A neural network and approximation type internal model control-based infusion heating control method

Abstract

The application discloses a kind of based on neural network and approximation type internal model control infusion heating control method, comprising the following steps: step S1: with the form of full power heating, the system response curve of infusion heating system is obtained, and infusion heating parameter is obtained based on system response curve;Step S2: based on infusion heating parameter, the equivalent controller model of infusion heating system is established;Step S3: construct the approximation type internal model control system based on neural network and equivalent controller model, realize the control of infusion heating.The application can be automatically adjusted according to temperature deviation, real-time correction, online optimization, and quickly and stably reach target heating temperature.

Description

Technical Field

[0001] This invention relates to the field of infusion control, and in particular to an infusion heating control method based on neural networks and approximation internal model control. Background Technology

[0002] In current infusion heaters, the key technology used for liquid heating and temperature control is typically the Proportional-Integral-Differential (PID) control algorithm. The PID algorithm is a regulator that controls based on the proportional, integral, and derivative of the deviation signal. It features a clear structure, simple principle, and ease of implementation, making it the most widely used regulator in industrial and agricultural production and the most commonly used control algorithm in temperature control systems. However, due to the complex dynamic characteristics of temperature control systems, such as pure time delay, nonlinearity, and time-varying behavior, it is difficult to establish an accurate mathematical model of the controlled object. Furthermore, in practical applications, traditional PID control algorithms require repeated adjustments of the controller parameters based on the deviation trend, relying on extensive debugging experience to manually tune the optimal control parameters so that the controlled variable can quickly, accurately, and without oscillations track the target setpoint. Moreover, once the parameters are determined, they cannot be automatically adjusted, corrected in real time, or optimized online based on the deviation. Therefore, when existing infusion heaters use traditional PID control technology to heat and maintain the temperature of liquids, problems such as slow heating rate, large temperature fluctuations, long temperature adjustment time, or even failure to stabilize at the target temperature may occur due to unreasonable adjustment of controller parameters, changes in liquid flow rate, changes in the heating object model or heat-conducting material. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the prior art and provide an infusion heating control method based on neural networks and approximation internal model control, which can automatically adjust, correct in real time, and optimize online according to temperature deviation, so as to quickly and stably reach the target heating temperature.

[0004] The objective of this invention is achieved through the following technical solution: an infusion heating control method based on neural networks and approximation internal model control, comprising the following steps:

[0005] Step S1: Obtain the system response curve of the infusion heating system in the form of full-power heating, and obtain the infusion heating parameters based on the system response curve;

[0006] Step S2: Based on the infusion heating parameters, establish an equivalent controller model for the infusion heating system;

[0007] Step S3: Construct an approximation internal model control system based on neural networks and equivalent controller models to achieve control of infusion heating.

[0008] The beneficial effects of this invention are: This invention integrates the internal model control algorithm with neural network technology, and derives the backpropagation algorithm of the equivalent approximation internal model controller based on neural network online learning. By adjusting a control parameter in real time online, the control performance of the infusion heating system can be significantly improved, the patient's infusion experience can be improved, and the discomfort caused by infusion can be relieved. Attached Figure Description

[0009] Figure 1 This is a schematic diagram illustrating the principle of the present invention. Detailed Implementation

[0010] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings, but the scope of protection of the present invention is not limited to the following description.

[0011] like Figure 1 As shown, an infusion heating control method based on neural networks and approximation internal model control is characterized by the following steps:

[0012] Step S1: Obtain the system response curve of the infusion heating system in the form of full-power heating, and obtain the infusion heating parameters based on the system response curve;

[0013] Step S1 includes:

[0014] S101. The infusion heating system is heated at full power, and the infusion temperature and corresponding heating time are continuously collected within a set time period from the start of heating. The system response curve of the infusion heating system is constructed with heating time as the horizontal axis and infusion temperature as the vertical axis.

