CN116500903B

CN116500903B - Breathing machine heating and humidifying control method and system based on artificial intelligence and breathing machine

Info

Publication number: CN116500903B
Application number: CN202310787069.2A
Authority: CN
Inventors: 欧桂康; 朱婷婷; 赵宁
Original assignee: Jiangsu Yuyue Medical Equipment and Supply Co Ltd
Current assignee: Jiangsu Yuyue Medical Equipment and Supply Co Ltd
Priority date: 2023-06-30
Filing date: 2023-06-30
Publication date: 2023-10-31
Anticipated expiration: 2043-06-30
Also published as: CN116500903A

Abstract

The invention discloses a ventilator heating and humidifying control method and system based on artificial intelligence for measuring, processing and controlling nasal information parameters of a ventilator and the ventilator. The method comprises the following steps: a placement system; constructing a neural network model taking an actual output temperature value of the heating plate, an ambient temperature value and a nasal airflow temperature value as input and taking a control action of the controller as output; and deploying the trained, verified, tested and evaluated neural network model into a breathing machine for use. The system comprises a heating plate, a temperature sensor, a fan, a nose information temperature and humidity sensor, an environment temperature and humidity sensor and a controller. The scheme applies the neural network model to compare the nose airflow temperature predicted value and the current nose airflow temperature value and then carries out corresponding judgment and operation, thereby ensuring that the nose person can obtain the optimal temperature and humidity of the breathing machine after adjustment.

Description

Breathing machine heating and humidifying control method and system based on artificial intelligence and breathing machine

Technical Field

The invention relates to the technical field of measurement, processing and control of nose information parameters of a breathing machine, in particular to a breathing machine heating and humidifying control method and system based on artificial intelligence for measuring, processing and controlling the nose information breathing temperature and humidity of the breathing machine and the breathing machine.

Background

The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.

In modern clinical medicine, a respirator is used as an effective means capable of replacing autonomous ventilation by manpower, is widely used for respiratory failure caused by various reasons, anesthesia respiratory management during major surgery, respiratory support treatment and emergency resuscitation, and occupies a very important position in the field of modern medicine. The breathing machine is a vital medical device which can prevent and treat respiratory failure, reduce complications, save and prolong the life of patients.

When the breathing machine is used, in order to increase the breathing comfort of a patient, a heating and humidifying device is usually added at the inhalation end of a patient circuit of the breathing machine to heat and humidify inhaled gas so as to avoid a series of uncomfortable conditions such as dry mouth, nasal cavity and throat of a user, nasal obstruction and the like. The heating and humidifying device of the breathing machine heats water to generate steam so as to achieve the function of heating and humidifying inhaled gas. Therefore, the control of temperature is particularly important in the heating and humidifying process. The traditional PID control method is to adjust the output by calculating the proportion, integral and derivative of the error signal, is a linear system based on feedback control, and uses the traditional PID control method to manually adjust the specific coefficient values of the proportion, integral and derivative of the error signal, has poor adaptability and cannot be well adapted to various uncertainties and complexities, thereby influencing the final heating and humidification of the breathing machine and causing discomfort when the patient uses the breathing machine. In the existing ventilator heating and humidifying control process, an ambient temperature sensor needs to be provided to detect the current ambient temperature. In addition, the existing manufacturer also has a method that a heating pipeline is arranged in the breathing machine, a pipeline temperature sensor is arranged at the tail end of the heating pipeline, the environment temperature sensor detects the environment temperature value of the current environment, and the pipeline temperature sensor detects the pipeline temperature value of air flow in the pipeline and respectively sends the environment temperature value and the pipeline temperature value to the controller; the controller matches the temperature compensation value in the supplementary thermometer according to the nasal information demand temperature value, the airflow speed value and the environment temperature value, and the method firstly needs to use a heating pipeline equipped by a manufacturer and has the limitation of the heating pipeline, and secondly, in the practical application test, even if the same test method is used in different positions of the machine, the obtained result is different.

Therefore, how to solve the problems of complicated adjustment, poor adaptability to various uncertainties and complexities, limitation of heating pipelines, uncertainty of an ambient temperature sensor and the like of the PID control method for heating and humidifying of the traditional breathing machine becomes the problem to be studied and solved by the invention.

Disclosure of Invention

The invention aims to provide a ventilator heating and humidifying control method and system based on artificial intelligence and a ventilator.

In order to achieve the above object, a first aspect of the present invention provides an artificial intelligence based ventilator heating and humidifying control method, in which a ventilator heating and humidifying control system is disposed in the ventilator, the ventilator heating and humidifying control system includes a heating plate, a temperature sensor, a fan, a nose temperature and humidity sensor, an environmental temperature and humidity sensor, and a controller, the artificial intelligence based ventilator heating and humidifying control method includes the following steps:

s100, acquiring an actual output temperature value of a heating plate detected by a temperature sensor, an ambient temperature value detected by a plurality of ambient temperature and humidity sensors and a nose gas flow temperature value detected by a nose gas temperature and humidity sensor by a controller;

s200, constructing a neural network model taking each temperature value acquired by a controller as input and taking the control action of the controller as output, wherein the step of constructing the neural network model comprises the following steps of:

S210, data acquisition, namely determining a sampling frequency, acquiring temperature value data with the sampling frequency, marking each temperature value of each data point, and storing the acquired data into a controller;

s220, preprocessing data, namely collecting all temperature values and corresponding time stamps of all data points in a standard limit as ventilator historical temperature data, and dividing the ventilator historical temperature data into a training set and a verification set;

s230, model selection and training, wherein a model consisting of an input layer, a hidden layer and an output layer is selected and training is performed by using a training set; the input layer is used for receiving the ventilator history temperature data of the training set, and the activation function of the ventilator history temperature data in the input layer is provided with characteristic weights and bias items corresponding to the input temperature values and the ventilator set temperature; the hidden layer is used for processing the data of the input layer and providing signals for the output layer; the output layer generates a predicted value for the temperature of the future nasal airflow and outputs a heating disc temperature control parameter of the current breathing machine; back propagation is carried out on the training set to update training parameters of the characteristic weights and the bias items, and training is carried out again by using the new training parameters;

S240, model verification, namely verifying the trained neural network model by using the historical temperature data of the breathing machine in the verification set, and comparing the trained neural network model by using a verification range value;

s250, model testing and evaluation, wherein real-time testing is utilized to evaluate whether the difference value between the obtained nasal airflow temperature value of the actual air outlet of the breathing machine and the set temperature of the breathing machine is within an evaluation range value or not after the control action obtained by the neural network model is executed;

and S300, deploying the constructed neural network model into a breathing machine for use, acquiring an environmental temperature value under the use environment, comparing the nasal airflow temperature predicted value obtained by the output layer by the neural network model, and judging whether the difference value between the nasal airflow temperature predicted value and the current nasal airflow temperature value is within a use range value, if not, re-executing the acquisition of the environmental temperature value under the use environment, and if so, executing the control of the breathing machine by using the output heating plate temperature control parameter.

The invention provides a ventilator heating and humidifying control system based on artificial intelligence, which is used for ventilator heating and humidifying control in the method of the first aspect and is characterized by comprising a heating plate, a temperature sensor, a fan, a nose information temperature and humidity sensor, an environment temperature and humidity sensor and a controller; wherein,,

A temperature sensor is arranged at a heating disc of the breathing machine, a nose temperature and humidity sensor is arranged at an air outlet of a fan of the breathing machine, and at least two environmental temperature and humidity sensors are arranged around the breathing machine;

the controller is electrically connected with the heating plate, the temperature sensor, the fan, the nose information temperature and humidity sensor and the environment temperature and humidity sensor, and the controller obtains the actual output temperature value of the heating plate detected by the temperature sensor, the environment temperature values detected by the environment temperature and humidity sensors and the nose information air flow temperature value detected by the nose information temperature and humidity sensor; the controller comprises a neural network model taking an actual output temperature value of the heating plate, an ambient temperature value and a nasal airflow temperature value as inputs and taking control actions of the controller as outputs, and the controller obtains the temperature control parameters of the heating plate through the neural network model and then transmits signals to the heating plate so that the heating plate can carry out temperature adjustment by the temperature control parameters of the heating plate, and the output nasal airflow temperature value corresponds to the set temperature of the breathing machine.

