of the ventilator within the monitoring time period as: , where b = 1, 2, 3,..., B, B is a positive integer, and b is the number of the flow rate of the collected gas within the monitoring time period ;

[0106] Generate the flow rate waveform of the output gas according to the flow rate data of the gas output from the output pipe of the ventilator within the monitoring time period , and integrate the flow rate waveform of the output gas within the monitoring time period ;

[0107] Obtain the gas reference flow rate waveform of the gas output from the output pipe of the ventilator, and integrate the gas reference flow rate waveform of the gas output from the output pipe of the ventilator within the monitoring time period ;

[0108] Calculate the flow waveform difference coefficient, and the calculation formula is: ; where ~ are the start time and end time of the monitoring time period , For the monitoring time period Generate a flow rate waveform of the gas output from the ventilator output pipe within, which is the reference flow rate waveform of the gas output from the ventilator output pipe.

[0109] As can be seen from the formula, the greater the flow rate waveform difference coefficient, the more significant the waveform difference, which may indicate problems with the air flow regulation mechanism of the device or other faults in the device, and there may be other reasons as follows:

[0110] Air flow inconsistency: The ventilator is designed to provide a stable and regular air flow to match the patient's breathing needs. If there is a significant difference between the actual air flow waveform (real-time waveform) and the normal air flow waveform (reference waveform), it may mean that the air flow regulation mechanism of the ventilator cannot work properly. This inconsistency may cause the patient's breathing discomfort and affect the treatment effect;

[0111] Device adjustment problem: A large waveform difference may reflect problems with the adjustment mechanism inside the device. For example, parameters such as the pressure, flow rate, and oxygen concentration of the air flow cannot be adjusted as expected. Maladjustment of the device may lead to unstable or abnormal air flow, thus affecting the patient's breathing comfort and treatment effect;

[0112] Patient state change: The patient's breathing state may change, such as changes in breathing frequency and tidal volume. If the difference between the waveform of the ventilator and the reference waveform increases, it may reflect that the change in the patient's state is not effectively responded to and adjusted by the ventilator. This may be due to the fact that the device adjustment mechanism cannot adapt to the change in the patient's state in real time, or the device cannot provide enough adjustment range to meet the patient's needs;

[0113] Airway obstruction or leakage: Obstruction or leakage in the airway may also cause an increase in the waveform difference. Obstruction will cause poor air flow, resulting in a difference between the actual air flow and the expected air flow, while leakage may cause the device to be unable to effectively provide the predetermined gas volume, thus triggering a significant waveform difference.

[0114] The acquisition logic of the air pressure output deviation coefficient is as follows: Obtain the pressure of the gas in the ventilator output pipeline during each breathing cycle of the patient within the monitoring time period and obtain the average value of the pressure in the ventilator output pipeline during each breathing cycle of the patient. Mark the average value of the pressure in the ventilator output pipeline during each breathing cycle of the patient as: Obtain the standard deviation of the average value of the pressure in the ventilator output pipeline during each breathing cycle of the patient, and mark the standard deviation of the average value of the pressure in the ventilator output pipeline during each breathing cycle of the patient as: ; where , ;

[0115] Obtain the monitoring time period The peak pressure of the gas in the ventilator output pipeline during each breathing cycle of the patient within the monitoring time period, and mark the peak pressure of the gas in the ventilator output pipeline during each breathing cycle of the patient as: ;

[0116] Set the pressure range threshold of the gas in the ventilator output pipeline, and mark the pressure range threshold as: ;

[0117] Calculate the air pressure output deviation coefficient, and the calculation formula is:

[0118] ;

[0119] Among them, is the air pressure output deviation coefficient.

[0120] It can be seen from the formula that the larger the air pressure output deviation coefficient, the more unstable the pressure of the gas in the ventilator output pipeline may be. Specifically, it may include:

[0121] There are problems with the airflow regulation of the ventilator. For example, too high a pressure peak may cause discomfort or injury to the patient, while too low a pressure peak may indicate insufficient gas supply, affecting the treatment effect;

[0122] There is a significant change in the average pressure, and the ventilator has problems in maintaining a stable airflow pressure, affecting the patient's breathing comfort and treatment effect;

[0123] A significant change in the peak and average values of the airway pressure may indicate a failure of the internal components of the device, such as problems may occur with components such as pressure sensors and airflow regulating valves.

