System and method for adjusting conductivity of dialysis liquid mixing device

Abstract

The invention discloses a dialysis liquid mixing device conductivity adjusting system and method. The dialysis liquid mixing device conductivity adjusting system comprises an RO water flow sensor, a B liquid conductivity sensor, a dialysate conductivity sensor C, an RO water pump, a B liquid infusion pump, an A liquid infusion pump, a B liquid feed-forward controller, a dialysate feed-forward controller, a B liquid parameter self-adaptive PID feedback controller and a dialysate parameter self-adaptive PID feedback controller. A feedforward + feedback control strategy is adopted to adaptively adjust parameter variables in the proportioning process, and large-batch automatic proportioning of dialysate can be achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical devices, and in particular to a conductivity regulating system and method for a dialysis liquid mixing device. Background Art

[0002] Currently, hemodialysis centers often use a centralized liquid distribution system to centrally prepare dialysate and then transport it to each dialysis site. Traditional liquid distribution systems mainly use open-loop proportional control, which has the following technical limitations:

[0003] 1. High model dependence: Control requires a precise mixing ratio model, and frequent manual calibration is required when the solution concentration changes;

[0004] 2. Weak anti-interference ability: Traditional PID parameters are fixed and cannot adapt to time-varying disturbances (such as differences in solution batches). Summary of the Invention

[0005] The purpose of the present invention is to provide a conductivity adjustment system and method for a dialysis mixing device in response to the problems and shortcomings described in the background technology.

[0006] A conductivity regulation system for a dialysis mixing device, comprising an RO water flow sensor, a B fluid conductivity sensor, a dialysate conductivity sensor C, an RO water pump, a B fluid infusion pump, an A fluid infusion pump, a B fluid feedforward controller, a dialysate feedforward controller, a B fluid parameter adaptive PID feedback controller, and a dialysate parameter adaptive PID feedback controller;

[0007] The RO water tank is connected to the B liquid mixer via an RO water pump. An RO water flow sensor is connected between the RO water pump and the B liquid mixer. A B liquid infusion pump is connected between the B liquid concentrate storage tank and the B liquid mixer. A B liquid conductivity sensor is installed inside the B liquid mixer.

[0008] The liquid A concentrate storage tank is connected to the liquid A mixer via the liquid A infusion pump. A dialysate conductivity sensor C is installed inside the liquid A mixer.

[0009] The liquid A mixer is connected to the liquid B mixer, and dialysate is formed after mixing, and the dialysate is finally formed in the liquid A mixer;

[0010] The RO water flow sensor and the B liquid infusion pump are electrically connected to the B liquid feedforward controller;

[0011] The RO water flow sensor and the A-liquid infusion pump are electrically connected to the dialysate feedforward controller;

[0012] The RO water flow sensor, the B liquid conductivity sensor, and the B liquid infusion pump are electrically connected to the B liquid parameter adaptive PID feedback controller;

[0013] The RO water flow sensor, the dialysate conductivity sensor C and the A-liquid infusion pump are electrically connected to the dialysate parameter self-adaptive PID feedback controller.

[0014] A method for adjusting the conductivity of a dialysis mixing device comprises the following steps:

[0015] B liquid infusion pump speed control:

[0016] Step 1: Signal acquisition: Real-time acquisition of RO water flow Q water 、B liquid mixer output solution conductivity C B ;

[0017] Step 2: Liquid B feedforward controller output, according to RO water flow rate n water Calculate the control speed n of the liquid B infusion pump B , the calculation formula is as follows:

[0018] n B =k B n water

[0019] Among them, k B is the ratio coefficient of liquid B, which is determined according to the concentration of liquid B concentrate and the target conductivity value of liquid B;

[0020] Step 3: The B liquid parameter adaptive PID feedback controller outputs the B liquid infusion pump control speed correction value;

[0021] Step 4: The output of the liquid B parameter adaptive PID feedback controller + the output of the liquid B feedforward controller are used to control the output speed control value of the liquid B infusion pump;

[0022] A liquid infusion pump speed control:

[0023] Step 5: Signal acquisition: Real-time acquisition of RO water flow Q water , the conductivity of the dialysate output from the A-liquid mixer C D ;