According to a third aspect of the invention, a ventilator is provided, and the ventilator comprises the ventilator heating and humidifying control system based on artificial intelligence according to the second aspect of the invention.

The content of the present invention is explained as follows:

1. through implementation of the technical scheme, the main comprehensive consideration is that the control accuracy and the response speed can be improved by controlling the heating plate of the breathing machine through the neural network model, and meanwhile, the requirement of manual adjustment is reduced, so that the automation degree and the safety of a breathing machine system are improved. The refining and optimizing comprises the following steps: the method comprises the steps of arranging a heating and humidifying control system of the breathing machine so as to simplify the structure and reduce the cost, so that the acquired temperature value of the nasal airflow is close to the actual inhalation of a user, and the reliability of environmental temperature acquisition is improved to the maximum extent; the method comprises the steps of constructing the neural network model, wherein the sampling frequency is determined according to the response speed of a breathing machine system and the change range of the gas temperature, so that the sampled data can reflect the real gas temperature change, reasonable data preprocessing is performed, the current environment temperature value data subjected to the data preprocessing has training value and representativeness, the model parameters are updated by back propagation on a training set, the control error is reduced to the maximum extent, the production of the future nasal airflow temperature predicted value is more accurate, the training set data subjected to the data preprocessing is adopted to verify the trained neural network model, various data acquired and preprocessed in the preamble step can be effectively utilized, the quality of the neural network model can be further improved by testing and evaluating the model by utilizing real-time testing, and the better effect is achieved; the method also comprises the step of comparing the nose airflow temperature predicted value with the current nose airflow temperature value by the neural network model during actual application, and then carrying out corresponding judgment and operation, so as to ensure that the nose person can obtain the optimal temperature and humidity of the breathing machine after adjustment.

2. In the first aspect of the foregoing technical solution, in step S100, the method includes disposing a temperature sensor at a heating plate of a ventilator, disposing a nose information temperature and humidity sensor at an air outlet of a ventilator fan, and disposing at least two environmental temperature and humidity sensors around the ventilator.

3. In the first aspect of the foregoing technical solution, in the step S210 of data acquisition, the method includes the steps of:

s211, setting a sampling frequency: the sampling frequency needs to be determined according to the response speed of the breathing machine system and the change range of the gas temperature so as to ensure that sampled data can reflect the real gas temperature change;

s212, improving data quality: removing abnormal values and noise from the acquired data, adopting smoothing filtering to process,where k represents the kth sampling point and N represents the total number of times that sampling is required;

s213, labeling: correctly labeling the set temperature of the breathing machine, the actual output temperature value of the heating disc and the nasal airflow temperature value of each data point, and facilitating subsequent model training and evaluation;

s214, data storage: the collected data is stored in a memory element of the controller for subsequent data analysis and mining.

When the steps are executed, the sampling frequency is determined according to the response speed of the breathing machine system and the change range of the gas temperature, so that the sampled data can reflect the real gas temperature change, the built neural network model is more accurate and intelligent, and the deployment application of the follow-up neural network model can also be matched with the real gas temperature change to carry out heating and humidifying control.

4. In the first aspect of the foregoing technical solution, in the step of S220 data preprocessing, an average value and a standard deviation of an ambient temperature value data set collected by a ventilator are calculated, a specification limit of normal data of the ventilator is determined, each individual ambient temperature value data is compared with the specification limit, and if the current ambient temperature value data is within the specification limit, the ambient temperature value data is retained; if the current environmental temperature value data is not within the standard limit, deleting the environmental temperature value data, the set temperature of the breathing machine under the data point and the actual output temperature value of the heating disc. The average value represents the central position of the temperature of the breathing machine in the measuring process, the standard deviation reflects the temperature variability of the breathing machine in the measuring process, so that the current environmental temperature value data subjected to data preprocessing has training value and representativeness, and the obtained historical temperature data of the breathing machine can be used for constructing the neural network model as data which are close to the actual environmental temperature value, the set temperature of the breathing machine, the actual output temperature value of the heating disc, the nasal airflow temperature value and the like, thereby improving the construction quality of the neural network model, training efficiency and accuracy, and enabling the neural network model to be well adapted to various uncertainties and complexity.

5. In the first aspect of the foregoing technical solution, in the step S220 of data preprocessing, the method includes the steps of:

s221, acquiring data of an environmental temperature value acquired by data, a set temperature of a breathing machine, an actual output temperature value of a heating disc and a nasal airflow temperature value;

s222, calculating the average value and standard deviation of the environmental temperature value, and using these data, calculating the average value and standard deviation of the temperature. And (3) injection: the average value represents the center position of the ventilator temperature of the measurement process, while the standard deviation reflects the ventilator temperature variability of the measurement process;

s223, determining a specification limit: adding three times of standard deviation to the average value of the ambient temperature values acquired by the breathing machine to obtain an upper standard limit, and subtracting three times of standard deviation from the average value of the ambient temperature values acquired by the breathing machine to obtain a lower standard limit; the upper and lower specification limits represent ranges in which the measurement of the ambient temperature value should fall under normal operating conditions of the ventilator;

s224, comparing the results: comparing each individual temperature measurement to a specification limit, the result being considered acceptable and normal if the result is within the specification limit;

s225, collecting the collected environmental temperature value, the ventilator set temperature, the actual output temperature value of the heating disc, the nasal airflow temperature value and the corresponding time stamp of each data point within the standard limit as ventilator historical temperature data, and dividing the ventilator historical temperature data into a training set and a verification set.

6. In the first aspect of the foregoing technical solution, in the step of S220 data preprocessing, when dividing the ventilator history temperature data into a training set and a verification set, a quantity ratio of the ventilator history temperature data included in the training set and the verification set is: 6/4 to 8/2, more preferably 7/3.

7. In the first aspect of the foregoing technical solution, in the step of selecting and training the model S230, for each temperature data in the acquired training set, an activation value corresponding to an activation function of an input layer is set, an output layer temperature error generated by the output layer is further calculated according to a forward propagation manner, then a reverse propagation error is calculated according to the output layer temperature error, a new training parameter related to a feature weight and an offset term is obtained according to the reverse propagation error, and the new training parameter is used as a new feature weight and an offset term of an environmental temperature value, a ventilator set temperature, an actual output temperature value of a heating disc, and a nasal airflow temperature value in the input layer to perform training. By adopting the mode to train the neural network model, the model parameters are updated by back propagation on the training set, so that the control error is reduced to the maximum extent, the production of the future nose information air flow temperature predicted value is more accurate, the output of the heating plate temperature control parameters and the temperature control of the heating plate are more in line with the expectations, and thus the nose information temperature required by a nose information person can be controlled more accurately.

