Intelligent control method for traditional Chinese medicine concentration process

Abstract

The invention discloses an intelligent control method for a traditional Chinese medicine concentration process and relates to an intelligent control method. The invention aims to solve the problem that a method for intelligently regulating and controlling the vacuum degree, the temperature and the online detection density of an evaporation chamber does not exist at present. The method comprises the following steps: step 1, establishing a concentration process temperature control model and a concentrated solution density soft measurement model; 2, designing a fuzzy rule in a temperature control model in the concentration process, and constructing a simulation model of a concentration temperature fuzzy self-tuning controller; introducing a fuzzy self-tuning control algorithm controlled by an expert; 3, performing simulation implementation and training on the concentrated solution density soft measurement model through matlab programming; continuously optimizing the obtained model result; adding a dynamic factor and introducing a genetic algorithm to optimize a local minimum value of the neural network and two points of a weight domain respectively; 4, whether the density is qualified or not is judged through a PLC program, and concentration and medicine discharging are conducted. The invention belongs to the technical field of traditional Chinese medicine production.

Description

Technical Field

[0001] The invention relates to an intelligent control method and belongs to the technical field of traditional Chinese medicine production. Background Art

[0002] The most important processes in the production process of traditional Chinese medicine are extraction, concentration and alcohol precipitation. These three processes involve a variety of physical changes and are a complex dynamic process system. In a traditional Chinese medicine production process, the soaked raw medicinal materials are first placed in the extraction tank for cyclic decoction to extract the available components in the liquid as much as possible, and at the same time, the volatile oil is separated by heating and condensation for use in the preparation of ingredients; the next step is to heat and concentrate the extracted liquid medicine to evaporate excess water to obtain the required liquid medicine that meets the density requirements; the subsequent alcohol precipitation stage is to separate the impurities in the liquid medicine through alcohol, and use the property that impurities are not easily soluble in alcohol to separate the impurity layer and the liquid medicine layer to obtain a purely refined liquid medicine, and finally the ingredients are packaged to obtain the finished liquid medicine that meets the process requirements.

[0003] During the concentration process, the vacuum degree and temperature of the evaporation chamber and the online density detection need to be adjusted to ensure that qualified traditional Chinese medicine products can be obtained in the end. However, there is currently no method that can intelligently control the vacuum degree and temperature of the evaporation chamber and the online density detection. Summary of the invention

[0004] The present invention aims to solve the problem that there is no method currently capable of intelligently regulating the vacuum degree and temperature of an evaporation chamber and online detection density, and further proposes an intelligent control method for a traditional Chinese medicine concentration process.

[0005] The technical solution adopted by the present invention to solve the above-mentioned problem is: the steps of the present invention include:

[0006] Step 1: Establish a temperature control model for the concentration process and a soft measurement model for the density of the concentrate;

[0007] Step 2: Design the fuzzy rules in the temperature control model during the concentration process. According to the formulated fuzzy rules, use the simulation editor in Matlab2018a to build a simulation model of the concentration temperature fuzzy self-tuning controller; introduce the fuzzy self-tuning control algorithm of expert control;

[0008] Step 3, simulate and train the soft measurement model of the concentrated liquid density through matlab programming; analyze the simulation results and continue to optimize the obtained model results; introduce the dynamic factor and the genetic algorithm to optimize the local minimum value and weight range of the neural network respectively;

[0009] Step 4: Use the PLC program to determine whether the density is qualified and concentrate the medicine.

[0010] Furthermore, the temperature control model of the concentration process constructed in step 1 specifically includes:

[0011] The state equation is constructed as:

[0012]

[0013] In formula (1), is the room temperature elastic coefficient, T is the evaporation chamber temperature, u p is the control action of the steam valve, τ is the time delay of the control function damaging the room temperature, α and β are the structural parameters of the evaporator;

[0014] The pressure equation is constructed as:

[0015]

[0016] In formula (2), P is the pressure inside the evaporator, u is the control function of the vacuum valve, and ε and θ are the structural parameters of the evaporator;

[0017] Let T = x 1 ,P=x 2 , combining equations (1) and (2), the state equation is:

[0018]

[0019] Perform Laplace transform on formula (3) and get G 1 (s) and G 2 (s) are temperature mathematical model and pressure mathematical model respectively:

[0020]

[0021] Only the temperature is studied, and the temperature control model obtained by formula (4) is:

[0022]

[0023] In formula (5), K is the static gain, τ is the lag time, and T is the coefficient of the inertia link.

[0024] Furthermore, the step of establishing a soft measurement model for the density of the concentrated liquid in step 1 includes:

[0025] The relationship between output density D and variable is:

[0026] D=F(P 1 ,P 2 ,T 1 ,H 1 )(1),

[0027] In formula (1), P 1Indicates the steam pipeline pressure, P 2 Indicates the vacuum degree of the first-effect evaporation chamber, T 1 Indicates the temperature of the first-effect evaporation chamber, H 1 represents the liquid level of the first-effect evaporation chamber; select P in formula (1) 1 , P 2 、T 1 , H 1 The input layer of the neural network constructed with four parameters as input of density soft measurement model is composed of these four auxiliary variables, and the output layer is the dominant variable density D. The input nodes of this model are 4, the output node is 1, and a single hidden layer structure is adopted to construct a three-layer BP neural network model;

[0028] The function that acts as a link in the forward propagation process is f(x), and the output layer is:

[0029]

[0030] The hidden layer output equation is:

[0031]

[0032] The activation function is a unipolar Sigmoid function with an infinite domain and a range of (0,1), and a continuously differentiable function definition:

[0033]

[0034] Import x 0 =-1, b 0 =-1, simplifying formulas (2) and (3) to:

[0035]

[0036] The modified derivation process of weights is as follows:

[0037] The output error is:

[0038]

[0039] Formula (7) is back-propagated to the hidden layer to obtain:

[0040]

[0041] Formula (8) is expanded back to the input layer to obtain:

[0042]

[0043] There is a functional relationship between the error and the weights of each layer. The error size can be changed by adjusting the weights, thereby continuously reducing the error value.