[0024] Step 6: Liquid A feedforward controller output, according to RO water flow rate n water Calculate the control speed n of the liquid A infusion pump A , the calculation formula is as follows:

[0025] n A =k A n water

[0026] Among them, k A is the ratio coefficient of liquid A, which is determined according to the concentration of liquid A concentrate, the concentration of liquid B concentrate and the target dialysate conductivity value;

[0027] Step 7: The A-liquid parameter adaptive PID feedback controller outputs the A-liquid infusion pump control speed correction value;

[0028] Step 8: Liquid A parameter adaptive PID feedback controller output + Liquid A feedforward controller output, control the output of Liquid A infusion pump.

[0029] Preferably, the specific correction method in step 3 is as follows:

[0030] 1. System model construction: Assume that the controlled object can be described by an autoregressive moving average (ARMA) model. The discrete-time system input is u(k) and the output is y(k). The model expression is:

[0031]

[0032] Among them, n a and n b are the orders of autoregression and sliding average, a i and b i are model parameters, ξ(k) is zero-mean white noise;

[0033] Convert it into vector form and define the parameter vector

[0034] Regression vector φ(k)=[-y(k-1),-y(k-2),…,-y(k-na),u(k),u(k-1),…,u(k-

[0035] nb)] T , then the system model can be expressed as y(k)=φ T (k)θ+ξ(k);

[0036] At the same time, the control law of the PID controller is:

[0037]

[0038] Where e(k) = r(k) - y(k) is the error, and r(k) is the set value;

[0039] 2. Initialization: Initialize parameter estimates Set it to zero vector, that is

[0040] Initialize the covariance matrix P(0) = αI, where α is a large positive number and I is the identity matrix;

[0041] Select the forgetting factor λ, the value range is 0<λ≤1, generally between 0.95-1;

[0042] Initialize PID parameter K p (0), Ki (0) and K d (0), can be set based on experience or trial and error;

[0043] Initialize the error integral and the error at the previous moment e(-1)=0;

[0044] 3. Sampling time operation: At each sampling time k, perform the following operations:

[0045] Data acquisition and error calculation: Measure the system set value r(k) and the actual output y(k), calculate the error e(k) = r(k) - y(k), and update the error integral

[0046] Calculate the regression vector: Calculate the regression vector φ based on the current and historical input and output data

[0047] (k)=[-y(k-1),-y(k-2),…,-y(k-na),u(k),u(k-1),…,u(k-nb)] T ;

[0048] Recursive least squares parameter estimation:

[0049] Calculate the gain vector:

[0050] Update parameter estimates:

[0051] Update the covariance matrix:

[0052] PID parameter adjustment: based on minimizing the sum of squared errors The goal is to adjust the PID parameters using the gradient descent method. First, the gradient of the error square sum function with respect to the PID parameters is calculated:

[0053] Approximately calculate the gradient, considering the contribution of the current moment error to the gradient,

[0054]

[0055] Where g is a constant obtained through system identification or empirical estimation;

[0056] Update PID parameters according to the gradient descent method:

[0057] K p (k+1)=K p (k)+η·2g·e 2 (k);

[0058]

[0059] Limit the updated PID parameters and set K p , K i and K d The upper and lower limits of

[0060] K p (k+1)=max(min(K p (k+1), K pmax ), K pmin ),

[0061] K i (k+1)=max(min(K i (k+1), K imax ), K imin ),

[0062] K d (k+1)=max(min(K d (k+1), K dmax ), K dmin ),

[0063] Where η is the learning rate, which ranges from 0 to 1;

[0064] Calculate the control quantity: According to the adjusted PID parameters Kp(k), K i (k) and K d (k), calculate the control quantity

[0065]

[0066] Output control quantity: Apply the calculated control quantity u(k) to the controlled object (liquid B infusion pump);

[0067] Update the time step: add 1 to the time step k, return to the data collection and error calculation steps, and continue the operation at the next sampling moment.