8. In the first aspect of the foregoing solution, in the step of selecting and training the model in S230, the step of performing back propagation on the training set to update the training parameters includes the following steps:

receiving data of the input layer at the hidden layer and calculating neuron input data,wherein->Input representing the j-th neuron of the g-th layer,>characteristic weights representing that the kth neuron of the g-th layer is connected to the jth neuron of the g-th layer,/->Represents the output of the kth neuron of the g-1 th layer,/o>A bias term representing a jth neuron of a g-th layer;

the hidden layer calculates the neuron output data,wherein->Representing the output of the jth neuron of the g-th layer, σ representing the activation function;

for each ventilator historical temperature data x in the training set, an input layer pair is setActivation value a of the response ^g The forward propagation formula is obtained: z ^g ＝ω ^g +a ^g-1 +b ^g ，a ^g ＝σ(z ^g ) Wherein g represents the number of layers, z represents the linear operation result of omega, a and b, omega represents the characteristic weight, a represents the activation value of z, and b represents the bias term;

further calculating an output layer temperature error value delta generated by the output layer ^G ，Wherein the formula is a differential operator expression of the output error formula, delta ^G Indicating the output layer temperature error value,/->Is->C represents a cost function, G represents the number of layers of the neural network, j represents the jth neuron;

From output layer temperature error delta ^G Obtaining a counter-propagating temperature error value delta ^g ，δ ^g ＝((ω ^g+1 ) ^T δ ^g+1 )·σ’(z ^g ) Wherein delta ^g Indicating the counter-propagating temperature error value, ω ^g+1 Neuron characteristic weight, delta, representing the g+1 layer ^g+1 A back propagation temperature error value representing the previous layer;

obtaining new training parameters related to characteristic weights and bias terms from the back propagation error, wherein the new training parameters of the characteristic weights areWherein the new training parameters of the bias term areWherein η represents the learning rate, m represents the sample size, x represents the summation variable, δ ^x，g A represents the difference between the output layer temperature error and the current temperature error ^x，g-1 Representing activationThe value, T, represents the sample set.

9. In the first aspect of the above technical solution, in the step of verifying the model S240, after verifying that the neural network model inputs the ambient temperature value and the nasal airflow temperature value in the verification set of the previous time node, the nasal airflow temperature predicted value output by the output layer and the difference value between the actual output temperature value of the heating disc and the nasal airflow temperature value and the actual output temperature value of the heating disc in the verification set of the next time node after the adjustment of the temperature control parameter of the heating disc of the current breathing machine are determined to be within the verification range value, if yes, the neural network model is determined to be successfully trained, and the next step is executed; if not, judging that the neural network model training fails, and retraining after adjusting the feature weights and the bias items in the step S230. The training set data after data preprocessing is adopted to verify the trained neural network model, the data based on the real change, such as the ambient temperature value, the breathing machine set temperature, the heating plate actual output temperature value, the nasal airflow temperature value and the like, which are acquired and preprocessed in the preamble step can be effectively utilized, the model verification is carried out at the moment, if the training of the neural network model is judged to be failed, only the feature weight and the bias item are required to be optimized in the training, the error due to human factors can be reduced, the accurate control is realized, the training amount and the time length are reduced for the training of the neural network model, and the neural network model can be trained well more quickly under the condition of guaranteeing the quality of the neural network model.

10. In the first aspect of the above technical solution, in the step of S250 model testing and evaluating, the method includes the steps of:

s251, transplanting the trained neural network model into a heating and humidifying control system of the breathing machine;

s252, sending the set temperature of the breathing machine to a heating and humidifying control system of the breathing machine;

s253, acquiring environmental temperature values detected by a plurality of environmental temperature and humidity sensors, and giving the environmental temperature values to a neural network model, and calculating to obtain heating plate temperature control parameters required to be newly issued by the heating plate;

s254, issuing a heating disc temperature control parameter to the heating disc, and acquiring a nose airflow temperature value of the actual air outlet of the breathing machine through a standard thermometer;

s255, judging whether the difference value between the nose gas flow temperature value of the actual gas outlet of the breathing machine and the set temperature of the breathing machine is in an evaluation range value, if so, judging that the neural network model passes the evaluation, and deploying in the step S300; if not, it is determined that the neural network model evaluation fails, and steps S100 and S200 are re-executed or step S230 is re-executed.

11. In the first aspect of the foregoing technical solution, in step S300, the using step includes:

s310, setting a nose-resting person to a heating and humidifying control system of the breathing machine according to the required temperature;

S320, acquiring environmental temperature values detected by a plurality of environmental temperature and humidity sensors by a controller, inputting the environmental temperature values into a neural network model, outputting a nose information airflow temperature predicted value and a heating plate temperature control parameter of a breathing machine, and acquiring a current nose information airflow temperature value detected by the nose information temperature and humidity sensors;

s330, comparing the nose gas flow temperature predicted value and the difference value of the current nose gas flow temperature value by a neural network model to determine whether the difference value is within a use range value, and if not, re-executing the step S320; if yes, executing the next step;

and S340, issuing a temperature control parameter of the heating plate to the heating plate so as to control the temperature and the humidity of the breathing machine.

12. In the first aspect of the above technical solution, in step S300, the collected environmental temperature value data is compared with a specification limit, and if the current environmental temperature value data is within the specification limit, the actual output temperature value of the hot plate, the environmental temperature value, and the nasal airflow temperature value under the current timestamp are used as inputs of the neural network model; and if the current environmental temperature value data is not in the standard limit, continuing to acquire the environmental temperature value data at the sampling frequency until the acquired environmental temperature value data is in the standard limit.

13. In the first aspect of the above-described technical means, the verification range value, the evaluation range value, and the use range value are each set to a value of + -0.1 ℃ to + -0.3 ℃, and more preferably + -0.2 ℃.

14. In the second aspect of the foregoing technical solution, the number of the environmental temperature and humidity sensors is four, the four environmental temperature and humidity sensors are distributed around the ventilator, and the data acquired by the four environmental temperature and humidity sensors calculate the current representative environmental temperature value according to an average weighting formula. In the scheme, the conventional heating pipeline can be removed by placing the nose information temperature and humidity sensor at the air outlet, so that the structure is simplified, the cost is reduced, and the acquired nose information air flow temperature value is close to the actual inhalation temperature of a user; by arranging a plurality of ambient temperature and humidity sensors around the breathing machine, the reliability of ambient temperature acquisition can be improved to the greatest extent, so that inaccuracy of results caused by damage of a single ambient temperature and humidity sensor or the fact that a certain sensor is close to a heat source is reduced to the greatest extent.

Due to the application of the scheme, compared with the prior art, the invention has the following advantages and effects:

through implementation of the technical scheme, the inventor performs comprehensive consideration in a large direction and performs further refinement and optimization in a small aspect in the research and development process of the scheme. The main comprehensive consideration is that the control precision and the response speed can be improved by controlling the heating plate of the breathing machine through the neural network model, and meanwhile, the requirement of manual adjustment is reduced, so that the automation degree and the safety of a breathing machine system are improved. The refining and optimizing comprises the following steps: the method comprises the steps of arranging a heating and humidifying control system of the breathing machine so as to simplify the structure and reduce the cost, so that the acquired temperature value of the nasal airflow is close to the actual inhalation of a user, and the reliability of environmental temperature acquisition is improved to the maximum extent; the method comprises the steps of constructing the neural network model, wherein the sampling frequency is determined according to the response speed of a breathing machine system and the change range of the gas temperature, so that the sampled data can reflect the real gas temperature change, reasonable data preprocessing is performed, the current environment temperature value data subjected to the data preprocessing has training value and representativeness, the model parameters are updated by back propagation on a training set, the control error is reduced to the maximum extent, the production of the future nasal airflow temperature predicted value is more accurate, the training set data subjected to the data preprocessing is adopted to verify the trained neural network model, various data acquired and preprocessed in the preamble step can be effectively utilized, the quality of the neural network model can be further improved by testing and evaluating the model by utilizing real-time testing, and the better effect is achieved; the method also comprises the step of comparing the nose airflow temperature predicted value with the current nose airflow temperature value by the neural network model during actual application, and then carrying out corresponding judgment and operation, so as to ensure that the nose person can obtain the optimal temperature and humidity of the breathing machine after adjustment. The details of refinement and optimization are closely served in a large comprehensive consideration direction, so that the establishment, training and deployment application of the neural network model of the scheme are advanced layer by layer, and finally the scheme of the invention is obtained. The scheme of the invention is comprehensively considered in a large direction and further refined and optimized in a small aspect, breaks through the limitation of the conventional temperature and humidity control mode of the breathing machine, provides a novel and reasonable control method and system, has unique and ingenious technical conception, and has outstanding substantive characteristics and remarkable progress.