[0044]

[0045] In formulas (10) and (11), the number η is a proportional coefficient with a value range of (0,1). The BP back propagation algorithm is an error signal δ algorithm based on gradient descent;

[0046] Formulas (10) and (11) are expanded to:

[0047]

[0048] Introducing error signals into the output layer and hidden layer and Formulas (10) and (11) are optimized as follows:

[0049] Δw j1 =ηδ 1 y b j (14)

[0050] Δv ij =ηδ j b x i (15)

[0051] The weight is proportional to the input. Expand the output layer and hidden layer separately, and we have:

[0052]

[0053] Combining the error expressions (8) and (9), we have:

[0054]

[0055] Combined with the derivative of the S activation function, the proportional coefficient is obtained:

[0056] δ 1 y =(dy)y(1-y)(20),

[0057] δ j b =[(dy)f'(net 1 ) j1 ]f'(net j )=(δ 1 y w j1 )b j (1-b j )(twenty one),

[0058] Substituting into formula (20) and (21), we get the weight modification equations for the hidden layer and the output layer respectively:

[0059] Δw j1 =ηδ 1 y b j＝ η(dy)y(1-y)b j (twenty two),

[0060] Δv ij =ηδ j b x i＝ η(δ 1 y w j1 )b j (1-b j )x i (twenty three),

[0061] The BP neural network continuously modifies the weights according to the above formula, and after multiple iterations, the error is reduced to obtain the best prediction output; the global sum of squares error E of the output of the concentrated density soft sensor model is q have:

[0062]

[0063] Furthermore, in step 2, based on the tracking of the deviation and the rate of change of the deviation during the control process, the self-tuning law of the fuzzy PID parameters is summarized as follows:

[0064] In the response stage, the error e is large, so the Kp value should be increased and the Kd value should be decreased to shorten the response time of the control system, and Ki=0 should be taken to suppress overshoot; in the following stage, the error e is moderate, and the moderate Ki and Kd values ​​should be selected to ensure a faster adjustment speed, and a smaller Kp value should be used to reduce overshoot; in the adjustment stage, the error e is small, and the Kp and Ki values ​​should be appropriately increased to reduce the oscillation of the system and improve the stability of the controller. At the same time, when ec is small, the Kd value is slightly larger, and when ec is large, the Kd value is reduced;

[0065] In the fuzzy PID temperature control system, the fuzzy rules meet the following conditions: when the temperature is too high, the temperature can be reduced significantly to make it close to the expected value, and the control amount should be increased at this time; on the contrary, when the temperature value is close to the expected value, a smaller control amount is applied to slow down the control rate and make the control system reach a stable state;

[0066] The two input values ​​of the designed fuzzy controller are error e and error change rate ec, while Kp, Ki, and Kd are its three outputs. Through the quantization of the fuzzy controller, the maximum membership method is used to obtain accurate output values, thereby adjusting the PID coefficients, accumulating and outputting: Kp = Kp' + ΔKp, Ki = Ki' + ΔKi, Kd = Kd' + ΔKd, acting on the controlled object, and finally obtaining the output graph;

[0067] Through the error changes and rules, the expert system rules are introduced as follows:

[0068] The constant A is greater than the constant B, ka is the amplification factor, ka>1, kb is the suppression factor, 0<kb<1, and em(k) is the deviation extreme value;

[0069] 1. The absolute value of the error e(k) is greater than a constant value A: At this time, no matter how the error changes, the maximum or minimum value of the controller can be directly output to obtain the fastest error adjustment, allowing the error to reduce its absolute value at the fastest rate;

[0070] 2. The absolute value of the error e(k) keeps increasing or is a fixed constant. At this time, e(k)ec(k)>0: If the error value is relatively large, greater than the constant B, a stronger controller is selected:

[0071] u(k)=u(k-1)+k a {[e(k)-e(k-1)]k p}+e(k)k i +[e(k)-2e(k-1)+e(k-2)]k d (6)

[0072] When the absolute value of the error increases but the error itself is very small, less than the smaller constant B, a weaker controller is selected to make the error change in the opposite direction and the absolute value decrease:

[0073] u(k)=u(k-1)+[e(k)-e(k-1)]k p +e(k)k i +[e(k)-2e(k-1)+e(k-2)]k d (7)

[0074] 3. The absolute value of the error e(k) keeps decreasing: the controller output can be kept unchanged to achieve dynamic balance;

[0075] Expressed as a control rule: if e(k)ec(k)<0and e(k)ec(k-1)>0,thenΔK p =ΔK i =ΔK d =0;

[0076] 4. When the error e(k) is at an extreme value: the control output is determined based on the deviation value at time k. When the absolute value of the deviation at this moment is greater than B, the control regulation is stronger:

[0077] u(k)=u(k-1)+k a e m (k)·k p (8)

[0078] When the absolute value of the deviation is less than B, the control regulation is weak:

[0079] u(k)=u(k-1)+k b e m (k)·k p (9)

[0080] 5. The absolute value of the error e(k) is less than a very small value: saturation needs to be reduced. The solution is to add an integral separation effect, and its mathematical expression is:

[0081]

[0082] In formula (10),

[0083] M+N∈Xe, obviously, with the change of deviation, the value changes continuously in the range of (0, 1), which reduces the steady-state error and has stronger anti-saturation performance;

[0084] Expressed as a control rule: if |e(k)|<ε, then ΔK p =ΔK i =ΔK d =0;

[0085] When the temperature deviation is different, select different control methods. When the deviation and the deviation change rate meet three and five, select expert control, and select fuzzy PID control for the rest.

[0086] Furthermore, the soft-sensing model of the BP neural network in the concentration section in step 3 is programmed using MATLAB, and is run and simulated in conjunction with its internal neural network toolbox. The specific steps include:

[0087] Step 1: Data processing;

[0088] Step 2: BP neural network design, training and testing;

[0089] Step 3: Performance evaluation of the neural network model;

[0090] Step 4: Run the Matlab program to obtain the simulation result curve.

[0091] Furthermore, the introduction of momentum factor in step 3 can pass through the minimum value, and add a value proportional to the previous weight change to each weight change to generate a new weight change. The value modification formula including the momentum factor is as follows:

[0092] ΔW(k+1)=(1-mc)ηδX+mcΔW(k)(26),

[0093] In formula (26), W is the weight matrix of a certain layer, X is the input vector of a certain layer, k is the number of training times, and mc is the introduced momentum factor; the value range of mc is 0 to 1. When mc = 0, the change of weight is adjusted according to the error decrease, and when mc = 1, the next weight change is the same as the previous weight change; the value of the momentum factor is 0.9, and the network model selects mc = 0.95. At this time, the judgment condition of the momentum method is:

[0094]

[0095] In formula (27), SSE(k) is the error, and the error change rate is 1.04.