[0068] Preferably, the specific correction method in step 7 is as follows:

[0069] 1. System model construction: Assume that the controlled object (liquid A infusion pump) can be described by an autoregressive moving average (ARMA) model. The discrete-time system input is u(k) and the output is y(k). The model expression is:

[0070]

[0071] Among them, n a and n b are the orders of autoregression and sliding average, a i and b iare model parameters, ξ(k) is zero-mean white noise;

[0072] Convert it into vector form and define the parameter vector Regressor φ(k)=[-y(k-1),-y(k-2),…,-y(k-na),u(k),u(k-1),…,u(k-nb)] T , then the system model can be expressed as y(k)=φ T (k)θ+ξ(k);

[0073] At the same time, the control law of the PID controller is:

[0074]

[0075] Where e(k) = r(k) - y(k) is the error, and r(k) is the set value;

[0076] 2. Initialization: Initialize parameter estimates Set it to zero vector, that is

[0077] Initialize the covariance matrix P(0) = αI, where α is a large positive number (such as α = 10 6 ), I is the identity matrix;

[0078] Select the forgetting factor λ, the value range is 0<λ≤1, generally between 0.95-1;

[0079] Initialize PID parameter K p (0), K i (0) and K d (0), can be set based on experience or trial and error;

[0080] Initialize the error integral and the error at the previous moment e(-1)=0;

[0081] 3. Sampling time operation: At each sampling time k, perform the following operations:

[0082] Data acquisition and error calculation: Measure the system set value r(k) and the actual output y(k), calculate the error e(k) = r(k) - y(k), and update the error integral

[0083] Calculate the regression vector: Calculate the regression vector φ based on the current and historical input and output data

[0084] (k)=[-y(k-1),-y(k-2),…,-y(k-na),u(k),u(k-1),…,u(k-nb)] T ;

[0085] Recursive least squares parameter estimation:

[0086] Calculate the gain vector:

[0087] Update parameter estimates:

[0088] Update the covariance matrix:

[0089] PID parameter adjustment: based on minimizing the sum of squared errors The goal is to adjust the PID parameters using the gradient descent method. First, calculate the gradient of the error square sum function with respect to the PID parameters:

[0090] Approximately calculate the gradient, considering the contribution of the current moment error to the gradient,

[0091]

[0092] Where g is a constant obtained through system identification or empirical estimation;

[0093] Update PID parameters according to the gradient descent method:

[0094] K p (k+1)=K p (k)+η·2g·e 2 (k);

[0095]

[0096] Limit the updated PID parameters and set K p , K i and K d The upper and lower limits of

[0097] K p (k+1)=max(min(K p (k+1), K pmax ), K pmin ),

[0098] K i (k+1)=max(min(K i (k+1), K imax ), K imin ),

[0099] K d (k+1)=max(min(K d (k+1), K dmax ), K dmin ),

[0100] Where η is the learning rate, which ranges from 0 to 1;

[0101] Calculate the control quantity: According to the adjusted PID parameter K p (k), K i (k) and K d (k), calculate the control quantity

[0102]

[0103] Output control quantity: Apply the calculated control quantity u(k) to the controlled object (liquid A infusion pump);

[0104] Update the time step: add 1 to the time step k, return to the data collection and error calculation steps, and continue the operation at the next sampling moment.

[0105] Beneficial effects of the present invention:

[0106] The present invention can realize large-scale automated proportioning of dialysate, and can adaptively adjust parameter variables in the proportioning process; when the solution concentration changes, there is no need for frequent manual calibration, and the infusion pump speed can be automatically adjusted according to the actual situation, which significantly improves the system's adaptability to different solution concentrations. When the RO water flow rate fluctuates, it can also respond quickly, suppress interference, and ensure that the dialysate conductivity is always stable within the target value range, greatly improving the accuracy and stability of dialysate preparation, and providing reliable quality dialysate for hemodialysis treatment. The parameter adaptive adjustment mechanism can automatically adapt to changes during the operation of the system, reducing the technical threshold and work intensity of the operator. In terms of system maintenance, since the frequent calibration and adjustment work caused by model dependence and fixed parameters is reduced, the system maintenance is simpler, the maintenance cost is reduced, and the reliability and maintainability of the equipment are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0107] Figure 1 This is a block diagram of the structure of the dialysis mixing device;