Drawings

FIG. 1 is a system block diagram of an artificial intelligence based ventilator warming and humidification control system in accordance with an embodiment of the present invention;

FIG. 2 is a schematic flow chart of a ventilator heating and humidifying control method based on artificial intelligence according to an embodiment of the invention;

FIG. 3 is a schematic diagram of model selection and training in accordance with an embodiment of the present invention;

FIG. 4 is a schematic diagram of a temperature change curve of a first ventilator when a temperature control experiment of a heating plate is performed according to an embodiment of the present invention;

FIG. 5 is a schematic diagram of a temperature change curve of a second ventilator when a temperature control experiment of a heating plate is performed according to an embodiment of the present invention;

FIG. 6 is a schematic diagram showing a temperature change curve of a third ventilator when a temperature control experiment of a heating plate is performed according to an embodiment of the present invention;

fig. 7 is a schematic diagram of a temperature change curve of a fourth ventilator when a temperature control experiment of a heating plate is performed in the embodiment of the present invention;

FIG. 8 is a schematic diagram of a temperature change curve of a fifth ventilator in a heating pan temperature control experiment according to an embodiment of the present invention;

fig. 9 is a schematic diagram of a temperature change curve of the sixth ventilator when the temperature control experiment of the heating plate is performed according to the embodiment of the present invention.

Each part in the figure:

100. a ventilator;

110. a controller;

120. a heating plate;

130. a temperature sensor;

140. A blower;

145. an air outlet;

150. a nose information temperature and humidity sensor;

160. an environmental temperature and humidity sensor.

Detailed Description

In order that the above objects, features and advantages of the application will be readily understood, a more particular description of the application will be rendered by reference to the appended drawings. In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. The present application may be embodied in many other forms than described herein and similarly modified by those skilled in the art without departing from the spirit of the application, whereby the application is not limited to the specific embodiments disclosed below.

The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the present disclosure. Singular forms such as "a," "an," "the," and "the" are intended to include the plural forms as well, as used herein.

The terms "first," "second," and the like, as used herein, do not denote a particular order or sequence, nor are they intended to be limiting, but rather are merely used to distinguish one element or operation from another in the same technical term.

As used herein, "connected" or "positioned" may refer to two or more components or devices in physical contact with each other, or indirectly, or in operation or action with each other.

As used herein, the terms "comprising," "including," "having," and the like are intended to be open-ended terms, meaning including, but not limited to.

The term (terms) as used herein generally has the ordinary meaning of each term as used in this field, in this disclosure, and in the special context, unless otherwise noted. Certain terms used to describe the present disclosure are discussed below, or elsewhere in this specification, to provide additional guidance to those skilled in the art in connection with the description herein.

The present invention aims to provide more comfortable respiratory support for a patient by warming and humidifying control of a ventilator. For example, in the scheme of the invention, related data is processed by adopting an artificial intelligence algorithm mode, and the neural network model is used for controlling the heating disc of the breathing machine, so that the control precision and the response speed can be improved, and meanwhile, the requirement of manual adjustment is reduced, thereby improving the automation degree and the safety of the breathing machine system.

An embodiment I of the present invention provides a ventilator heating and humidifying control method based on artificial intelligence, the ventilator heating and humidifying control method includes the following steps:

s100, arranging a temperature sensor at a heating disc of a breathing machine, arranging a nose information temperature and humidity sensor at an air outlet of a fan of the breathing machine, and arranging at least two environment temperature and humidity sensors around the breathing machine; the method comprises the steps that a controller obtains an actual output temperature value of a heating plate detected by a temperature sensor, an ambient temperature value detected by a plurality of ambient temperature and humidity sensors and a nose information airflow temperature value detected by a nose information temperature and humidity sensor;

S200, constructing a neural network model taking an actual output temperature value of the heating plate, an ambient temperature value and a nasal airflow temperature value as inputs and taking a control action of a controller as output, wherein the step of constructing the neural network model comprises the following steps of:

s210, data acquisition, namely determining a sampling frequency according to the response speed of a breathing machine system and the change range of the gas temperature, acquiring environmental temperature value data according to the sampling frequency, marking the breathing machine set temperature, the actual output temperature value of a heating disc and the nasal airflow temperature value of each data point, and storing the acquired data into a controller;

s220, preprocessing data, namely preprocessing the acquired data, collecting an environmental temperature value, a breathing machine set temperature, an actual output temperature value of a heating disc, a nasal airflow temperature value and a corresponding time stamp of each data point within a standard limit as breathing machine historical temperature data, and dividing the breathing machine historical temperature data into a training set and a verification set;

s230, model selection and training, wherein a model consisting of an input layer, a hidden layer and an output layer is selected and training is performed by using a training set; the input layer is used for receiving the ventilator history temperature data of the training set, and the activation function of the ventilator history temperature data in the input layer is provided with characteristic weights and bias items corresponding to the input environment temperature value, the ventilator set temperature, the actual output temperature value of the heating disc and the nasal airflow temperature value; the hidden layer is used for processing the data of the input layer and providing signals for the output layer; the output layer generates a predicted value for the temperature of the future nasal airflow and outputs a heating disc temperature control parameter of the current breathing machine; back propagation is carried out on the training set to update training parameters of the characteristic weights and the bias items, and training is carried out again by using the new training parameters;

s250, testing and evaluating the model by utilizing a real-time test, and evaluating whether the difference value between the obtained nasal airflow temperature value of the actual air outlet of the breathing machine and the set temperature of the breathing machine is within an evaluation range value or not after the control action obtained by the neural network model is executed;

and S300, deploying the trained, verified, tested and evaluated neural network model into a breathing machine for use, acquiring an ambient temperature value under the use environment, comparing the nasal airflow temperature predicted value obtained by the output layer by the neural network model, and judging whether the difference value between the nasal airflow temperature predicted value and the current nasal airflow temperature value is within the use range value, if not, re-executing the acquisition of the ambient temperature value under the use environment, and if so, executing the breathing machine control by using the output heating plate temperature control parameter.

By implementing the first embodiment, the control accuracy and the response speed can be improved by controlling the heating plate of the breathing machine through the neural network model, and meanwhile, the requirement of manual adjustment is reduced, so that the automation degree and the safety of a breathing machine system are improved; meanwhile, the method comprises the steps of arranging a heating and humidifying control system of the breathing machine so as to simplify the structure and reduce the cost, so that the acquired temperature value of the nasal airflow is close to the actual inhalation of a user, and the reliability of environmental temperature acquisition is improved to the maximum extent; the method comprises the steps of constructing the neural network model, wherein the sampling frequency is determined according to the response speed of a breathing machine system and the change range of the gas temperature, so that the sampled data can reflect the real gas temperature change, reasonable data preprocessing is performed, the current environment temperature value data subjected to the data preprocessing has training value and representativeness, the model parameters are updated by back propagation on a training set, the control error is reduced to the maximum extent, the production of the future nasal airflow temperature predicted value is more accurate, the training set data subjected to the data preprocessing is adopted to verify the trained neural network model, various data acquired and preprocessed in the preamble step can be effectively utilized, the quality of the neural network model can be further improved by testing and evaluating the model by utilizing real-time testing, and the better effect is achieved; the method also comprises the step of comparing the nose airflow temperature predicted value with the current nose airflow temperature value by the neural network model during actual application, and then carrying out corresponding judgment and operation, so as to ensure that the nose person can obtain the optimal temperature and humidity of the breathing machine after adjustment.