[0015] S102. Obtain infusion heating parameters based on the system response curve, including:

[0016] Determine the initial temperature , that is, the infusion temperature at the start of heating;

[0017] Determine the steady-state temperature That is, the infusion temperature tends to stabilize. If, during full-power heating, the infusion temperature remains stable for n consecutive preset time intervals... If the infusion temperature change does not exceed the preset threshold within n time intervals, then the n time intervals will be... The average infusion temperature is taken as the steady-state temperature;

[0018] Based on the initial temperature steady-state temperature Intermediate temperature used in computational modeling , :

[0019] ;

[0020] ;

[0021] intermediate temperature The corresponding time on the system response curve is denoted as . , to intermediate temperature The corresponding time on the system response curve is denoted as . ;

[0022] Calculate the static gain of the temperature object The time constant of the temperature object Lag time of temperature object :

[0023] ;

[0024] ;

[0025] .

[0026] Step S2: Based on the infusion heating parameters, establish an equivalent controller model for the infusion heating system;

[0027] Step S2 includes:

[0028] S201. Process of establishing the internal model of the infusion heating system :

[0029] ;

[0030] In the formula, This describes the actual control process of the infusion heating system. It is a complex variable;

[0031] S202. Inner mold process Decomposed into minimum phase part and unstable zero-point part :

[0032] ;

[0033] ;

[0034] S203. Introduce a first-order filter element:

[0035] ;

[0036] In the formula, These are the filter coefficients. For the filtering order, and Divide and establish an internal model controller :

[0037] ;

[0038] S204. Establish a mathematical model for the equivalent controller of the infusion system:

[0039] ;

[0040] in, Indicates the equivalent controller. Indicates the internal terminal controller. Indicates the internal model process, It is a complex variable;

[0041] In order to make the internal mold controller The transfer function is a complex variable Let the rational proper fraction function be the order of the denominator of the low-pass filter. The value of ,Will , Substitute the value In the middle, we get:

[0042] ;

[0043] Equivalent controller By adjusting the control parameters This improves the control performance of the infusion heating system.

[0044] Step S3: Construct an approximation internal model control system based on neural networks and equivalent controller models to achieve control of infusion heating.

[0045] Step S3 includes:

[0046] S301. Construct an approximation internal model control system based on neural networks and equivalent controller models, including a temperature deviation calculation module, an equivalent controller, a neural network model, an infusion heating control system, and a temperature sensor module;

[0047] The temperature sensor module is used to collect the liquid temperature of the infusion tube in the infusion heating control system in real time, i.e., the actual temperature. The output of the temperature sensor module is connected to the temperature deviation calculation module and the neural network module respectively. The output of the temperature deviation calculation module is connected to the equivalent controller and the neural network model respectively. The output of the neural network model is connected to the control input of the equivalent controller. The output of the equivalent controller is connected to the infusion heating control system.

[0048] Step S302. Based on the approximation internal model control system, control of the infusion heating process is achieved:

[0049] The infusion tubing is attached to the temperature sensor module, and the current fluid temperature in the infusion tubing is sampled every 1 second, which is the actual temperature of the infusion heating. And perform the following procedure:

[0050] (1) The data is fed into the temperature deviation calculation module and the neural network model.

[0051] (2) Temperature deviation calculation module calculates the target temperature of the infusion heating system. Compared with actual temperature Temperature deviation between ;

[0052] (3) The neural network calculation model is based on the target temperature Actual temperature and temperature deviation Dynamic output equivalent controller Control parameters in ;

[0053] (4) Equivalent controller Control parameters based on neural network model output Temperature deviation output by the temperature deviation calculation module Output equivalent control information ;

[0054] (5) Infusion heating control system, used to control the infusion based on equivalent control information This enables heating control, adjusting the actual infusion temperature, and theoretically... .

[0055] The neural network calculation model is based on the target temperature. Actual temperature and temperature deviation Dynamic output equivalent controller Control parameters in ,include:

[0056] The neural network model is constructed using a backpropagation (BP) neural network structure with 3 nodes in the input layer, 5 nodes in the hidden layer, and 1 node in the output layer.