[0124] Comprehensively analyze the flow information and air pressure information, conduct a weighted analysis of the gas matching variation coefficient, flow waveform difference coefficient, and air pressure output deviation coefficient, construct a fault assessment model, and generate a fault assessment coefficient. The expression of the fault assessment coefficient is: ; Among them, is the fault assessment coefficient, , , are the proportionality coefficients of the gas matching variation coefficient, flow waveform difference coefficient, and air pressure output deviation coefficient respectively, , , are all greater than 0.

[0125] Set a threshold for the fault evaluation coefficient, compare the fault evaluation coefficient with the threshold. If the fault evaluation coefficient is greater than the threshold, change the second warning signal of the ventilator to the first warning signal. If the fault evaluation coefficient is less than the threshold, continue the above process until the second warning signal disappears or is changed to the first warning signal.

[0126] The present invention determines the possible fault time of the ventilator during dynamic adjustment of gas output through the long short-term memory algorithm, and evaluates the probability of faults in the process of the ventilator dynamically matching the patient's breathing by actually collecting the real-time data in the output pipeline during the operation of the ventilator, which helps to take preventive measures in the initial stage of the fault, prevent the harm caused by the accumulation of faults, and improve the reliability of ventilator fault identification.

[0127] Embodiment 2

[0128] As Figure 2 shows a schematic structural diagram of a method for predicting ventilator faults based on big data according to the present invention, including a data acquisition module, an LSTM model warning module, a data analysis module, and a fault evaluation module;

[0129] The data acquisition module is used to collect the input variables required by the LSTM model warning module, including the residual oxygen concentration, the residual carbon dioxide concentration, the gas exchange rate, and the tidal volume deviation;

[0130] The LSTM model warning module is used to construct an LSTM model, analyze the prediction results at each time step, and perform warning classification based on the prediction results of the LSTM model;

[0131] The data analysis module is used to analyze the physical data collected in the ventilator output pipeline, including the gas matching coefficient of variation, the flow waveform difference coefficient, and the air pressure output deviation coefficient, and send them to the fault evaluation module;

[0132] The fault evaluation module is used to construct a fault evaluation model, generate a fault evaluation coefficient, quantitatively evaluate the probability of faults, and update the warning level.

[0133] The above formulas are all dimensionless and take their numerical calculations. The formulas are obtained by software simulation of a large amount of collected data to get a formula closest to the real situation. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0134] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center in a wired or wireless (such as infrared, wireless, microwave, etc.) manner. The computer-readable storage medium can be any available medium that can be accessed by a computer, or a data storage device such as a server or data center that includes one or more collections of available media. The available media can be magnetic media (such as floppy disks, hard disks, magnetic tapes), optical media (such as DVDs), or semiconductor media. The semiconductor media can be a solid-state drive.

[0135] It should be understood that in various embodiments of the present application, the order numbers of the above processes do not indicate the order of execution, and the order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0136] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. A professional technician can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present application.

[0137] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments, and will not be described herein again.

[0138] In several embodiments provided by this application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection to each other can be through some interfaces. The indirect coupling or communication connection of the device or unit can be in electrical, mechanical, or other forms.

[0139] If the above functions are implemented in the form of software function units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the essence of the technical solution of this application, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of this application. The foregoing storage media include: USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs, and other media that can store program codes.