A further description of one preferred embodiment of the present invention will be provided below.

In the preferred embodiment, a ventilator heating and humidifying control system is arranged in the ventilator, and the ventilator heating and humidifying control system comprises a heating disc, a temperature sensor, a fan, a nose information temperature and humidity sensor, an environment temperature and humidity sensor and a controller.

The ventilator heating and humidifying control method based on artificial intelligence of the preferred embodiment comprises the following steps:

s100, arranging a temperature sensor at a heating disc of a breathing machine, arranging a nose information temperature and humidity sensor at an air outlet of a fan of the breathing machine, and arranging at least two environment temperature and humidity sensors around the breathing machine; the controller acquires the actual output temperature value of the heating plate detected by the temperature sensor, the ambient temperature values detected by the ambient temperature and humidity sensors and the nasal airflow temperature value detected by the nasal airflow temperature and humidity sensor. By placing the nose information temperature and humidity sensor at the air outlet, a conventional heating pipeline can be removed, so that the structure is simplified, the cost is reduced, and the acquired nose information air flow temperature value is close to the actual inhalation temperature of a user; by arranging a plurality of ambient temperature and humidity sensors around the breathing machine, the reliability of ambient temperature acquisition can be improved to the greatest extent, so that inaccuracy of results caused by damage of a single ambient temperature and humidity sensor or the fact that a certain sensor is close to a heat source is reduced to the greatest extent.

s210, data acquisition, namely setting a sampling frequency, acquiring environmental temperature value data with the sampling frequency, marking the set temperature of the breathing machine and the actual output temperature value of the heating disc at each data point, and storing the acquired data into a controller; the sampling frequency is determined according to the response speed of the breathing machine system and the change range of the gas temperature, so that the sampled data can reflect the real gas temperature change, the built neural network model is more accurate and intelligent, and the deployment and application of the follow-up neural network model can also be matched with the real gas temperature change to carry out heating and humidifying control.

In the step S210 of data acquisition, the steps include:

S212, improving data quality: removing abnormal values and noise from the acquired data, adopting smoothing filtering to process,wherein each letter is defined as k representing the kth sampling point and N representing the total number of times (or called filter length) that the sampling is required;

s213, labeling: accurately marking the set temperature of the breathing machine and the actual output temperature value of the heating plate at each data point, so that the subsequent model training and evaluation are convenient;

S220, preprocessing the acquired environmental temperature value data, calculating the average value and standard deviation of an environmental temperature value data set acquired by the breathing machine, determining the standard limit of normal data of the breathing machine, comparing each independent environmental temperature value data with the standard limit, and if the current environmental temperature value data is within the standard limit, reserving the environmental temperature value data; if the current environmental temperature value data is not within the standard limit, deleting the environmental temperature value data, the set temperature of the breathing machine under the data point and the actual output temperature value of the heating disc; collecting the collected environmental temperature value, the ventilator set temperature, the actual output temperature value of the heating disc, the nasal airflow temperature value and the corresponding time stamp under each data point within the standard limit as ventilator historical temperature data, and dividing the ventilator historical temperature data into a training set and a verification set; the average value represents the central position of the temperature of the breathing machine in the measuring process, the standard deviation reflects the temperature variability of the breathing machine in the measuring process, so that the current environmental temperature value data subjected to data preprocessing has training value and representativeness, and the obtained historical temperature data of the breathing machine can be used for constructing the neural network model as data which are close to the actual environmental temperature value, the set temperature of the breathing machine, the actual output temperature value of the heating disc, the nasal airflow temperature value and the like, thereby improving the construction quality of the neural network model, training efficiency and accuracy, and enabling the neural network model to be well adapted to various uncertainties and complexity.

In the step of preprocessing S220, the method includes the steps of:

s224, comparing the results: each individual temperature measurement is compared to a specification limit. If the result is within specification limits, then the result is considered acceptable and normal;

In the step of S220 of data preprocessing, when the ventilator history temperature data is divided into a training set and a verification set, the number ratio of the ventilator history temperature data contained in the training set to the verification set is: 6/4 to 8/2, more preferably 7/3.

S230, model selection and training, wherein a model consisting of an input layer, a hidden layer and an output layer is selected and training is performed by using a training set; the input layer is used for receiving the ventilator history temperature data of the training set, and the activation function of the ventilator history temperature data in the input layer is provided with characteristic weights and bias items corresponding to the input environment temperature value, the ventilator set temperature, the actual output temperature value of the heating disc and the nasal airflow temperature value; the hidden layer is used for processing the data of the input layer and providing signals for the output layer; the output layer generates a predicted value for the temperature of the future nasal airflow and outputs a heating disc temperature control parameter of the current breathing machine; in the model selection and training step, setting an activation value corresponding to an input layer activation function for each acquired temperature data in a training set, further calculating an output layer temperature error generated by an output layer according to a forward propagation mode, then calculating a reverse propagation error according to the output layer temperature error, obtaining new training parameters related to a characteristic weight and a bias item by the reverse propagation error, and training by taking the training parameters as the new characteristic weight and the bias item of the environment temperature value, the ventilator set temperature, the actual output temperature value of a heating disc and the nasal airflow temperature value in the input layer; by adopting the mode to train the neural network model, the training parameters are updated by back propagation on the training set so as to maximally reduce the control error, the production of the future nasal information air flow temperature predicted value is more accurate, the output of the heating plate temperature control parameters and the temperature control of the heating plate are more in line with the expectations, and thus, the nasal information temperature required by a nasal information person can be controlled more accurately.

In the step of model selection and training at S230, the step of back-propagating on the training set to update the training parameters comprises the following:

receiving data of input layer at hidden layer and calculating neuron input numberAccording to this, the process is carried out,wherein->Input representing the j-th neuron of the g-th layer,>characteristic weights representing that the kth neuron of the g-th layer is connected to the jth neuron of the g-th layer,/->Represents the output of the kth neuron of the g-1 th layer,/o>A bias term representing a jth neuron of a g-th layer;

for each ventilator history temperature data x in the training set, setting an activation value a corresponding to the input layer ^g The forward propagation formula is obtained: z ^g ＝ω ^g +a ^g-1 +b ^g ，a ^g ＝σ(z ^g ) Wherein g represents the number of layers, z represents the linear operation result of omega, a and b, omega represents the characteristic weight, a represents the activation value of z, and b represents the bias term;

further calculating an output layer temperature error generated by the output layer,wherein the formula is a differential operator expression of the output error formula, delta ^G Indicating output layer temperatureError value, & gt>Is->C represents a cost function, G represents the number of layers of the neural network, j represents the jth neuron;

obtaining new training parameters related to characteristic weights and bias terms from the back propagation error, wherein the new training parameters of the characteristic weights areWherein the new training parameters of the bias term areWherein η represents the learning rate, m represents the sample size, x represents the summation variable, δ ^x，g A represents the difference between the output layer temperature error and the current temperature error ^x，g-1 Represents an activation value, and T represents a sample set.