[0057] The output of the i-th node in the input layer is :

[0058]

[0059] in, , , The superscript [] indicates the number of network layers. A network layer of 1 indicates an input layer, a network layer of 2 indicates a hidden layer, and a network layer of 3 indicates an output layer.

[0060] Forward computation of the hidden layer of the network:

[0061]

[0062] in, Let l be the weighted input of the l-th node in the hidden layer. This is the output of the l-th node in the hidden layer. The weights connecting the i-th node in the input layer to the l-th node in the hidden layer;

[0063] Forward computation of the network output layer:

[0064]

[0065] in, This represents the weighted input of the j-th node in the network output layer. The output of the j-th node in the network output layer, i.e., the control parameters. Since the network output layer has only one node, .

[0066] The neural network model uses gradient descent with momentum, following the negative gradient direction, and sets a learning rate. momentum coefficient For weight parameters Real-time adjustments are made, and the internal model equivalent controller is dynamically output. Control parameters in The adjustment method is as follows:

[0067] Take the error function:

[0068] ;

[0069] Network output layer gradient:

[0070]

[0071] The backpropagation formula for the output layer is:

[0072]

[0073] in, Substituting the error of the third layer into the gradient of the network output layer, we get:

[0074]

[0075] Gradient of hidden layer:

[0076] .

[0077] This invention addresses the challenge of precise temperature control in existing infusion heating systems, which primarily focus on optimizing and improving the heating structure of the equipment. However, innovative approaches to achieving precise temperature control from the perspectives of control theory and technology are severely lacking. Therefore, this invention integrates internal model control algorithms with neural network technology, deriving a backpropagation algorithm for an equivalent approximation internal model controller based on online neural network learning. By adjusting a single control parameter in real time, the control performance of the infusion heating system can be significantly improved, enhancing the patient's infusion experience and alleviating discomfort caused by infusions.

[0078] The foregoing description illustrates and describes a preferred embodiment of the present invention. However, as previously stated, it should be understood that the present invention is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and environments, and can be altered within the scope of the inventive concept described herein through the foregoing teachings or techniques or knowledge in related fields. Any modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention should be within the protection scope of the appended claims.

Claims

1. A method for controlling infusion heating based on neural networks and approximation internal model control, characterized in that: Includes the following steps: Step S1: Obtain the system response curve of the infusion heating system in the form of full-power heating, and obtain the infusion heating parameters based on the system response curve; Step S2: Based on the infusion heating parameters, establish an equivalent controller model for the infusion heating system; Step S3: Construct an approximation internal model control system based on neural networks and equivalent controller models to achieve control of infusion heating.

2. The infusion heating control method based on neural network and approximation internal model control according to claim 1, characterized in that: Step S1 includes: S101. The infusion heating system is heated at full power, and the infusion temperature and corresponding heating time are continuously collected within a set time period from the start of heating. The system response curve of the infusion heating system is constructed with heating time as the x-axis and infusion temperature as the y-axis. S102. Obtain infusion heating parameters based on the system response curve, including: Determine the initial temperature , that is, the infusion temperature at the start of heating; Determine the steady-state temperature That is, the infusion temperature tends to stabilize. If, during full-power heating, the infusion temperature remains stable for n consecutive preset time intervals... If the infusion temperature change does not exceed the preset threshold within n time intervals, then the n time intervals will be... The average infusion temperature is taken as the steady-state temperature; Based on the initial temperature steady-state temperature Intermediate temperature used in computational modeling , : ； ； intermediate temperature The corresponding time on the system response curve is denoted as . , to intermediate temperature The corresponding time on the system response curve is denoted as . ; Calculate the static gain of the temperature object The time constant of the temperature object Lag time of temperature object : ； ； 。 3. The infusion heating control method based on neural network and approximation internal model control according to claim 1, characterized in that: Step S2 includes: S201. Process of establishing the internal model of the infusion heating system : ； In the formula, This describes the actual control process of the infusion heating system. It is a complex variable; S202. Inner mold process Decomposed into minimum phase part and unstable zero-point part : ； ； S203. Introduce a first-order filter element: ； In the formula, These are the filter coefficients. For the filtering order, and Divide and establish an internal model controller : ； S204. Establish a mathematical model for the equivalent controller of the infusion system: ； in, Indicates the equivalent controller. Indicates the internal terminal controller. Indicates the internal model process, It is a complex variable; In order to make the internal mold controller The transfer function is a complex variable Let the rational proper fraction function be the order of the denominator of the low-pass filter. The value of ,Will , Substitute the value In the middle, we get: ； Equivalent controller By adjusting the control parameters This improves the control performance of the infusion heating system.