[0140] The above is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed by this application can easily think of changes or substitutions, which should all be covered by the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

Claims

1. A ventilator failure prediction method based on big data, characterized in that: The specific steps include: S1: Use the long short-term memory algorithm to predict the process of the ventilator dynamically adjusting the gas volume, and determine the abnormal situation in advance when the ventilator may fail; S2: Obtain prediction results at different time steps through the output results of the LSTM model, analyze the prediction results, evaluate possible abnormal conditions, perform early warning classification, and determine whether to detect the output pipeline when the ventilator is running; S3: Collect the flow information and air pressure information in the output pipe when the ventilator is running, build a fault assessment model based on the comprehensive analysis of the flow information and air pressure information, and quantify the possibility of ventilator failure; S4: According to the results of the fault assessment model, the fault assessment coefficient of the ventilator dynamically matching the patient's breathing process is obtained, and compared with the fault assessment coefficient threshold, and the warning level of the ventilator is reconfirmed; The long short-term memory algorithm predicts the process of the ventilator dynamically adjusting the gas volume, including: Determine the input variables and output variables of the long short-term memory algorithm, wherein the input variables of the long short-term memory algorithm are residual oxygen concentration, residual carbon dioxide concentration, gas exchange rate, and tidal volume deviation, and the output variable is the adjustment deviation, that is, the difference between the total amount of gas output from the ventilator output pipe and the patient's inhaled tidal volume after each dynamic matching of the ventilator; Keep the time step of adjusting the deviation consistent with the output time series generated by the smallest time step of the input variables; Build an LSTM model and use LSTM units to process time series data. Each LSTM unit contains a memory unit and a gating mechanism, and passes the output of the LSTM to the fully connected layer. According to the output results of the LSTM model, the predicted adjustment deviation is obtained, and the mean and standard deviation of the adjustment deviation are calculated according to the adjustment deviation of different time steps; Carry out early warning classification, including: Set the mean value threshold of the adjustment deviation and the standard deviation threshold of the adjustment deviation, compare the mean value and standard deviation of the adjustment deviation with the mean value threshold of the adjustment deviation and the standard deviation threshold of the adjustment deviation, and generate the following situations: If the average value threshold of the adjustment deviation is greater than the average value of the adjustment deviation, a first warning signal is generated; If the average value threshold of the adjustment deviation is less than the average value of the adjustment deviation and the standard deviation of the adjustment deviation is greater than the standard deviation threshold of the adjustment deviation, a second warning signal is generated; If the mean value threshold of the adjustment deviation is less than the mean value of the adjustment deviation and the standard deviation of the adjustment deviation is less than the standard deviation threshold of the adjustment deviation, no warning signal is generated; Collect flow information in the output pipe when the ventilator is running, including: Determine the time to generate the second warning signal, collect flow information and air pressure information in the output pipeline when the ventilator is running, quantify the flow information through the gas matching variation coefficient and the flow waveform difference coefficient, and quantify the air pressure information through the air pressure output deviation coefficient; The logic for obtaining the gas matching coefficient of variation is: The total gas concentration output from the ventilator output pipeline at the current time point is obtained, and the total gas concentration output from the ventilator output pipeline at the current time point is marked as: , obtain the total gas concentration output from the patient output pipeline at the current time point, and mark the total gas concentration output from the patient output pipeline at the current time point as: , where g = 1, 2, 3, ..., C, C is a positive integer greater than 3, and g is the monitoring time period The number of the patient's breathing cycle; The total gas concentration output from the ventilator output pipe and the total gas concentration output from the patient output pipe at the last time point are obtained and marked as: and ; Calculate the first-order difference of the total gas concentration output in the ventilator output pipeline, the calculation formula is: ;in, is the first-order difference of the total gas concentration output in the ventilator output pipe; Calculate the first-order difference of the total gas concentration output in the patient output pipeline, and the calculation formula is: ;in, is the first-order difference of the total gas concentration output in the patient output pipeline; Calculate the first-order difference of the dynamically matched gas using the formula: ;in, The first-order difference of the dynamic matching gas; Calculate the second-order difference of the total gas concentration output in the ventilator output pipeline, and the calculation formula is: ;in, is the second-order difference of the total gas concentration output in the ventilator output pipe; Calculate the second-order difference of the total gas concentration output in the patient output pipeline, and the calculation formula is: ;in, is the second-order difference of the total gas concentration output in the patient output pipeline; Calculate the second-order difference of the dynamically matched gas using the formula: ;in, The second-order difference for the dynamic matching gas; Calculate the second-order difference of the dynamic matching gas during the monitoring period The mean value within is calculated as: ;in, To dynamically match the second-order difference of the gas during the monitoring period The mean within ; Calculate the second-order difference of the dynamic matching gas during the monitoring period The standard deviation within is calculated as: ; Calculate the gas matching coefficient of variation using the formula: ;in, is the gas matching coefficient of variation; The logic for obtaining the flow waveform difference coefficient is: Obtain the monitoring time period through the monitoring device in the ventilator output pipeline The flow rate of the gas output from the output tube of the ventilator and the monitoring time period The flow rate of the output gas from the output tube of the internal ventilator is marked as: , where b=1, 2, 3, ..., B, B is a positive integer, and b is the monitoring time period The number of the flow rate of the gas collected; According to the monitoring time period The output gas flow rate data of the output tube of the internal ventilator generates the output gas flow rate waveform, and the output gas flow rate waveform is monitored during the monitoring period. Integrate within Obtain the gas reference flow rate waveform of the output gas of the ventilator output tube, and the gas reference flow rate waveform of the output gas of the ventilator output tube during the monitoring period Integrate within Calculate the flow waveform difference coefficient, the calculation formula is: ;in, ~ Monitoring time period The start and end time of Monitoring time period The output gas flow rate data of the internal ventilator output tube generates the output gas flow rate waveform. It is the gas reference flow rate waveform of the gas output from the output tube of the ventilator; The logic for obtaining the air pressure output deviation coefficient is: Get monitoring time period The pressure of the gas in the ventilator output pipe during each breathing cycle of the patient is obtained, and the average value of the pressure in the ventilator output pipe during each breathing cycle of the patient is obtained, and the average value of the pressure in the ventilator output pipe during each breathing cycle of the patient is marked as: , obtain the standard deviation of the average pressure in the ventilator output pipe during each breathing cycle of the patient, and mark the standard deviation of the average pressure in the ventilator output pipe during each breathing cycle of the patient as: ;in, , ; Get monitoring time period The peak value of the gas pressure in the ventilator output pipe during each breathing cycle of the patient is marked as: ; Set the pressure range threshold of the gas in the ventilator output pipeline and mark the pressure range threshold as: ; Calculate the air pressure output deviation coefficient using the following formula: ;in, is the pressure output deviation coefficient; Build a fault assessment model, including: The gas matching variation coefficient, flow waveform difference coefficient and pressure output deviation coefficient are weighted and analyzed to construct a fault assessment model and generate a fault assessment coefficient. The expression of the fault assessment coefficient is: ;in, is the fault assessment coefficient, , , They are the proportional coefficients of gas matching variation coefficient, flow waveform difference coefficient, and pressure output deviation coefficient. , , Both are greater than 0.