[0096] Furthermore, in step 3, a genetic algorithm is introduced to optimize the weights and thresholds of the network. The specific steps include:

[0097] Step (a), chromosome encoding and population initialization: design and build a neural network structure model according to input and output variables, determine the number of input and output nodes and the overall number of layers, the input layer node of the density soft measurement BP network is 4, the output layer node is 1, and the number of layers is three. All weights and thresholds of the BP neural network are encoded, and the length of the code is calculated as:

[0098] S = n × m + m × l + m + l (28),

[0099] In formula (28), n is the input layer node, m is the hidden layer node, n×m is the total weight of the input layer and the hidden layer, and l is the output layer node, then m×l is the total weight of the hidden layer and the output layer, m is the threshold of the hidden layer, l is the threshold of the output layer, n=4, m=10, l=1 determined above, the calculated coding length is 61, and the population size is 50;

[0100] Step (ii), select the fitness function; the fitness function is:

[0101]

[0102] The larger the value of the fitness function, the smaller the deviation value of the neural network model, and the better the adaptability of the neural network;

[0103] Step (iii), performing genetic manipulation; the steps of performing genetic manipulation are:

[0104] Step a: Perform selection operation: The probability of an individual being selected is based on the proportion of its fitness value, and the selection is carried out in a roulette wheel manner. The larger the proportion in the roulette wheel, the better the fitness; if k is the total number of individuals in the population, f i is the value of an individual’s fitness, so the probability of the gene being selected can be expressed as:

[0105]

[0106] Step b: Perform crossover operation: generate new evolved individuals by crossovering certain positions of two individuals, improve the structure of individuals, and the crossover probability is b. c The value of is selected in the interval [0,1];

[0107] Step c, perform mutation operation: The mutation operation selects gene points in the parent sample individuals and replaces them with random numbers to ensure the diversity of population individuals and improve the search ability of genetic algorithms applied to the global network. The mutation probability b m Take the value in [0,1], and choose 0.1;

[0108] Step (iv), loop operation; continuously loop step (ii) and step (iii), repeatedly train the neural network until the result of genetic neural network training meets the error requirement or the number of iterations reaches the target; the individual records the weight and threshold of each iteration, and updates them once after one cycle; genetic operation is used to continuously update and correct the weight and threshold to obtain the optimal individual value, so that the neural network estimation result is more accurate.

[0109] The beneficial effects of the present invention are as follows: the present invention combines fuzzy control theory with conventional PID algorithm by fuzzy self-tuning PID control, and derives fuzzy mathematical relationships between three parameters Kp, Ki, and Kd of the PID controller and the deviation e and the deviation change rate ec through experience, and performs online adjustment on the parameters according to fuzzy control theory to meet the control needs of different deviations and deviation change rates, so as to ensure good dynamic and static characteristics of the controlled variable; compared with conventional PID, the present invention has the advantages of good robustness, high fault tolerance and strong stability; the present invention introduces expert self-tuning control to reduce overshoot faster and accelerate the stabilization speed, and at the same time, adds expert rules on the basis of the fuzzy rule base to realize segmented control and improve system performance; the present invention designs a density soft measurement algorithm based on BP neural network, selects parameter samples in the concentration process, and performs online detection on the density before the discharge of the concentration section; after establishing the soft measurement model of the density of the concentrated liquid, the present invention implements and trains the model through matlab programming, analyzes the simulation results, and optimizes the obtained model results. The momentum factor was increased and the genetic algorithm was introduced to optimize the local minimum value and weight threshold of the neural network. The simulation verification showed that the curves of the expected value and the predicted value of the soft measurement model after the genetic algorithm 3 were basically consistent, and a soft measurement model with good performance was obtained. Combining it with the hardware instrument, the online detection of the density parameters in the concentration stage of traditional Chinese medicine in the project was more accurate. Finally, the PLC program was used to judge whether the density was qualified, and the medicine was concentrated and discharged to complete the control. BRIEF DESCRIPTION OF THE DRAWINGS

[0110] Figure 1 It is the block diagram of the fuzzy self-tuning controller;

[0111] Figure 2 It is a schematic diagram of the membership function of the error e and the error change rate ec;

[0112] Figure 3 It is a schematic diagram of the membership function of Kp;

[0113] Figure 4 It is a schematic diagram of the membership function of Ki;

[0114] Figure 5 It is the membership function diagram of Kd;

[0115] Figure 6 This is the fuzzy self-tuning PID flow chart;

[0116] Figure 7 It is a schematic diagram of fuzzy PID control simulation;

[0117] Figure 8 It is the block diagram of expert fuzzy control system;

[0118] Fig. 9It is a schematic diagram of the BP network soft sensor model of concentrated density;

[0119] Fig.10 This is a schematic diagram of the training comparison results of the BP network model for density prediction;

[0120] Fig.11 It is a schematic diagram of the comparison results of density prediction by the BP network model with momentum factor added;

[0121] Fig.12 This is a schematic diagram of the training comparison results of density prediction based on the BP network model optimized by genetic algorithm;

[0122] Fig.13 It is the error comparison diagram of three network models;

[0123] Fig.14 It is a schematic diagram of density judgment;

[0124] Fig.15 It is a schematic diagram of density detection;

[0125] Fig.16 It is a schematic diagram of a request for medicine;

[0126] Fig. 9 In the above equation, X=(x 1 ,x 2 ...x i ...x 4 ) T is the input vector, B=(b 1 ,b 2 ...b j ...b m ) T is the hidden layer output vector, y is the estimated output, d is the expected output, V = (v 1 ,v 2 ...v j ...v m ) T is the weight matrix between the input layer and the hidden layer, W = (w 11 ,w 21 ... j1 ... m1 ) T is the weight matrix between the hidden layer and the output layer. DETAILED DESCRIPTION

[0127] Specific implementation method 1: Figures 1 to 16 As shown, a method for intelligent control of a Chinese medicine concentration process comprises the following specific steps:

[0128] Step 1: Establish a temperature control model for the concentration process and a soft measurement model for the density of the concentrate;

[0129] The construction of the concentration process temperature control model specifically includes:

[0130] The state equation is constructed as:

[0131]

[0132] In formula (1), is the room temperature elastic coefficient, T is the evaporation chamber temperature, u p is the control action of the steam valve, τ is the time delay of the control function damaging the room temperature, α and β are the structural parameters of the evaporator;

[0133] The pressure equation is constructed as:

[0134]

[0135] In formula (2), P is the pressure inside the evaporator, u is the control function of the vacuum valve, and ε and θ are the structural parameters of the evaporator;

[0136] Let T = x 1 ,P=x 2 , combining equations (1) and (2), the state equation is:

[0137]

[0138] Perform Laplace transform on formula (3) and get G 1 (s) and G 2 (s) are temperature mathematical model and pressure mathematical model respectively:

[0139]

[0140] Only the temperature is studied, and the temperature control model obtained by formula (4) is:

[0141]

[0142] In formula (5), K is the static gain, τ is the lag time, and T is the coefficient of the inertia link;

[0143] The steps to establish the soft-sensing model of concentrated liquid density include:

[0144] The relationship between output density D and variable is:

[0145] D=F(P 1 ,P 2 ,T 1 ,H 1 )(1),

[0146] In formula (1), P 1 Indicates the steam pipeline pressure, P2 Indicates the vacuum degree of the first-effect evaporation chamber, T 1 Indicates the temperature of the first-effect evaporation chamber, H 1 represents the liquid level of the first-effect evaporation chamber; select P in formula (1) 1 , P 2 、T 1 , H 1 The input layer of the neural network constructed with four parameters as input of density soft measurement model is composed of these four auxiliary variables, and the output layer is the dominant variable density D. The input nodes of this model are 4, the output node is 1, and a single hidden layer structure is adopted to construct a three-layer BP neural network model;

[0147] The function that acts as a link in the forward propagation process is f(x), and the output layer is:

[0148]

[0149] The hidden layer output equation is:

[0150]

[0151] The activation function is a unipolar Sigmoid function with an infinite domain and a range of (0,1), and a continuously differentiable function definition:

[0152]

[0153] Import x 0 =-1, b 0 =-1, simplifying formulas (2) and (3) to:

[0154]

[0155] The modified derivation process of weights is as follows:

[0156] The output error is:

[0157]

[0158] Formula (7) is back-propagated to the hidden layer to obtain:

[0159]

[0160] Formula (8) is expanded back to the input layer to obtain:

[0161]

[0162] There is a functional relationship between the error and the weights of each layer. The error size can be changed by adjusting the weights, thereby continuously reducing the error value.

[0163]

[0164] In formulas (10) and (11), the number η is a proportional coefficient with a value range of (0,1). The BP back propagation algorithm is an error signal δ algorithm based on gradient descent;

[0165] Formulas (10) and (11) are expanded to:

[0166]

[0167] Introducing error signals into the output layer and hidden layer and Formulas (10) and (11) are optimized as follows:

[0168] Δw j1 =ηδ 1 y b j (14)

[0169] Δv ij =ηδ j b x i (15)

[0170] The weight is proportional to the input. Expand the output layer and hidden layer separately, and we have:

[0171]

[0172] Combining the error expressions (8) and (9), we have:

[0173]

[0174] Combined with the derivative of the S activation function, the proportional coefficient is obtained:

[0175] δ 1 y =(dy)y(1-y)(20),

[0176] δ j b =[(dy)f'(net 1 ) j1 ]f'(net j )=(δ 1 y w j1 )b j (1-b j )(twenty one),

[0177] Substituting into formula (20) and (21), we get the weight modification equations for the hidden layer and the output layer respectively:

[0178] Δw j1 =ηδ 1 y b j＝ η(dy)y(1-y)b j (twenty two),

[0179] Δv ij =ηδ j b x i＝ η(δ 1 y w j1 )b j (1-b j )x i (twenty three),

[0180] The BP neural network continuously modifies the weights according to the above formula, and after multiple iterations, the error is reduced to obtain the best prediction output; the global sum of squares error E of the output of the concentrated density soft sensor model is q have:

[0181]

[0182] Step 2: Design the fuzzy rules in the temperature control model during the concentration process. According to the formulated fuzzy rules, use the simulation editor in Matlab2018a to build a simulation model of the concentration temperature fuzzy self-tuning controller; introduce the fuzzy self-tuning control algorithm of expert control;

[0183] Through the error changes and rules, the expert system rules are introduced as follows:

[0184] The constant A is greater than the constant B, ka is the amplification factor, ka>1, kb is the suppression factor, 0<kb<1, and em(k) is the deviation extreme value;

[0185] 1. The absolute value of the error e(k) is greater than a constant value A: At this time, no matter how the error changes, the maximum or minimum value of the controller can be directly output to obtain the fastest error adjustment, allowing the error to reduce its absolute value at the fastest rate;

[0186] 2. The absolute value of the error e(k) keeps increasing or is a fixed constant. At this time, e(k)ec(k)>0: If the error value is relatively large, greater than the constant B, a stronger controller is selected:

[0187] u(k)=u(k-1)+k a {[e(k)-e(k-1)]k p}+e(k)k i +[e(k)-2e(k-1)+e(k-2)]k d (6)

[0188] When the absolute value of the error increases but the error itself is very small, less than the smaller constant B, a weaker controller is selected to make the error change in the opposite direction and the absolute value decrease:

[0189] u(k)=u(k-1)+[e(k)-e(k-1)]k p +e(k)k i +[e(k)-2e(k-1)+e(k-2)]k d (7)

[0190] 3. The absolute value of the error e(k) keeps decreasing: the controller output can be kept unchanged to achieve dynamic balance;

[0191] Expressed as a control rule: if e(k)ec(k)<0and e(k)ec(k-1)>0,thenΔK p =ΔK i =ΔK d =0;

[0192] 4. When the error e(k) is at an extreme value: the control output is determined based on the deviation value at time k. When the absolute value of the deviation at this moment is greater than B, the control regulation is stronger:

[0193] u(k)=u(k-1)+k a e m (k)k p (8)

[0194] When the absolute value of the deviation is less than B, the control regulation is weak:

[0195] u(k)=u(k-1)+k b e m (k)k p (9)

[0196] 5. The absolute value of the error e(k) is less than a very small value: saturation needs to be reduced. The solution is to add an integral separation effect, and its mathematical expression is:

[0197]

[0198] In formula (10),

[0199] M+N∈Xe, obviously, with the change of deviation, the value changes continuously in the range of (0, 1), which reduces the steady-state error and has stronger anti-saturation performance;

[0200] Expressed as a control rule: if |e(k)|<ε, then ΔK p =ΔK i =ΔK d =0;

[0201] When the temperature deviation is different, choose different control methods. When the deviation and the deviation change rate meet three and five, choose expert control, and the rest choose fuzzy PID control;

[0202] Step 3, simulate and train the soft measurement model of the concentrated liquid density through matlab programming; analyze the simulation results and continue to optimize the obtained model results; introduce the dynamic factor and the genetic algorithm to optimize the local minimum value and weight range of the neural network respectively;

[0203] The introduction of momentum factor can pass through the minimum value, and a value proportional to the previous weight change is added to each weight change to generate a new weight change. The value modification formula including the momentum factor is as follows:

[0204] ΔW(k+1)=(1-mc)ηδX+mcΔW(k)(26),

[0205] In formula (26), W is the weight matrix of a certain layer, X is the input vector of a certain layer, k is the number of training times, and mc is the introduced momentum factor; the value range of mc is 0 to 1. When mc = 0, the change of weight is adjusted according to the error decrease, and when mc = 1, the next weight change is the same as the previous weight change; the value of the momentum factor is 0.9, and the network model selects mc = 0.95. At this time, the judgment condition of the momentum method is:

[0206]

[0207] In formula (27), SSE(k) is the error, and the error change rate is 1.04;

[0208] Genetic algorithm is introduced to optimize the weights and thresholds of the network. The specific steps include:

[0209] Step (a), chromosome encoding and population initialization: design and build a neural network structure model according to input and output variables, determine the number of input and output nodes and the overall number of layers, the input layer node of the density soft measurement BP network is 4, the output layer node is 1, and the number of layers is three. All weights and thresholds of the BP neural network are encoded, and the length of the code is calculated as:

[0210] S = n × m + m × l + m + l (28),

[0211] In formula (28), n is the input layer node, m is the hidden layer node, n×m is the total weight of the input layer and the hidden layer, and l is the output layer node, then m×l is the total weight of the hidden layer and the output layer, m is the threshold of the hidden layer, l is the threshold of the output layer, n=4, m=10, l=1 determined above, the calculated coding length is 61, and the population size is 50;

[0212] Step (ii), select the fitness function; the fitness function is:

[0213]

[0214] The larger the value of the fitness function, the smaller the deviation value of the neural network model, and the better the adaptability of the neural network;

[0215] Step (iii), performing genetic manipulation; the steps of performing genetic manipulation are:

[0216] Step a: Perform selection operation: The probability of an individual being selected is based on the proportion of its fitness value, and the selection is carried out in a roulette wheel manner. The larger the proportion in the roulette wheel, the better the fitness; if k is the total number of individuals in the population, f i is the value of an individual’s fitness, so the probability of the gene being selected can be expressed as:

[0217]

[0218] Step b: Perform crossover operation: generate new evolved individuals by crossovering certain positions of two individuals, improve the structure of individuals, and the crossover probability is b. c The value of is selected in the interval [0,1];

[0219] Step c, perform mutation operation: The mutation operation selects gene points in the parent sample individuals and replaces them with random numbers to ensure the diversity of population individuals and improve the search ability of genetic algorithms applied to the global network. The mutation probability b m Take the value in [0,1], and choose 0.1;

[0220] Step (iv), loop operation; continuously looping step (ii) and step (iii), repeatedly training the neural network until the result of genetic neural network training meets the error requirement or the number of iterations reaches the target; the individual records the weight and threshold of each iteration, and updates them once after one cycle; genetic operations are used to continuously update and correct the weight and threshold to obtain the optimal individual value, so that the neural network estimation result is more accurate;

[0221] Step 4: Use the PLC program to determine whether the density is qualified and concentrate the medicine.

[0222] Among them, the size of the input and output variables of fuzzy control is described in the form of language as {negative large, negative medium, negative small, zero, positive small, positive medium, positive large}. At the same time, in order to improve the accuracy of the system, when the error e approaches zero, positive zero and negative zero are divided to increase the resolution of the error. Therefore, the fuzzy subset of variable e is set to [NB, NM, NS, NO, PO, PS, PM, PB], and the fuzzy subsets of the remaining four variables ec, ΔKp, ΔKi, ΔKd are: {NB, NM, NS, ZO, PS, PM, PB}.

[0223] Therefore, for this fuzzy controller, the fuzzy rules of PID parameters are shown in Table 3.1, Table 3.2 and Table 3.3 below:

[0224] Table 3.1ΔKp fuzzy control rules table

[0225]

[0226]

[0227] Table 3.2ΔKi fuzzy control rules table

[0228]

[0229] Table 3.2ΔKi fuzzy control rules table

[0230]

[0231] Table 3.3ΔKd fuzzy control rules table

[0232]

[0233] Fuzzy control rules are expressed by "if...then" statements, and the above rule table is described by 56 languages:

[0234] According to the actual engineering experience on site, the temperature deviation range is generally [-30,30], and the temperature difference change rate range is [-6,6]. In order to facilitate modeling and simulation, the fuzzy domain of the fuzzy self-tuning temperature controller is set as:

[0235] e=[-3,-2,-1,0,1,2,3], ec=[-3,-2,-1,0,1,2,3], ΔKp=[-0.3,-0.2,-0.1,0,0.1,0. 2,0.3], ΔKi=[-0.06,-0.04,-0.02,0,0.02,0.04,0.06], ΔKd=[-3,-2,-1,0,1,2,3].

[0236] The basic domain of the fuzzy controller is converted to the fuzzy domain through the quantization factor conversion, and the defuzzification is converted and output through the proportional factor. According to the mapping, the quantization factors of the error and the error change rate are K e =30 / 3=10,K ec =6 / 3=2.

[0237] Because the initial segment response of the system is large, the input and output variables of each fuzzy subset adopt triangular membership functions. To ensure the sensitivity of control, the edge adopts Gaussian function which can improve robustness.