4. The infusion heating control method based on neural network and approximation internal model control according to claim 1, characterized in that: Step S3 includes: S301. Construct an approximation internal model control system based on neural networks and equivalent controller models, including a temperature deviation calculation module, an equivalent controller, a neural network model, an infusion heating control system, and a temperature sensor module; The temperature sensor module is used to collect the liquid temperature of the infusion tube in the infusion heating control system in real time, i.e., the actual temperature. The output of the temperature sensor module is connected to the temperature deviation calculation module and the neural network module respectively. The output of the temperature deviation calculation module is connected to the equivalent controller and the neural network model respectively. The output of the neural network model is connected to the control input of the equivalent controller. The output of the equivalent controller is connected to the infusion heating control system. Step S302. Based on the approximation internal model control system, control of the infusion heating process is achieved: The infusion tubing is attached to the temperature sensor module, and the current fluid temperature in the infusion tubing is sampled every 1 second, which is the actual temperature of the infusion heating. And perform the following procedure: (1) The data is fed into the temperature deviation calculation module and the neural network model. (2) Temperature deviation calculation module calculates the target temperature of the infusion heating system. Compared with actual temperature Temperature deviation between ; (3) The neural network calculation model is based on the target temperature Actual temperature and temperature deviation Dynamic output equivalent controller Control parameters in ; (4) Equivalent controller Control parameters based on neural network model output Temperature deviation output by the temperature deviation calculation module Output equivalent control information ; (5) Infusion heating control system, used to control the infusion based on equivalent control information This enables heating control, adjusting the actual infusion temperature, and theoretically... .

5. The infusion heating control method based on neural network and approximation internal model control according to claim 4, characterized in that: The neural network calculation model is based on the target temperature. Actual temperature and temperature deviation Dynamic output equivalent controller Control parameters in ,include: The neural network model is constructed using a backpropagation (BP) neural network structure with 3 nodes in the input layer, 5 nodes in the hidden layer, and 1 node in the output layer. The output of the i-th node in the input layer is : in, , , The superscript [] indicates the number of network layers. A network layer of 1 indicates an input layer, a network layer of 2 indicates a hidden layer, and a network layer of 3 indicates an output layer. Forward computation of the hidden layer of the network: in, Let l be the weighted input of the l-th node in the hidden layer. This is the output of the l-th node in the hidden layer. The weights connecting the i-th node in the input layer to the l-th node in the hidden layer; Forward computation of the network output layer: in, This represents the weighted input of the j-th node in the network output layer. The output of the j-th node in the network output layer, i.e., the control parameters. Since the network output layer has only one node, .

6. The infusion heating control method based on neural network and approximation internal model control according to claim 5, characterized in that: The neural network model uses gradient descent with momentum, following the negative gradient direction, and sets a learning rate. momentum coefficient For weight parameters Real-time adjustments are made, and the internal model equivalent controller is dynamically output. Control parameters in The adjustment method is as follows: Take the error function: ； Network output layer gradient: The backpropagation formula for the output layer is: in, Substituting the error of the third layer into the gradient of the network output layer, we get: Gradient of hidden layer: 。

Images