CN102258960A - Method for automatically generating formula of fluidized bed based on neural network system - Google Patents
Method for automatically generating formula of fluidized bed based on neural network system Download PDFInfo
- Publication number
- CN102258960A CN102258960A CN2011101386320A CN201110138632A CN102258960A CN 102258960 A CN102258960 A CN 102258960A CN 2011101386320 A CN2011101386320 A CN 2011101386320A CN 201110138632 A CN201110138632 A CN 201110138632A CN 102258960 A CN102258960 A CN 102258960A
- Authority
- CN
- China
- Prior art keywords
- network system
- nerve network
- fluid bed
- raw material
- parameter
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a method for automatically generating the formula of a fluidized bed based on a neural network system. The method comprises the following steps of: 1, acquiring a training sample of a neural network by the neural network system, and storing in a data storage module; 2, simulating and training by the neural network system; and 3, applying by the neural network system and automatically generating the formula. A production formula is set by a neural network technology, parameters of the production formula for controlling the running of the fluidized bed are automatically generated by using raw material parameters, and the formula is not needed to be set by experienced formula engineers according to experiences, so that performance of a product and production efficiency and the scientificalness and accuracy of the formula setting are improved.
Description
Technical field
The present invention relates to a kind of automatic production technology, be specifically related to a kind of fluid bed prescription and generate method automatically based on nerve network system.
Background technology
At present, the prescription of granulating process is to the performance important influence of pharmaceutical products.Usually prescription relies on test and workman's experience to obtain basically, and this just makes the design of pharmacy procedure bear the character of much blindness.Therefore, obtain formulation parameter automatically, not only can improve the performance of product, and enhance productivity with effective model.Because have complicated non-linear relation between the parameters of granulating process prescription and pharmacy procedure, most cases can only lean on long-term test to grope, and is felt to obtain by the workman, lacks science and accuracy, the experience grasp cycle is long, also occurs systematic error easily.
Neutral net (Neural Network, NN) be to be the dynamical system of topological structure with the digraph by artificial foundation, state is corresponding to carry out information processing by continuous or interrupted input is done for it, have fault-tolerant, association, supposition, memory, self adaptation and characteristics such as processing certainly, can finish the Nonlinear Mapping of complicated input and output preferably, parallel processing capability is strong.Structurally, can be divided into input layer, output layer and hidden layer to a neutral net, as shown in Figure 2.Each node correspondence input variable one by one of input layer, the corresponding output variable of the node of output layer can have a plurality of.Be hidden layer (invisible concerning the user of neutral net) between input layer and output layer, the number of plies of hidden layer and the number of every node layer have determined the complexity of neutral net.The number of the extremely contained node of hidden layer by training decision neural network structure, and the connected mode between the node are set up the non-linear relation between input variable and the output variable.
Therefore neutral net is specially adapted to study the non-linear relation model between granulation parameter and the prescription.Literature search through prior art finds still do not have the technical scheme of fluid bed prescription generation method aspect at present.
Summary of the invention
The invention provides a kind of fluid bed prescription and generate method automatically, improve the performance and the production efficiency of product, improve the science and the accuracy of granulating process prescription or pharmacy procedure parameter based on nerve network system.
For achieving the above object, the present invention discloses a kind of fluid bed prescription generation method based on nerve network system, is characterized in that the method includes the steps of:
The collection of step 1 nerve network system is used for the training sample of the neutral net of fluid bed;
The fluid bed dry run of step 1.1 reality;
The raw material parameter of step 1.2 default fluid bed operation;
Step 1.3 system provides and the corresponding formulation parameter of raw material parameter according to the running of fluid bed;
Step 1.4 nerve network system storage corresponding raw material parameter of fluid bed and formulation parameter, acquisition is used for the training sample of neural network training;
Step 2.1 nerve network system imports each the raw material parameter in the training sample, as the input layer of neutral net;
Step 2.2 nerve network system imports each formulation parameter in the training sample, as the output layer neuron of neutral net;
Step 2.3 nerve network system calculates the number of hidden neuron, hidden neuron number N=number of training/2+e, and e gets the integer value between 0 to 9 at random;
Step 2.4 nerve network system utilizes training sample to carry out simulation training, determines the weights and the threshold value of neutral net;
The training sample raw material parameter of step 2.5 nerve network system forward-propagating input layer input, and judge whether the output of output layer is consistent with the formulation parameter of the training sample that imports output layer, if, then jump to step 2.8, if not, then jump to step 2.6;
Step 2.6 nerve network system reverse propagated error, and learn, the weight and the threshold value of modification or each node of each layer of iteration reduce cost function;
Step 2.7 nerve network system judges whether cost function can reduce again, if, then finished the mapping of input layer and output layer, then jump to step 2.6, if not, and jump to step 2.8;
Step 2.8 nerve network system simulation training is finished, and sets up the relation between input layer and the output layer neuron;
Step 3 nerve network system is used, and generates the prescription of fluid bed automatically;
Step 3.1 nerve network system input raw material parameter is to the neutral net input layer;
Step 3.2 nerve network system is handled the raw material parameter, and output layer generates formulation parameter;
The formulation parameter that step 3.3 nerve network system output nerve network generates, the running of control fluid bed.