[0238] Among them, the design points of density soft measurement BP neural network include:

[0239] Selection and design of training sets

[0240] Sixty groups were selected for training and testing, of which forty groups were used for training and the remaining twenty groups were used for testing. The selected sample data are shown in Tables 4.1 and 4.2 below:

[0241] Table 4.1 Density soft measurement training set sample data

[0242]

[0243] Table 4.1 Density soft measurement training set sample data

[0244]

[0245]

[0246] Table 4.2 Density soft measurement test set sample data

[0247]

[0248] Table 4.2 Density soft measurement test set sample data

[0249]

[0250] Data preprocessing

[0251] The link function of the network output layer is an S-shaped activation function, whose value is between [0,1], and its value outside this range is too small to be ignored. Therefore, before training, the training data is normalized to the interval [0,1] to make the range of variation of the selected data close, shorten the training time of the neural network, and increase the convergence speed.

[0252] The normalized expression converted to the [0,1] interval is:

[0253]

[0254] In formula (25), x is the sample value before conversion, y is the sample value after conversion, and x min and x max is the minimum and maximum value of the sample initial data;

[0255] Hidden nodes selection

[0256] First, set a small number of hidden nodes, and then increase the number of hidden nodes one by one each time training, and compare the training results of the same sample to find the interval of hidden node values ​​with smaller errors; the empirical formula for hidden node selection can be used: Where n is the number of input layer nodes, l is the number of output layer nodes. According to the test, the range of α is 4 to 10, so the hidden nodes of this model are selected in the range of 6 to 12;

[0257] The network error comparison of the number of hidden nodes within the value range is obtained, and the error comparison of the number of hidden layer nodes is shown in Table 5.3. Obviously, when the number of hidden nodes is 10, the network error is the smallest and the approximation effect is the best. Although the error values ​​of 11 and 12 nodes are also excellent, the number of hidden nodes is large, and the network training time will increase accordingly. Considering all factors, it is the most reasonable to choose 10 nodes.

[0258] Table 4.3 Error comparison of different numbers of hidden nodes

[0259]

[0260] Comparison and selection of training methods

[0261] According to the selected sample set, under the same other conditions, the number of hidden nodes is selected as 10, and the six algorithms are trained. The comparison table of training effects is shown in Table 4.4:

[0262] Table 4.4 Comparison of the training effects of different training methods on BP network

[0263]

[0264] From the table analysis, we know that the number of iterations of the quasi-Newton algorithm and the Levenberg-Marquardt algorithm is small, but the error of the quasi-Newton algorithm is large, while the errors of the Levenberg-Marquardt algorithm and the Bayesian algorithm are similar, but the number of iterations of the Bayesian algorithm is too many and the convergence speed is slow; the Levenberg-Marquardt algorithm is selected to train the model. The algorithm has a short training cycle, fast training time, and small training error, which speeds up the training speed and improves the generalization ability of the soft sensor model;

[0265] Learning rate determination

[0266] The initial learning rate of the selected density prediction neural network model is 0.01.

[0267] Among them, the soft measurement model of BP neural network in the concentration section was constructed and programmed using matlab2018a software, and simulation analysis was performed in combination with its internal neural network toolbox;

[0268] The writing and implementation of the program is divided into the following steps:

[0269] According to the above design, the soft-sensing model of the BP neural network in the concentration section is programmed using matlab2018a, and combined with its internal neural network toolbox for running simulation. The programming and implementation of the program are divided into the following steps:

[0270] (1) Data processing:

[0271] Loading input data

[0272] Loading output data

[0273] 40 training set inputs

[0274] 40 training set outputs

[0275] 20 test set inputs

[0276] 20 test set outputs

[0277] Sample data is normalized to (0,1):

[0278] (2) BP neural network design, training and testing:

[0279] Create a network with the initial hidden nodes set to 10

[0280] Select the Levenberg-Marquardt training method

[0281] Set the training parameters, this is the number of training times, set it to a maximum of 10,000 times and stop

[0282] Set the minimum error required for the target, here it is set to 0.001

[0283] Set the learning rate, the standard initial network is set to 0.01

[0284] The network weights and thresholds are stored in net.IW, net.LW, and net.b.

[0285] Train the network

[0286] Perform simulation and call sim function to implement network testing

[0287] Data normalization

[0288] (3) Performance evaluation of neural network models

[0289] error

[0290] Relative error

[0291] The absolute coefficient of the regression curve is used to evaluate the goodness of fit of the prediction results of the neural network model.

[0292] The output results are expected data, predicted data, error and relative error.

[0293] When constructing a genetically optimized neural network, the GAOT genetic algorithm toolbox was selected. This toolbox was developed and launched by Carolina State University in the United States. When using it, it should be configured in Matlab in advance.

[0294] The GA-BP network is given a genetic algebra of 100. According to the set parameters, the GA operation is performed to improve the network structure. The constructed neural network density soft measurement model is retrained using the optimized weights and thresholds. The simulation results are as follows: Fig.12 As shown;

[0295] Depend on Fig.12 It can be seen that the actual value of the BP neural network density soft measurement model optimized by genetic algorithm is basically consistent with the output value of the model, and the density value detected during the concentration process can be predicted more perfectly.

[0296] The prediction values, absolute errors, and relative errors of the initial BP neural network soft measurement model, the neural network soft measurement model with momentum factor, and the neural network soft measurement model with genetic algorithm are compared to obtain the following Table 4.5. After calculation, it is found that the average absolute error of the initial BP neural network is 0.41%, the average absolute error of the improved concentrated liquid density soft measurement model with momentum factor added is 0.24%, and the absolute error of the predicted data of the finally optimized BP neural network concentrated liquid density soft measurement model based on genetic algorithm is 0.13%. By comparison, it can be seen that the soft measurement model based on the BP neural network with momentum factor genetic algorithm is most consistent with the actual value and has the highest accuracy.

[0297] Table 4.5 Data comparison of three neural network soft sensor models

[0298]

[0299]

[0300] Compare the absolute errors of the predicted values ​​of the three models with the actual values ​​and draw a curve, such as Fig.13As shown in the figure, curve a is the error of the initial standard BP network soft measurement model, curve b is the error curve of the soft measurement model with a momentum factor of 0.95, and curve c is the error curve of the soft measurement model with a genetic algorithm added to optimize the weight threshold on the basis of adding the momentum factor. By comparison, it can be seen that the oscillation amplitude of curve c is the smallest, and the model error after the two optimizations is the lowest.