S240, verifying the trained neural network model by using the historical temperature data of the breathing machine in the verification set, judging the predicted value of the nasal airflow temperature output by the output layer and the difference value between the actual output temperature value of the heating disc and the nasal airflow temperature value in the later time node after the ambient temperature value and the nasal airflow temperature value of the previous time node are input into the neural network model, and judging that the neural network model is successfully trained if the difference value is within a verification range value (+ -0.2 ℃); if not, judging that the neural network model training fails, and retraining after adjusting the characteristic weight and the bias term in the step S230; the training set data after data preprocessing is adopted to verify the trained neural network model, the data based on the real change, such as the ambient temperature value, the breathing machine set temperature, the heating plate actual output temperature value, the nasal airflow temperature value and the like, which are acquired and preprocessed in the preamble step can be effectively utilized, the model verification is carried out at the moment, if the training of the neural network model is judged to be failed, only the feature weight and the bias item are required to be optimized in the training, the error due to human factors can be reduced, the accurate control is realized, the training amount and the time length are reduced for the training of the neural network model, and the neural network model can be trained well more quickly under the condition of guaranteeing the quality of the neural network model.

S250, testing and evaluating the model by utilizing a real-time test, and evaluating whether the difference value between the obtained nasal airflow temperature value of the actual air outlet of the breathing machine and the set temperature of the breathing machine is within an evaluation range value or not after the control action obtained by the neural network model is executed.

In the step of S250 model test and evaluation, the steps include:

s253, the environmental temperature values detected by the environmental temperature and humidity sensors are fed to a neural network model, and the temperature control parameters of the heating plate, which are required to be newly issued, are calculated;

s255, the difference value between the nose gas flow temperature value of the actual gas outlet of the breathing machine and the set temperature of the breathing machine is within a set range value (+ -0.2 ℃), if yes, the neural network model is judged to pass through the evaluation, and the deployment of the step S300 is carried out; if not, it is determined that the neural network model evaluation fails, and steps S100 and S200 are re-executed or step S230 is re-executed.

By testing and evaluating the model by using the real-time test, the quality of the neural network model can be further improved, so that a better effect is achieved.

S300, deploying the trained, verified, tested and evaluated neural network model into a breathing machine for use, wherein the using steps comprise:

s330, comparing the nose gas flow temperature predicted value and the difference value of the current nose gas flow temperature value by a neural network model to determine whether the difference value is within a range of use value (+ -0.2 ℃), re-executing the step S320, and executing the next step if the two values are consistent;

In step S300, the collected environmental temperature value data is compared with a specification limit, and if the current environmental temperature value data is within the specification limit, the actual output temperature value of the hot plate, the environmental temperature value and the nasal airflow temperature value under the current timestamp are used as the input of the neural network model; and if the current environmental temperature value data is not in the standard limit, continuing to acquire the environmental temperature value data at the sampling frequency until the acquired environmental temperature value data is in the standard limit.

In the process of using the neural network model, the neural network model compares the nose airflow temperature predicted value with the current nose airflow temperature value and then carries out corresponding judgment and operation, so as to ensure that the nose person can obtain the optimal temperature and humidity of the breathing machine after adjustment.

In a second embodiment, the present invention provides a ventilator heating and humidifying control system based on artificial intelligence, which is used for ventilator heating and humidifying control in the method in the first embodiment of the present invention, and the ventilator heating and humidifying control system includes a heating plate 120, a temperature sensor 130, a fan 140, a nose temperature and humidity sensor 150, an environmental temperature and humidity sensor 160, and a controller 110; wherein,,

a temperature sensor 130 is arranged at the heating plate 120 of the breathing machine 100, a nose temperature and humidity sensor 150 is arranged at the air outlet 145 of the fan 140 of the breathing machine 100, and at least two environment temperature and humidity sensors 160 are arranged around the breathing machine 100;

the controller 110 is electrically connected to the heating plate 120, the temperature sensor 130, the fan 140, the nose temperature and humidity sensor 150, and the ambient temperature and humidity sensor 160, and the controller 110 obtains the actual output temperature values of the heating plate detected by the temperature sensor 130, the ambient temperature values detected by the ambient temperature and humidity sensors 160, and the nose airflow temperature value detected by the nose temperature and humidity sensor 150; the controller 110 includes a neural network model that takes an actual output temperature value of the heating plate, an ambient temperature value, and a nasal airflow temperature value as inputs, and takes a control action of the controller 110 as an output, and the controller 110 obtains a heating plate temperature control parameter through the neural network model and then transmits a signal to the heating plate 120, so that the heating plate 120 performs temperature adjustment according to the heating plate temperature control parameter, and the output nasal airflow temperature value corresponds to a set temperature of the ventilator 100.

In the second embodiment of the present invention, preferably, the number of the environmental temperature and humidity sensors is four, the four environmental temperature and humidity sensors are distributed around the ventilator, and the data acquired by the four environmental temperature and humidity sensors calculate the current representative environmental temperature value according to an average weighting formula.

An embodiment III of the present invention provides a ventilator, which includes the ventilator heating and humidifying control system based on artificial intelligence according to the embodiment II of the present invention.

Since most of the common respirators have rectangular cubic structures, the invention is described in detail by taking environmental temperature and humidity sensors arranged on the front, back, left and right surfaces of the respirators as an example.

As shown in figure 1, in the breathing machine provided by the invention, a breathing machine heating and humidifying control system is deployed, the breathing machine heating and humidifying control system comprises a heating disc, temperature sensors, fans, nose information temperature and humidity sensors, environment temperature and humidity sensors and a controller, wherein the number of the environment temperature and humidity sensors is four, the four environment temperature and humidity sensors are respectively arranged at the front, back, left and right sides of the breathing machine, the controller selects MCU with the model of stm32f103rct6, the environment temperature and humidity sensors select HTU21D, the heating disc is a common heating disc temperature sensor (NTC) of breathing support equipment and is NTCS0805E3104FXT, and the fans are common fans of the breathing equipment. In this example, the gas flow is 30L/Min based on the user demand temperature of 33-58.

The flow can be referred to as follows:

s100, a controller (MCU) collects an actual output temperature value of the heating plate detected by the temperature sensor, an environment temperature value detected by the four environment temperature and humidity sensors and a nose airflow temperature value detected by the nose temperature and humidity sensor, and a current representative environment temperature value is calculated according to an average weighting formula.

And S200, constructing a neural network model taking an actual output temperature value of the heating plate, an ambient temperature value and a nasal airflow temperature value as inputs and taking a control action of a controller as an output, wherein the control action of the controller comprises prediction of the nasal airflow temperature in the future and output of temperature control parameters of the heating plate.

The specific implementation steps of step S200 are as follows:

and (3) data acquisition: and collecting a large amount of environmental temperature value data, and simultaneously recording the set temperature of the breathing machine, the actual output temperature value of the heating disc and the nasal airflow temperature value at each data point.

A) Sampling frequency: the sampling frequency, in this example 100hz, needs to be determined based on the response speed of the ventilator system and the range of variation in the gas temperature.

B) Quality of data: the acquired data should be freed from outliers and noise and processed by smoothing y (temperature k) =y (temperature k-1) + [ x (temperature k) -x (temperature k-N) ]/acquisition temperature data length N.

C) Labeling: the set temperature and the actual output temperature of the heating plate at each data point are correctly marked, so that the subsequent model training and evaluation are convenient.

D) And (3) data storage: and storing the acquired data in an MCU memory so as to facilitate subsequent data analysis and mining.