[0108] Figure 2 This is the conductivity regulation control block diagram; DETAILED DESCRIPTION

[0109] A conductivity regulation system for a dialysis mixing device, comprising an RO water flow sensor, a B fluid conductivity sensor, a dialysate conductivity sensor, an RO water pump, a B fluid infusion pump, an A fluid infusion pump, a B fluid feedforward controller, a dialysate feedforward controller, a B fluid parameter adaptive PID feedback controller, and a dialysate parameter adaptive PID feedback controller;

[0110] The RO water tank is connected to the B liquid mixer via an RO water pump. An RO water flow sensor is connected between the RO water pump and the B liquid mixer. A B liquid infusion pump is connected between the B liquid concentrate storage tank and the B liquid mixer. A B liquid conductivity sensor is installed inside the B liquid mixer.

[0111] The liquid A concentrate storage tank is connected to the liquid A mixer via the liquid A infusion pump. A dialysate conductivity sensor C is installed inside the liquid A mixer.

[0112] The liquid A mixer is connected to the liquid B mixer, and dialysate is formed after mixing, and the dialysate is finally formed in the liquid A mixer;

[0113] The RO water flow sensor and the B liquid infusion pump are electrically connected to the B liquid feedforward controller;

[0114] The RO water flow sensor and the A-liquid infusion pump are electrically connected to the dialysate feedforward controller;

[0115] The RO water flow sensor, the B liquid conductivity sensor, and the B liquid infusion pump are electrically connected to the B liquid parameter adaptive PID feedback controller;

[0116] The RO water flow sensor, the dialysate conductivity sensor C and the A-liquid infusion pump are electrically connected to the dialysate parameter self-adaptive PID feedback controller.

[0117] Determine the RO water flow rate based on the dialysate dosage requirement, control the RO water pump to meet the RO water flow requirement, determine the B solution concentration and dialysate concentration based on the ion concentration requirement in the dialysate, and then determine the B solution conductivity setting value and the dialysate conductivity setting value;

[0118] The entire system adopts a feedforward + feedback control strategy, in which the feedforward control determines the control speed of the B liquid infusion pump and the A liquid infusion pump based on the RO water flow rate measured in real time by the RO water flow sensor. That is, the speed of the B liquid infusion pump is controlled according to the B liquid conductivity set value and the real-time RO water flow measurement value, and the speed of the A liquid infusion pump is controlled according to the dialysate conductivity set value and the real-time RO water flow measurement value.

[0119] However, during the real-time control process, due to fluctuations in the RO water flow rate or changes in the concentrations of saturated solution A (liquid in the A concentrate storage tank) and saturated solution B (liquid in the B concentrate storage tank), the actual controlled B fluid conductivity and dialysate conductivity may deviate from the set values. In this case, the system will perform feedback control on the speeds of the B fluid infusion pump and the A fluid infusion pump based on the B fluid conductivity deviation value and the dialysate conductivity deviation value. The basic control logic is shown in Table 1.

[0120] Table 1 Speed ​​control logic of liquid B infusion pump and liquid A infusion pump

[0121] Where: C BOis the conductivity setting value of liquid B, C B The conductivity value of the solution output by the B liquid mixer is the value detected in real time by the B liquid conductivity sensor. DO is the dialysate conductivity setting value, C D is the conductivity value of the dialysate, that is, the value detected in real time by the dialysate conductivity sensor, n B Control the speed of the liquid B infusion pump, n A Control the speed of the liquid A infusion pump.

[0122]

[0123]

[0124] A method for adjusting the conductivity of a dialysis mixing device comprises the following steps:

[0125] B liquid infusion pump speed control:

[0126] Step 1: Signal acquisition: Real-time acquisition of RO water flow Q water 、B liquid mixer output solution conductivity C B ;

[0127] Step 2: Liquid B feedforward controller output, according to RO water flow rate n water Calculate the control speed n of the liquid B infusion pump B , the calculation formula is as follows:

[0128] n B =k B n water

[0129] Among them, k B is the ratio coefficient of liquid B, which is determined according to the concentration of liquid B concentrate and the target conductivity value of liquid B;