2. A ventilator failure prediction method based on big data according to claim 1, characterized in that: Reconfirm the ventilator's warning level, including: A fault assessment coefficient threshold is set, and the fault assessment coefficient is compared with the fault assessment coefficient threshold. If the fault assessment coefficient is greater than the fault assessment coefficient threshold, the second warning signal of the ventilator is converted into a first warning signal. If the fault assessment coefficient is less than the fault assessment coefficient threshold, the process of steps S1 to S4 is continued until the second warning signal disappears or the second warning signal is converted into the first warning signal.

3. A ventilator fault prediction system based on big data, used to implement a ventilator fault prediction method based on big data as described in any one of claims 1-2, characterized in that: It includes data acquisition module, LSTM model warning module, data analysis module and fault assessment module; The data acquisition module is used to collect the input variables required by the LSTM model early warning module, including residual oxygen concentration, residual carbon dioxide concentration, gas exchange rate, and tidal volume deviation; LSTM model warning module, which is used to build an LSTM model, analyze the prediction results of each time step, and classify warnings based on the prediction results of the LSTM model; A data analysis module is used to analyze the physical data collected in the ventilator output pipeline, including the gas matching variation coefficient, the flow waveform difference coefficient and the gas pressure output deviation coefficient, and send them to the fault assessment module; The fault assessment module is used to build a fault assessment model, generate fault assessment coefficients, quantitatively assess the possibility of faults, and update the warning level.

Images