Raw material parameter in the above-mentioned step 1.3 comprises raw material type, material density, raw material granularity, slurry viscosity, slurries ratio, slurry temperature and hydrophily.
Formulation parameter in the above-mentioned step 1.3 comprises the low rotating speed, height of stirring and stirs rotating speed, a low grain rotating speed, high grain rotating speed, total time, height and stir that beginning, low grain beginning, high grain beginning, whitewashing duration, whitewashing begin, the terminal point electric current.
A kind of fluid bed prescription automatic creation system and generation method thereof based on neutral net of the present invention compared with prior art, its advantage is, the present invention sets factory formula by nerual network technique, automatically generate the factory formula parameter of control fluid bed running with the raw material parameter, do not need experienced prescription engineer to set prescription with experience, enhance product performance and production efficiency, improve science and accuracy that prescription is set.
Description of drawings
Fig. 1 is the method flow diagram of the generation method of a kind of fluid bed prescription automatic creation system based on nerve network system of the present invention;
Fig. 2 is the structural representation of neutral net;
Fig. 3 is the method flow diagram of the simulation training method of the generation method of a kind of fluid bed prescription automatic creation system based on nerve network system of the present invention.
The specific embodiment
Below in conjunction with description of drawings the specific embodiment of the present invention.
The present invention has illustrated that a kind of fluid bed prescription based on nerve network system generates method automatically.Can generate the prescription of control fluid bed running automatically according to the raw material parameter.
Below in conjunction with Fig. 1 and Fig. 3 a kind of fluid bed prescription generation method based on nerve network system of the present invention is described, the method includes the steps of.
The collection of step 1 nerve network system is used for the training sample of the BP neutral net of fluid bed.
The fluid bed dry run of step 1.1 reality.
Step 1.2 is set the raw material parameter of fluid bed operation.
Step 1.3 system provides the corresponding formulation parameter of raw material parameter with above-mentioned setting according to fluid bed and fluid bed control system running thereof.
Corresponding raw material parameter and formulation parameter that step 1.4 system will be used to control the fluid bed running are mapped one by one, and store the training sample that conduct is used for neural network training.
This training sample comprises the raw material parameter that fluid bed will be handled raw material, and controls the formulation parameter that fluid bed operates with the corresponding fluid bed control system of raw material parameter.Wherein, the raw material parameter comprises raw material type, material density, raw material granularity, slurry viscosity, slurries ratio, slurry temperature and hydrophily.Formulation parameter is meant that the low rotating speed, height of stirring stirs rotating speed, a low grain rotating speed, high grain rotating speed, total time, height and stir that beginning, low grain beginning, high grain beginning, whitewashing duration, whitewashing begin, the terminal point electric current.
Before carrying out simulation training, set earlier a threshold value, if cost function greater than this threshold value, then neutral net continues training, smaller or equal to this threshold value, illustrates that cost function need not reduce again.This threshold value is little, and the formulation parameter of explanation training sample output is very approaching with the formulation parameter that training obtains.
Step 2.1 nerve network system imports each the raw material parameter in the training sample, and each raw material parameter is as the neuron of an input layer of neutral net, and its neuron number is the number of each the raw material parameter in the training sample.