[0301] The essence of density measurement is to detect the density of the concentrated liquid. After judgment, the concentrated medicine that meets the conditions can be discharged and the concentration is completed. In actual engineering, step7 can be used to program density detection processing. The program segment of this process is as follows Figures 14 to 16 As shown:

[0302] Among them, NS03 is the three stages of concentration, Kg_Pause is the cut-off switch for concentrator pause, MDJC is density detection, An_MdJc is manual intervention density detection, An_NsEndSp_Ait is the end of manual intervention concentration, Wincc sets the qualified density value, PvMdOkT is the current density qualified time, Ait_No is unqualified density after detection, Ait_OK is qualified density after detection, and CyQQ is the drug delivery request.

[0303] The above is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technician familiar with this profession can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any simple modification, equivalent replacement and improvement made to the above embodiments without departing from the content of the technical solution of the present invention, based on the technical essence of the present invention, within the spirit and principles of the present invention, still fall within the protection scope of the technical solution of the present invention.

Claims

1. An intelligent control method for a traditional Chinese medicine concentration process, characterized in that: The specific steps include: Step 1: Establish a temperature control model for the concentration process and a soft measurement model for the density of the concentrate; Step 2: Design the fuzzy rules in the temperature control model during the concentration process. According to the formulated fuzzy rules, use the simulation editor in Matlab2018a to build a simulation model of the concentration temperature fuzzy self-tuning controller; introduce the fuzzy self-tuning control algorithm of expert control; Step 3, simulate and train the soft measurement model of the concentrated liquid density through matlab programming; analyze the simulation results and continue to optimize the obtained model results; introduce the dynamic factor and the genetic algorithm to optimize the local minimum value and weight range of the neural network respectively; Step 4: Use the PLC program to determine whether the density is qualified and concentrate the medicine.

2. The intelligent control method for traditional Chinese medicine concentration process according to claim 1, characterized in that: The temperature control model of the concentration process constructed in step 1 specifically includes: The state equation is constructed as: In formula (1), is the room temperature elastic coefficient, T is the evaporation chamber temperature, u p is the control action of the steam valve, τ is the time delay of the control function damaging the room temperature, α and β are the structural parameters of the evaporator; The pressure equation is constructed as: In formula (2), P is the pressure inside the evaporator, u is the control function of the vacuum valve, and ε and θ are the structural parameters of the evaporator; Let T = x1, P = x2, and combine equations (1) and (2), we get the state equation: By performing Laplace transform on formula (3), we can obtain G1(s) and G2(s), which are the temperature mathematical model and pressure mathematical model respectively: Only the temperature is studied, and the temperature control model obtained by formula (4) is: In formula (5), K is the static gain, τ is the lag time, and T is the coefficient of the inertia link.

3. The intelligent control method for traditional Chinese medicine concentration process according to claim 1, characterized in that: The steps of establishing the soft-sensing model of the density of the concentrated liquid in step 1 include: The relationship between output density D and variable is: D=F(P1,P2,T1,H1)(1), In formula (1), P1 represents the steam pipeline pressure, P2 represents the vacuum degree of the first-effect evaporation chamber, T1 represents the temperature of the first-effect evaporation chamber, and H1 represents the liquid level of the first-effect evaporation chamber; the input layer of the neural network constructed by selecting the four parameters P1, P2, T1, and H1 in formula (1) as the input of the density soft measurement model consists of these four auxiliary variables, and the output layer is the dominant variable density D. The input nodes of the model are 4, the output node is 1, and a single hidden layer structure is used to construct a three-layer BP neural network model; The function that plays a link role in the forward propagation process is f(x), and the output layer is: The hidden layer output equation is: The activation function is a unipolar Sigmoid function with an infinite domain and a range of (0,1), and a continuously differentiable function definition: Introducing x0=-1, b0=-1, formulas (2) and (3) are simplified to: The modified derivation process of weights is as follows: The output error is: Formula (7) is back-propagated to the hidden layer to obtain: Formula (8) is expanded back to the input layer to obtain: There is a functional relationship between the error and the weights of each layer. The error size can be changed by adjusting the weights, thereby continuously reducing the error value. In formulas (10) and (11), the number η is a proportional coefficient with a value range of (0,1). The BP back propagation algorithm is an error signal δ algorithm based on gradient descent; Formulas (10) and (11) are expanded to: Introducing error signals into the output layer and hidden layer and Formulas (10) and (11) are optimized as follows: Δw j1 =ηδ1 y b j (14), Δv ij =hd j b x i (15), The weight is proportional to the input. Expand the output layer and hidden layer separately, and we have: Combining the error expressions (8) and (9), we have: Combined with the derivative of the S activation function, the proportional coefficient is obtained: δ1 y =(d-y)y(1-y)(20), δ j b =[(dy)f'(net1)w j1 ]f'(net j )=(δ1 y In j1 )b j (1-b j )(21), Substituting into formula (20) and (21), we get the weight modification equations for the hidden layer and the output layer respectively: Δw j1 =ηδ1 y b j＝ η(d-y)y(1-y)b j (22), Δv ij =hd j b x i＝ n(δ1 y w j1 )b j (1-b j )x i (23), The BP neural network continuously modifies the weights according to the above formula, and after multiple iterations, the error is reduced to obtain the best prediction output; the global sum of squares error E of the output of the concentrated density soft sensor model is q have:

