| CPC A61M 16/101 (2014.02) [A61M 16/202 (2014.02); G16H 50/30 (2018.01); A61M 2230/06 (2013.01); A61M 2230/205 (2013.01); A61M 2230/42 (2013.01)] | 6 Claims |
|
1. A control method for an oxygen generator, comprising following steps:
step S101, obtaining a first physical sign data of a user according to a first preset time interval A, performing step S102 when the first physical sign data meets one of preset threshold conditions, otherwise performing step S104;
wherein the first physical sign data comprises: blood oxygen saturation, heart rate and breathing frequency;
step S102, switching the oxygen generator to a direct current oxygen outlet mode, and inputting the first physical sign data of the user into a first model, outputting a value as a first opening adjustment value of a direct current oxygen outlet solenoid valve in a first preset time period B;
step S103, adjusting an opening of the direct current oxygen outlet solenoid valve to the first opening adjustment value in the first preset time period B;
step S104, switching the oxygen generator to a pulse oxygen outlet mode, and obtaining a second physical sign data of the user according to a second preset time interval D in a second preset time period C to generate a feature sequence;
the feature sequence comprises N sequence units, and an ith sequence unit represents the second physical sign data of the user at an ith time point, wherein 1≤i≤N and N=C/D;
the second physical sign data comprises: blood oxygen saturation, heart rate, breathing frequency and respiratory waveform graph;
wherein a horizontal axis of the respiratory waveform graph represents the time point and a vertical axis represents airway pressure;
step S105, inputting the feature sequence into a second model, outputting a value as a second opening adjustment value when a pulse oxygen outlet solenoid valve is opened in a third preset time period E; and
step S106, adjusting an opening of the pulse oxygen outlet solenoid valve when opened to the second opening adjustment value in the third preset time period E;
wherein the preset threshold conditions comprise: the blood oxygen saturation is less than a first threshold; the heart rate is greater than or equal to a second threshold value; the breathing frequency is greater than or equal to ta third threshold; wherein the first threshold, the second threshold and the third threshold are all self-defined parameters;
a calculation formula of the first model is as follows:
 wherein Open1 represents the first opening adjustment value, Blood, Heartand Breath respectively represent the blood oxygen saturation, the heart rate and the breathing frequency; W1, W2 and W3 respectively represent a first weight parameter, a second weight parameter and a third weight parameter, α, β, and γ respectively represent a first influence coefficient, a second influence coefficient and a third influence coefficient;
the second model comprises N hidden layers, and an ith hidden layer inputs the ith sequence unit of the feature sequence and outputs a hidden state;
the hidden state output by an Nth hidden layer is input into a classifier, and a classification space of the classifier represents the second opening adjustment value when the pulse oxygen outlet solenoid valve is opened in the third preset time period E;
a calculation formula of the second model comprises:
 wherein Xi represents the ith sequence unit of the feature sequence input by the ith hidden layer, and hi and hi-1 respectively represent a hidden state output by the ith hidden layer and an i-1th hidden layer, h0 is assigned as 0, ri, zi and
 , respectively represent reset gate, update gate and a candidate hidden state of the ith hidden layer, Wir, Wir,2 and bir respectively represent the first weight parameter, the second weight parameter and an offset parameter corresponding to the reset gate of the ith hidden layer, Wiz,1, Wiz,2 and biz respectively represent the first weight parameter, the second weight parameter and an offset parameter corresponding to the update gate of the ith hidden layer, Wih,1Wih,2 and bih respectively represent the first weight parameter, the second weight parameter and the offset parameter corresponding to the candidate hidden state of the ith hidden layer, ⊙ represents Hadamard product, ReLU represents ReLU activation function, and tanh represents hyperbolic tangent activation function;obtaining a sample label of a training sample for training the second model comprises following steps:
step S301, before using the oxygen generator, self-defining standard values of the blood oxygen saturation, the heart rate and the breathing frequency of the user as a first vector;
step S302, randomly generating one opening value as the second opening adjustment value when the pulse oxygen outlet solenoid valve is opened in the third preset time period E;
step S303, after the third preset time period E, measuring the blood oxygen saturation, the heart rate and the breathing frequency of the user as a second vector, and calculating a Euclidean distance between the first vector and the second vector;
step S304, repeating the step S301 to the step S303 to obtain the second opening adjustment value corresponding to a minimum value of the Euclidean distance between the first vector and the second vector as the sample label of the training sample; and
step S305, repeating the step S304 until sample labels of K training samples are obtained, wherein K is a self-defined parameter.
|