[0130] Step 3: The B liquid parameter adaptive PID feedback controller outputs the B liquid infusion pump control speed correction value;

[0131] The specific correction methods are as follows:

[0132] 1. System model construction: Assume that the controlled object (liquid B infusion pump) can be described by an autoregressive moving average (ARMA) model. The discrete-time system input is u(k) and the output is y(k). The model expression is:

[0133]

[0134] Among them, n a and n b are the orders of autoregression and sliding average, a i and b iare model parameters, ξ(k) is zero-mean white noise;

[0135] Convert it into vector form and define the parameter vector

[0136] Regression vector φ(k)=[-y(k-1),-y(k-2),…,-y(k-na),u(k),u(k-1),…,u(k-

[0137] nb)] T , then the system model can be expressed as y(k)=φ T (k)θ+ξ(k);

[0138] At the same time, the control law of the PID controller is:

[0139]

[0140] Where e(k) = r(k) - y(k) is the error, and r(k) is the set value;

[0141] 2. Initialization: Initialize parameter estimates Set it to zero vector, that is

[0142] Initialize the covariance matrix P(0) = αI, where α is a large positive number (such as α = 10 6 ), I is the identity matrix;

[0143] Select the forgetting factor λ, the value range is 0<λ≤1, generally between 0.95-1;

[0144] Initialize PID parameter K p (0), K i (0) and K d (0), can be set based on experience or trial and error;

[0145] Initialize the error integral and the error at the previous moment e(-1)=0;

[0146] 3. Sampling time operation: At each sampling time k, perform the following operations:

[0147] Data acquisition and error calculation: Measure the system set value r(k) and the actual output y(k), calculate the error e(k) = r(k) - y(k), and update the error integral Calculate the regression vector: Based on the current and historical input and output data, calculate the regression vector φ(k) = [-y(k-1), -y(k-2), ..., -y(k-na), u(k), u(k-1), ..., u(k-nb)] T; Recursive least squares parameter estimation:

[0148] Calculate the gain vector: Update parameter estimates: Update the covariance matrix: PID parameter adjustment: based on minimizing the sum of squared errors The goal is to adjust the PID parameters using the gradient descent method. First, the gradient of the error square sum function with respect to the PID parameters is calculated: the gradient is calculated approximately, and the contribution of the error at the current moment to the gradient is considered.

[0149]

[0150] Where g is a constant obtained through system identification or empirical estimation;

[0151] Update PID parameters according to the gradient descent method:

[0152] K p (k+1)=K p (k)+η·2g·e 2 (k);

[0153]

[0154] Limit the updated PID parameters and set K p , K i and K d The upper and lower limits, that is, K p (k+1)=max(min(K p (k+1), K pmax ), K pmin ),

[0155] K i (k+1)=max(min(K i (k+1), K imax ), K imin ),

[0156] K d (k+1)=max(min(K d (k+1), K dmax ), K dmin ),

[0157] Where η is the learning rate, which ranges from 0 to 1;

[0158] Calculate the control quantity: According to the adjusted PID parameters Kp(k), K i (k) and K d (k), calculate the control quantity

[0159]

[0160] Output control quantity: Apply the calculated control quantity u(k) to the controlled object (liquid B infusion pump); Update time step: Add 1 to the time step k, return to the data acquisition and error calculation steps, and continue the operation at the next sampling time;

[0161] Step 4: The output of the liquid B parameter adaptive PID feedback controller + the output of the liquid B feedforward controller are used to control the output speed control value of the liquid B infusion pump.