4. The intelligent control method for Chinese medicine concentration process according to claim 1, characterized in that: In step 2, based on the tracking of the deviation and the rate of change of the deviation during the control process, the self-tuning rules of the fuzzy PID parameters are summarized as follows: In the response stage, the error e is relatively large, so the value of Kp should be increased and the value of Kd should be decreased to shorten the response time of the control system, and Ki=0 should be taken to suppress overshoot; In the following stage, the error e is moderate, and the values ​​of Ki and Kd are moderate to ensure a faster adjustment speed. At the same time, a smaller Kp value is used to reduce overshoot. In the adjustment stage, the error e is small, and the values ​​of Kp and Ki should be appropriately increased to reduce system oscillation and improve controller stability. At the same time, when ec is small, the Kd value is slightly larger, and when ec is large, the Kd value is reduced. In the fuzzy PID temperature control system, the fuzzy rules meet the following requirements: when the temperature is too high, the temperature can be reduced significantly to make it approach the expected value, and the control amount should be increased at this time; On the contrary, when the temperature value is close to the expected value, a smaller control amount is applied to slow down the control rate and make the control system reach a stable state; The designed fuzzy controller has two input values, namely error e and error change rate ec, while Kp, Ki and Kd are its three outputs. Through the quantization of the fuzzy controller, the maximum membership method is used to obtain the accurate output value, so as to adjust the PID coefficient, and after accumulation and output: Kp = Kp' + ΔKp, Ki = Ki' + ΔKi, Kd = Kd' + ΔKd, act on the controlled object, and finally get the output graph; Through the error changes and rules, the expert system rules are introduced as follows: The constant A is greater than the constant B, ka is the amplification factor, ka>1, kb is the suppression factor, 0<kb<1, and em(k) is the deviation extreme value; 1. The absolute value of the error e(k) is greater than a constant value A: At this time, no matter how the error changes, the maximum or minimum value of the controller can be directly output to obtain the fastest error adjustment, allowing the error to reduce its absolute value at the fastest rate; 2. The absolute value of the error e(k) keeps increasing or is a fixed constant. At this time, e(k)ec(k)>0: If the error value is relatively large, greater than the constant B, a stronger controller is selected: u(k)=u(k-1)+k a {[e(k)-e(k-1)]k p }+e(k)k i +[e(k)-2e(k-1)+e(k-2)]k d (6), When the absolute value of the error increases but the error itself is very small, less than the smaller constant B, a weaker controller is selected to make the error change in the opposite direction and the absolute value decrease: u(k)=u(k-1)+[e(k)-e(k-1)]k p +e(k)k i +[e(k)-2e(k-1)+e(k-2)]k d (7), 3. The absolute value of the error e(k) keeps decreasing: the controller output can be kept unchanged to achieve dynamic balance; Expressed as a control rule: if e(k)ec(k)<0and e(k)ec(k-1)>0,thenΔK p =ΔK i =ΔK d =0; 4. When the error e(k) is at an extreme value: the control output is determined based on the deviation value at time k. When the absolute value of the deviation at this moment is greater than B, the control regulation is stronger: u(k)=u(k-1)+k a and m (k)·k p (8), When the absolute value of the deviation is less than B, the control regulation is weak: u(k)=u(k-1)+k b and m (k)·k p (9), 5. The absolute value of the error e(k) is less than a very small value: saturation needs to be reduced. The solution is to add an integral separation effect, and its mathematical expression is: Formula (10) M+N∈Xe, obviously, with the change of deviation, the value changes continuously in the range of (0, 1), which reduces the steady-state error and has stronger anti-saturation performance; Expressed as a control rule: if |e(k)|<ε, then ΔK p =ΔK i =ΔK d =0; When the temperature deviation is different, select different control methods. When the deviation and the deviation change rate meet three and five, select expert control, and select fuzzy PID control for the rest.

5. The intelligent control method for Chinese medicine concentration process according to claim 1, characterized in that: In step 3, the soft-sensing model of the BP neural network in the concentration section is programmed using MATLAB and combined with its internal neural network toolbox for operation simulation. The specific steps include: Step 1: Data processing; Step 2: BP neural network design, training and testing; Step 3: Performance evaluation of the neural network model; Step 4: Run the Matlab program to obtain the simulation result curve.

6. The intelligent control method for Chinese medicine concentration process according to claim 1, characterized in that: The introduction of momentum factor in step 3 can pass through the minimum value, and add a value proportional to the previous weight change to each weight change to generate a new weight change. The value modification formula including the momentum factor is as follows: ΔW(k+1)=(1-mc)ηδX+mcΔW(k)(26), In formula (26), W is the weight matrix of a certain layer, X is the input vector of a certain layer, k is the number of training times, and mc is the introduced momentum factor; the value range of mc is 0 to 1. When mc = 0, the change of weight is adjusted according to the error decrease, and when mc = 1, the next weight change is the same as the previous weight change; the value of the momentum factor is 0.9, and the network model selects mc = 0.

95. At this time, the judgment condition of the momentum method is: In formula (27), SSE(k) is the error, and the error change rate is 1.

04.

7. The intelligent control method for Chinese medicine concentration process according to claim 1, characterized in that: In step 3, a genetic algorithm is introduced to optimize the weights and thresholds of the network. The specific steps include: Step (a), chromosome encoding and population initialization: design and build a neural network structure model according to input and output variables, determine the number of input and output nodes and the overall number of layers, the input layer node of the density soft measurement BP network is 4, the output layer node is 1, and the number of layers is three. All weights and thresholds of the BP neural network are encoded, and the length of the code is calculated as: S = n × m + m × l + m + l (28), In formula (28), n is the input layer node, m is the hidden layer node, n×m is the total weight of the input layer and the hidden layer, and l is the output layer node, then m×l is the total weight of the hidden layer and the output layer, m is the threshold of the hidden layer, l is the threshold of the output layer, n=4, m=10, l=1 determined above, the calculated coding length is 61, and the population size is 50; Step (ii), select the fitness function; the fitness function is: The larger the value of the fitness function, the smaller the deviation value of the neural network model, and the better the adaptability of the neural network; Step (iii), performing genetic manipulation; the steps of performing genetic manipulation are: Step a: Perform selection operation: The probability of an individual being selected is based on the proportion of its fitness value, and the selection is carried out in a roulette wheel manner. The larger the proportion in the roulette wheel, the better the fitness; if k is the total number of individuals in the population, f i is the value of an individual’s fitness, so the probability of the gene being selected can be expressed as: Step b: Perform crossover operation: generate new evolved individuals by crossovering certain positions of two individuals, improve the structure of individuals, and the crossover probability is b. c The value of is selected in the interval [0,1]; Step c, perform mutation operation: The mutation operation selects gene points in the parent sample individuals and replaces them with random numbers to ensure the diversity of population individuals and improve the search ability of genetic algorithms applied to the global network. The mutation probability b m Take the value in [0,1], and choose 0.1; Step (iv), loop operation; continuously loop step (ii) and step (iii), repeatedly train the neural network until the result of genetic neural network training meets the error requirement or the number of iterations reaches the target; the individual records the weight and threshold of each iteration, and updates them once after one cycle; genetic operation is used to continuously update and correct the weight and threshold to obtain the optimal individual value, so that the neural network estimation result is more accurate.

Images