2. Data preprocessing: the collected original data is processed, abnormal values are removed through a three-sigma method, and the method specifically comprises the following steps:

according to the property of normal distribution, the data within the standard deviation of plus or minus 3 times of the temperature mean value acquired by the breathing machine is regarded as normal data, and the temperature value exceeding the range is judged as an abnormal value. Calculating the average value and standard deviation of a temperature set acquired by a breathing machine; calculating upper and lower limits according to the average value and the standard deviation; for each temperature point, judging whether the temperature point is within an upper limit and a lower limit, and if the temperature point is not within the upper limit and the lower limit, regarding the temperature point as an abnormal value and performing elimination processing:

1) Temperature data of the ventilator is collected according to the sampling frequency of the complaints.

2) Calculation of mean and standard deviation: the mean and standard deviation of the temperature were calculated using these data. And (3) injection: the average value represents the center position of the ventilator temperature of the measurement process, while the standard deviation reflects the ventilator temperature variability of the measurement process.

3) Determining a specification limit: and adding or subtracting three times of standard deviation to the temperature average value acquired by the breathing machine to obtain upper and lower standard limits. These specifications represent the range in which the temperature measurement process should fall under normal operating conditions.

4) Comparison result: each individual temperature measurement is compared to a specification limit. If the result is within specification limits, then the result is considered acceptable and normal;

5) And collecting the collected environmental temperature value, the respirator set temperature, the actual output temperature value of the heating disc, the nasal airflow temperature value and the corresponding time stamp under each data point within the standard limit as the respirator historical temperature data, and dividing the respirator historical temperature data into a training set and a verification set.

3. Model selection and training: a multi-layer perceptron (MLP) is selected and model training is performed using the preprocessed data.

A portion of the data is typically partitioned as a validation set during training for model tuning and optimization. The model consists of an input layer, a hidden layer and an output layer. The input layer receives ventilator historical temperature data including a time stamp, ambient temperature values of the sensors of temp 1-temp 4. The hidden layer processes the input and provides signals to the output layer. The output layer generates a future nasal airflow temperature predicted value and outputs a heating plate temperature control Parameter (PWM) of the current breathing machine. Let us assume that we have a training sample X comprising n input features, each feature being denoted xi. We also have a target variable y, which represents the desired temperature value. Input layer (receiving ventilator historic temperature) =f1 (ω) ₁ x ₁ +ω ₂ x ₂ +...+ω _i x _i +b)；Characteristic weights representing that the kth neuron of the g-th layer is connected to the jth neuron of the g-th layer,/->Represents the output of the kth neuron of the g-1 th layer,/o>A bias term representing the jth neuron of the g-th layer, receiving data of the input layer at the hidden layer and calculating neuron input data,/v> Representing the output of the jth neuron of the g-th layer, σ representing the activation function, ++>Setting an activation value corresponding to an input layer for each acquired temperature data x to obtain a forward propagation formula: z ^g ＝ω ^g +a ^g-1 +b ^g ，a ^g ＝σ(z ^g ) The method comprises the steps of carrying out a first treatment on the surface of the Further calculating the temperature error value delta generated by the output layer _G :/>From output layer temperature error delta ^G Obtaining a counter-propagating temperature error value delta ^g ，δ ^g ＝((ω ^g+1 ) ^T δ ^g+1 )·σ’(z ^g ) Thereby obtaining training parameters，/>

4. After the model is trained, the trained neural network model is verified by using the ventilator historical temperature data in the verification set, and the verification range value is compared.

5. Model test and evaluation: the model is tested and verified by utilizing a real-time test, and the specific method is as follows:

1) Transplanting the trained MLP model into a temperature control system (MCU);

2) Issuing a target temperature to a temperature control system;

3) Collecting temp 1-temp 4 temperature data, and feeding the data to MLP, and calculating to obtain PWM required to be newly issued by heating disc;

4) And sending PWM to the heating disc, and observing that the actual air outlet temperature of the breathing machine and the set temperature are different by plus or minus 0.2 ℃ through a standard thermometer.

6. Deployment and application: and deploying the trained AI model into a breathing machine, and starting the breathing machine to obtain the temperature and humidity airflow required by the nasal stutterer.

After the scheme of the invention is adopted, the applicant carries out a comparison experiment on the breathing machine adopting the artificial intelligence-based breathing machine heating and humidifying control method and the breathing machine adopting the conventional PID control method, wherein the first group of three breathing machines adopting the scheme and the second group of three breathing machines adopting the PID control method are respectively a heating disc temperature control experiment and a nasal airflow temperature control experiment, and the purposes and the reasons of the two experiments are that the temperature control experiment of the heating disc is carried out because the control of the breathing machine on the temperature and the humidity is realized by the heating temperature control of the heating disc, the accuracy and the stability of the temperature control of the heating disc are very important, and the accuracy of the temperature control of the heating disc can also be directly reflected on the nasal airflow temperature; the snuff air flow temperature control experiment is performed because the snuff air flow temperature is taken as the temperature when a snuffer inhales, can bring the most visual feeling to the snuffer, and is also taken as the most direct data parameter after being subjected to the influence of the environmental temperature and the control variable is corrected by the control method, so that the snuff air flow temperature control experiment is considered to be the most important point in the scheme of the invention.

The heating plate temperature control experiment and the snuff air flow temperature control experiment are described in detail below.

The experimental object:

the first ventilator adopts the first ventilator based on the artificial intelligence ventilator heating and humidifying control method;

the second breathing machine adopts the breathing machine heating and humidifying control method based on artificial intelligence;

the third ventilator adopts the artificial intelligence-based ventilator heating and humidifying control method;

a fourth ventilator using a non-artificial intelligent temperature control algorithm (PID), wherein the PID coefficient is repaired for a certain time;

fifth ventilator, fifth ventilator with other algorithms (PID) of non-artificial intelligence temperature control, wherein the PID coefficients are adjusted only for a short time or hardly;

and a sixth ventilator using a non-artificial intelligent temperature control algorithm (PID), wherein the PID coefficient is manually adjusted for a long time.

The temperature control experiment of the heating plate is carried out under the conditions that the ambient temperature is 20.5 ℃, the humidity is 36%, the atmospheric pressure is 101.8Kpa, and the heating temperature of the preset heating plate is 56 ℃.

The specific test conditions are as follows: and uniformly selecting 5 points on the surface of each heating disc of six experimental objects, and fixing the K-type thermocouple on each point by using thermocouple glue/heat conducting silica gel/high-temperature adhesive tape.

Experimental results referring to fig. 4 to 9, temperature change curves of the heating plates of the first to sixth ventilators with time are shown.

Fig. 4 shows a first ventilator using the artificial intelligence temperature control algorithm of the present example, which is smooth and stable overall except for a small Xu Guochong at the beginning.

Fig. 5 shows a second ventilator using the artificial intelligence temperature control algorithm of the present example, which is smooth and stable overall except for a small Xu Guochong at the beginning.

Fig. 6 shows a third ventilator using the artificial intelligence temperature control algorithm of the present example, which is smooth and stable except for a small Xu Guochong at the beginning.

The three groups of experimental data can obtain that the actual temperature is basically consistent with the preset temperature in the whole experimental period, and no larger shaking occurs, so that the experimental expectation is met.

Fig. 7 shows a fourth ventilator (repaired for a certain period of time, when not completely repaired) using other algorithms (PID) of non-artificial intelligence temperature control, and it can be seen that the initial temperature is less than Xu Guochong, the overall temperature jitter is greater, and if the problem is to be repaired, a great deal of time and effort is required to fix the PID coefficient.

Fig. 8 shows a fifth ventilator (integration operation is not adjusted, is adjusted in a short time or is hardly adjusted) using other algorithms (PID) of non-artificial intelligence temperature control, and it can be seen that the initial temperature is smoothed for a period of time after a small overshoot, and the subsequent temperature jitter is large, and if the problem is to be repaired, a great deal of time and effort is required to adjust the PID coefficient.