[0162] A liquid infusion pump speed control:

[0163] Step 5: Signal acquisition: Real-time acquisition of RO water flow Q water , the conductivity of the dialysate output from the A-liquid mixer C D ;

[0164] Step 6: Liquid A feedforward controller output, according to RO water flow rate n water Calculate the control speed n of the liquid A infusion pump A , the calculation formula is as follows:

[0165] n A =k A n water

[0166] Among them, k A is the ratio coefficient of liquid A, which is determined according to the concentration of liquid A concentrate, the concentration of liquid B concentrate and the target dialysate conductivity value;

[0167] Step 7: The A-liquid parameter adaptive PID feedback controller outputs the A-liquid infusion pump control speed correction value;

[0168] The specific correction methods are as follows:

[0169] 1. System model construction: Assume that the controlled object (liquid A infusion pump) can be described by an autoregressive moving average (ARMA) model. The discrete-time system input is u(k) and the output is y(k). The model expression is:

[0170]

[0171] Among them, n a and n b are the orders of autoregression and sliding average, a i and b i are model parameters, ξ(k) is zero-mean white noise;

[0172] Convert it into vector form and define the parameter vector Regressor φ(k)=[-y(k-1),-y(k-2),…,-y(k-na),u(k),u(k-1),…,u(k-nb)] T ,

[0173] Then the system model can be expressed as y(k)=φ T (k)θ+ξ(k);

[0174] At the same time, the control law of the PID controller is:

[0175]

[0176] Where e(k) = r(k) - y(k) is the error, and r(k) is the set value;

[0177] 2. Initialization: Initialize parameter estimates Set it to zero vector, that is

[0178] Initialize the covariance matrix P(0) = αI, where α is a large positive number (such as α = 10 6 ), I is the identity matrix;

[0179] Select the forgetting factor λ, the value range is 0<λ≤1, generally between 0.95-1;

[0180] Initialize PID parameter K p (0), K i (0) and K d (0), can be set based on experience or trial and error;

[0181] Initialize the error integral and the error at the previous moment e(-1)=0;

[0182] 3. Sampling time operation: At each sampling time k, perform the following operations:

[0183] Data acquisition and error calculation: Measure the system set value r(k) and the actual output y(k), calculate the error e(k) = r(k) - y(k), and update the error integral

[0184] Calculate the regression vector: Calculate the regression vector φ based on the current and historical input and output data

[0185] (k)=[-y(k-1),-y(k-2),…,-y(k-na),u(k),u(k-1),…,u(k-nb)] T ;

[0186] Recursive least squares parameter estimation:

[0187] Calculate the gain vector:

[0188] Update parameter estimates:

[0189] Update the covariance matrix:

[0190] PID parameter adjustment: based on minimizing the sum of squared errors The goal is to adjust the PID parameters using the gradient descent method. First, calculate the gradient of the error square sum function with respect to the PID parameters:

[0191] Approximately calculate the gradient, considering the contribution of the current moment error to the gradient,

[0192]

[0193] Where g is a constant obtained through system identification or empirical estimation;

[0194] Update PID parameters according to the gradient descent method:

[0195] K p (k+1)=K p (k)+η·2g·e 2 (k);

[0196]

[0197] Limit the updated PID parameters and set K p , K i and K d The upper and lower limits of

[0198] K p (k+1)=max(min(K p (k+1), K pmax ), K pmin ),

[0199] K i (k+1)=max(min(K i (k+1), K imax ), K imin ),

[0200] K d (k+1)=max(min(K d (k+1), K dmax ), K dmin ),

[0201] Where η is the learning rate, which ranges from 0 to 1;

[0202] Calculate the control quantity: According to the adjusted PID parameter K p (k), K i (k) and K d (k), calculate the control quantity

[0203]

[0204] Output control quantity: Apply the calculated control quantity u(k) to the controlled object (liquid A infusion pump);

[0205] Update time step: add 1 to time step k, return to the data collection and error calculation steps, and continue the operation at the next sampling moment;

[0206] Step 8: Liquid A parameter adaptive PID feedback controller output + Liquid A feedforward controller output, control the output of Liquid A infusion pump.