FIG. 9 shows a ventilator No. six (manually adjusted for a long time) employing other algorithms (PID) for non-artificial intelligence temperature control, which can see that the first long segment has a large temperature jitter and then flattens out, and if the problem is to be repaired, a certain time effort is required to set the PID coefficient.

From fig. 7, 8 and 9, it can be concluded that, with the conventional PID temperature control algorithm, temperature jitter is extremely easy to occur, and it takes a lot of time to set the PID coefficients.

Compared with two groups of test results of the heating plate temperature control experiment, the artificial intelligence algorithm of the scheme does not need engineers to spend a great deal of time on PID coefficient setting, reduces the operation threshold while saving time, and has higher stability compared with the conventional temperature control.

The temperature control experiment of the nasal airflow is carried out at the ambient temperature of 20.5 ℃, the humidity of 36 percent and the atmospheric pressure of 101.8Kpa, and different temperatures are set according to different gear positions of the breathing machine, wherein the 0 gear is closed, the 3 gear is 23 ℃, the 4 gear is 25 ℃, and the 5 gear is 28 ℃.

The specific test conditions are as follows: and obtaining the nasal airflow temperature value of the actual air outlet of each breathing machine by using a standard thermometer.

The results of the ventilator number one are shown in Table 1.

TABLE 1

The experimental results of the ventilator No. two are shown in table 2.

TABLE 2

The results of the experiment of the third ventilator are shown in table 3.

TABLE 3 Table 3

The results of the experiment of the fourth ventilator are shown in table 4.

TABLE 4 Table 4

The experimental results of the ventilator No. five are shown in table 5.

TABLE 5

The results of the experiment on the ventilator number six are shown in table 6.

TABLE 6

Comparing the two groups of test results of the nose breath flow temperature control test, wherein the table 1, the table 2 and the table 3 are artificial intelligent temperature control algorithms, and the error between the preset value and the actual measured value is less than or equal to 0.2 ℃; tables 4, 5, and 6 each had a nose temperature error value of greater than 3 degrees celsius controlled by the conventional PID method. Therefore, the nasal resting temperature of the nasal resting temperature required by a person suffering from nasal resting can be controlled more accurately by the scheme.

The temperature control experiment of the heating disc and the temperature control experiment of the nasal air flow can be obtained, and the temperature control method and the temperature control system of the breathing machine based on the artificial intelligence, which are disclosed by the embodiment of the invention, have the advantages that the output of temperature control parameters of the heating disc and the temperature control of the heating disc are more in line with expectations, and the nasal air temperature required by a nasal person can be controlled more accurately, so that the problems that the conventional PID control method for heating and humidifying the breathing machine is complicated in adjustment, cannot be well adapted to various uncertainties and complexities, the limitation of a heating pipeline, the uncertainty of an environment temperature sensor and the like are solved, and the purpose of the invention is achieved.

The above embodiments are provided to illustrate the technical concept and features of the present invention and are intended to enable those skilled in the art to understand the content of the present invention and implement the same, and are not intended to limit the scope of the present invention. All equivalent changes or modifications made in accordance with the spirit of the present invention should be construed to be included in the scope of the present invention.

Claims

1. The ventilator heating and humidifying control method based on artificial intelligence is characterized by comprising the following steps of:

2. The artificial intelligence based ventilator warming and humidifying control method according to claim 1, wherein:

in the step S100, a temperature sensor is arranged at a heating disc of the breathing machine, a nose temperature and humidity sensor is arranged at an air outlet of a fan of the breathing machine, and at least two environment temperature and humidity sensors are arranged around the breathing machine.

3. The method for controlling warming and humidification of an artificial intelligence based ventilator according to claim 1, wherein in the step of collecting S210 data, the method comprises the steps of:

4. The artificial intelligence based ventilator warming and humidifying control method according to claim 1, wherein: in the step S220 of data preprocessing, calculating the average value and standard deviation of an environmental temperature value data set acquired by a breathing machine, determining the standard limit of normal data of the breathing machine, comparing each independent environmental temperature value data with the standard limit, and if the current environmental temperature value data is within the standard limit, reserving the environmental temperature value data; if the current environmental temperature value data is not within the standard limit, deleting the environmental temperature value data, the set temperature of the breathing machine under the data point and the actual output temperature value of the heating disc.

5. The artificial intelligence based ventilator warming and humidifying control method according to claim 4, wherein: in the step of preprocessing S220, the method includes the steps of:

s222, calculating an average value and a standard deviation of the environmental temperature value, and calculating the average value and the standard deviation of the temperature by using the data, wherein the average value represents the central position of the temperature of the breathing machine in the measuring process, and the standard deviation reflects the temperature variability of the breathing machine in the measuring process;

6. The artificial intelligence based ventilator warming and humidification control method of claim 1, wherein in the step of S220 data preprocessing, when ventilator history temperature data is divided into a training set and a verification set, a quantity ratio of ventilator history temperature data contained in the training set and the verification set is: 6/4 to 8/2.

7. The artificial intelligence based ventilator warming and humidifying control method according to claim 1, wherein: in the step of selecting and training the model S230, for each temperature data in the acquired training set, an activation value corresponding to the activation function of the input layer is set, an output layer temperature error generated by the output layer is further calculated according to the forward propagation mode, then a reverse propagation error is calculated according to the output layer temperature error, new training parameters related to feature weights and bias items are obtained according to the reverse propagation error, and the new training parameters are used as new feature weights and bias items of the environmental temperature value, the set temperature of the breathing machine, the actual output temperature value of the heating disc and the nasal airflow temperature value in the input layer for training.

8. The artificial intelligence based ventilator warming and humidification control method of claim 7, wherein: in the step of model selection and training at S230, the step of back-propagating on the training set to update the training parameters comprises the following:

9. The artificial intelligence based ventilator heating and humidifying control method according to claim 1, wherein in the step of verifying the model S240, after verifying that the neural network model inputs the ambient temperature value and the nasal airflow temperature value in the verification set of the previous time node, the nasal airflow temperature predicted value output by the output layer and the difference value between the actual output temperature value of the heating plate and the nasal airflow temperature value and the actual output temperature value of the heating plate in the verification set of the next time node after the adjustment of the temperature control parameter of the heating plate of the current ventilator are judged to be within the verification range value, if yes, the neural network model is judged to be successfully trained, and the next step is executed; if not, judging that the neural network model training fails, and retraining after adjusting the feature weights and the bias items in the step S230.

10. The artificial intelligence based ventilator warming and humidification control method of claim 1, wherein in the step of S250 model testing and evaluation, comprising the steps of:

11. The artificial intelligence based ventilator warming and humidification control method of claim 1, wherein in step S300, the using step thereof comprises:

12. The ventilator warming and humidifying control method based on artificial intelligence according to claim 1, wherein,

13. The artificial intelligence based ventilator warming and humidification control method of any one of claims 1 to 12, wherein the validation range value, the evaluation range value, the usage range value take on values between ± 0.1 ℃ and ± 0.3 ℃.

14. A ventilator warming and humidification control system based on artificial intelligence for use in the method of any one of claims 1 to 13, the system comprising: the ventilator heating and humidifying control system comprises a heating disc, a temperature sensor, a fan, a nose temperature and humidity sensor, an environment temperature and humidity sensor and a controller; wherein,,

15. The artificial intelligence based ventilator warming and humidification control system of claim 14, wherein: the number of the environmental temperature and humidity sensors is four, the four environmental temperature and humidity sensors are distributed around the breathing machine, and the current representative environmental temperature value is calculated according to an average weighting formula by the data acquired by the four environmental temperature and humidity sensors.

16. A ventilator comprising an artificial intelligence based ventilator warming and humidification control system of any of claims 14 or 15.