Claims

1. A conductivity adjustment system for a dialysis mixing device, characterized by: Including RO water flow sensor, B liquid conductivity sensor, dialysate conductivity sensor C, RO water pump, B liquid infusion pump, A liquid infusion pump, B liquid feedforward controller, dialysate feedforward controller, B liquid parameter adaptive PID feedback controller, dialysate parameter adaptive PID feedback controller; The RO water tank is connected to the B liquid mixer via an RO water pump. An RO water flow sensor is connected between the RO water pump and the B liquid mixer. A B liquid infusion pump is connected between the B liquid concentrate storage tank and the B liquid mixer. A B liquid conductivity sensor is installed inside the B liquid mixer. The liquid A concentrate storage tank is connected to the liquid A mixer via the liquid A infusion pump. A dialysate conductivity sensor C is installed inside the liquid A mixer. The liquid A mixer is connected to the liquid B mixer, and dialysate is formed after mixing, and the dialysate is finally formed in the liquid A mixer; The RO water flow sensor and the B liquid infusion pump are electrically connected to the B liquid feedforward controller; The RO water flow sensor and the A-liquid infusion pump are electrically connected to the dialysate feedforward controller; The RO water flow sensor, the B liquid conductivity sensor, and the B liquid infusion pump are electrically connected to the B liquid parameter adaptive PID feedback controller; The RO water flow sensor, the dialysate conductivity sensor C and the A-liquid infusion pump are electrically connected to the dialysate parameter self-adaptive PID feedback controller.

2. A method for adjusting the conductivity of a dialysis mixing device, characterized in that: The steps include: B liquid infusion pump speed control: Step 1: Signal acquisition: Real-time acquisition of RO water flow Q water 、B liquid mixer output solution conductivity C B ; Step 2: Liquid B feedforward controller output, according to RO water flow rate n water Calculate the control speed n of the liquid B infusion pump B , the calculation formula is as follows: n B =k B n water Among them, k B is the ratio coefficient of liquid B, which is determined according to the concentration of liquid B concentrate and the target conductivity value of liquid B; Step 3: The B liquid parameter adaptive PID feedback controller outputs the B liquid infusion pump control speed correction value; Step 4: The output of the liquid B parameter adaptive PID feedback controller + the output of the liquid B feedforward controller are used to control the output speed control value of the liquid B infusion pump; A liquid infusion pump speed control: Step 5: Signal acquisition: Real-time acquisition of RO water flow Q water , the conductivity of the dialysate output from the A-liquid mixer C D ; Step 6: Liquid A feedforward controller output, according to RO water flow rate n water Calculate the control speed n of the liquid A infusion pump A , the calculation formula is as follows: n A =k A n water Among them, k A is the ratio coefficient of liquid A, which is determined according to the concentration of liquid A concentrate, the concentration of liquid B concentrate and the target dialysate conductivity value; Step 7: The A-liquid parameter adaptive PID feedback controller outputs the A-liquid infusion pump control speed correction value; Step 8: Liquid A parameter adaptive PID feedback controller output + Liquid A feedforward controller output, control the output of Liquid A infusion pump.

3. The method for adjusting the conductivity of a dialysis mixing device according to claim 2, wherein: The specific correction method in step 3 is as follows:

1. System model construction: Assume that the controlled object can be described by an autoregressive moving average (ARMA) model. The discrete-time system input is u(k) and the output is y(k). The model expression is: Among them, n a and n b are the orders of autoregression and sliding average, a i and b i are model parameters, ξ(k) is zero-mean white noise; Convert it into vector form and define the parameter vector Regression vector φ(k)=[-y(k-1),-y(k-2),…,-y(k-na),u(k),u(k-1),…,u(k-nb)] T , then the system model can be expressed as y(k)=φ T (k)θ+ξ(k); At the same time, the control law of the PID controller is: Where e(k) = r(k) - y(k) is the error, and r(k) is the set value; 2. Initialization: Initialize parameter estimates Set it to zero vector, that is Initialize the covariance matrix P(0) = αI, where α is a large positive number and I is the identity matrix; Select the forgetting factor λ, the value range is 0<λ≤1; Initialize PID parameter K p (0), K i (0) and K d (0), can be set based on experience or trial and error; Initialize the error integral and the error at the previous moment e(-1)=0; 3. Sampling time operation: At each sampling time k, perform the following operations: Data acquisition and error calculation: Measure the system set value r(k) and the actual output y(k), calculate the error e(k) = r(k) - y(k), and update the error integral Calculate the regression vector: Based on the current and historical input and output data, calculate the regression vector φ(k) = [-y(k-1), -y(k-2), ..., -y(k-na), u(k), u(k-1), ..., u(k-nb)] T ; Recursive least squares parameter estimation: Calculate the gain vector: Update parameter estimates: Update the covariance matrix: PID parameter adjustment: based on minimizing the sum of squared errors The goal is to adjust the PID parameters using the gradient descent method. First, the gradient of the error square sum function with respect to the PID parameters is calculated: Approximately calculate the gradient, considering the contribution of the current moment error to the gradient, Where g is a constant obtained through system identification or empirical estimation; Update PID parameters according to the gradient descent method: K p (k+1)=K p (k)+η·2g·e 2 (k); K d (k+1)=Kd(k)+η·2g·e(k)[e(k)-e(k-1)]; Limit the updated PID parameters and set K p , K i and K d The upper and lower limits of K p (k+1)=max(min(K p (k+1),K pmax ),K pmin ), K i (k+1)=max(min(K i (k+1),K imax ),K imin ), K d (k+1)=max(min(K d (k+1),K dmax ),K dmin ), Where η is the learning rate, which ranges from 0 to 1; Calculate the control quantity: According to the adjusted PID parameters Kp(k), K i (k) and K d (k), calculate the control quantity Output control quantity: Apply the calculated control quantity u(k) to the controlled object (liquid B infusion pump); Update the time step: add 1 to the time step k, return to the data collection and error calculation steps, and continue the operation at the next sampling moment.

4. The method for adjusting the conductivity of a dialysis mixing device according to claim 3, wherein: The specific correction method in step 7 is as follows:

1. System model construction: Assume that the controlled object can be described by an autoregressive moving average (ARMA) model. The discrete-time system input is u(k) and the output is y(k). The model expression is: Among them, n a and n b are the orders of autoregression and sliding average, a i and b i are model parameters, ξ(k) is zero-mean white noise; Convert it into vector form and define the parameter vector Regressor φ(k)=[-y(k-1),-y(k-2),…,-y(k-na),u(k),u(k-1),…,u(k-nb)] T , then the system model can be expressed as y(k)=φ T (k)θ+ξ(k); At the same time, the control law of the PID controller is: Where e(k) = r(k) - y(k) is the error, and r(k) is the set value; 2. Initialization: Initialize parameter estimates Set it to zero vector, that is Initialize the covariance matrix P(0) = αI, where α is a large positive number and I is the identity matrix; Select the forgetting factor λ, the value range is 0<λ≤1; Initialize PID parameter K p (0), K i (0) and K d (0), can be set based on experience or trial and error; Initialize the error integral and the error at the previous moment e(-1)=0; 3. Sampling time operation: At each sampling time k, perform the following operations: Data acquisition and error calculation: Measure the system set value r(k) and the actual output y(k), calculate the error e(k) = r(k) - y(k), and update the error integral Calculate the regression vector: Calculate the regression vector φ based on the current and historical input and output data (k)=[-y(k-1),-y(k-2),…,-y(k-na),u(k),u(k-1),…,u(k-nb)] T ; Recursive least squares parameter estimation: Calculate the gain vector: Update parameter estimates: Update the covariance matrix: PID parameter adjustment: based on minimizing the sum of squared errors The goal is to adjust the PID parameters using the gradient descent method. First, calculate the gradient of the error square sum function with respect to the PID parameters: Approximately calculate the gradient, considering the contribution of the current moment error to the gradient, Where g is a constant obtained through system identification or empirical estimation; Update PID parameters according to the gradient descent method: K p (k+1)=K p (k)+η·2g·e 2 (k); Limit the updated PID parameters and set K p , K i and K d The upper and lower limits of K p (k+1)=max(min(K p (k+1),K pmax ),K pmin ), K i (k+1)=max(min(K i (k+1),K imax ),K imin ), K d (k+1)=max(min(K d (k+1),K dmax ),K dmin ), Where η is the learning rate, which ranges from 0 to 1; Calculate the control quantity: According to the adjusted PID parameter K p (k), K i (k) and K d (k), calculate the control quantity Output control quantity: Apply the calculated control quantity u(k) to the controlled object (liquid A infusion pump); Update the time step: add 1 to the time step k, return to the data collection and error calculation steps, and continue the operation at the next sampling moment.

Images