Step 2.2 nerve network system imports each formulation parameter in the training sample, and each formulation parameter is as the neuron of an output layer of neutral net, and its neuron number is the number of each formulation parameter in the training sample.
Step 2.3 nerve network system calculates the number of neutral net hidden neuron, and the hidden neuron number N=number of training/2+e(e of neutral net gets the integer value between 0 to 9, and the value of this e is produced at random by nerve network system).
Step 2.4 nerve network system utilizes above-mentioned training sample simulation training based on the BP neutral net that the fluid bed prescription generates, and determines the weights and the threshold value of BP neutral net.Wherein initial learn speed is 0.02, the momentum system is 0.98, the error of network is 0.01, adopt the batch processing gradient descent method of drive amount to train the BP neutral net, its simulation training method is at present most widely used general and theoretical ripe a kind of method, it uses the gradient search technology that cost function is minimized, to finish the mapping from the input layer to the output layer.
The gradient search technology is a kind of iteration optimizing algorithm to certain criterion function.If
Be criterion function,
It is a vector.
Be
The point
Gradient, be a vector, its direction is
Fastest-rising direction; The negative gradient direction then is
Reduce the fastest direction.Therefore,, walk, can full out reach maximum point along gradient direction if ask the maximum of certain function; Otherwise, walk along the negative gradient direction, can full out reach smallest point, the gradient search technology is to ask the minimizing iterative algorithm of function.
Cost function adopts mean square error, the i.e. square root of the mean value of the quadratic sum of error.In the present invention with the raw material parameter (raw material type in the training sample, material density, raw material granularity, slurry viscosity, the slurries ratio, slurry temperature and hydrophily) input layer of input neural network, the output layer of neutral net will be exported corresponding formulation parameter (the low rotating speed that stirs, height stirs rotating speed, low grain rotating speed, high grain rotating speed, total time, height stirs beginning, low grain beginning, high grain beginning, the whitewashing duration, the whitewashing beginning, the terminal point electric current), the square root of the mean value of the quadratic sum of the difference of the corresponding formulation parameter in this formulation parameter and the training sample is exactly the cost function of fluid bed.
Step 2.5 nerve network system with the raw material parameter of the training sample of input layer input in neutral net forward-propagating under existing weight and threshold value.Initial weight and threshold value are nonzero value at random, when the error between the output of this value in next can constantly adjust to output and training sample according to algorithm is little till.Forward-propagating is meant that input signal passes to output layer through hidden layer from input layer.Nerve network system also judges whether handle the output parameter that obtains output layer through hidden layer consistent with the formulation parameter of the training sample that imports output layer, if, then jump to step 2.8, if not, do not obtain the output expected, then jump to step 2.6.
Step 2.6 nerve network system reverse propagated error, backpropagation is exactly by former connecting path backwards calculation with error signal (formulation parameter of the formulation parameter of training sample output and neutral net output), learn, weights and threshold value by the modification of gradient descent method or each layer of iteration neuron and each node, cost function is reduced, and this cost function promptly is above-mentioned mean square error.
Step 2.7 nerve network system is compared cost function with preset threshold before the simulation training, judge whether cost function can reduce again, if then cost function then continues training through network greater than prior preset threshold, and jump to step 2.6, if not, then cost function is less than or equal to this threshold value of prior setting, illustrates that then cost function need not reduce again, promptly finish the mapping of input layer and output layer, and jumped to step 2.8.
Step 2.8 is finished based on the BP neutral net simulation training that the fluid bed prescription generates, and sets up the relation between input layer and the output layer neuron.
After one kind of step 3 is used for nerve network system that the fluid bed prescription generates and trains, be applied to the obtaining of formulation parameter of fluid bed in the practical work, realize the automatic generation of Recipe in the production process.
Step 3.1 nerve network system input raw material parameter is to neutral net.Nerve network system is with the neuron of each raw material parameter as the neutral net input layer.The raw material parameter comprises raw material type, material density, raw material granularity, slurry viscosity, slurries ratio, slurry temperature and hydrophily, a kind of parameter of the corresponding raw material of the variable of each input layer of neutral net.According to the raw material parameter, set the value of neutral net input layer variable.
Step 3.2 nerve network system is handled the raw material parameter of input layer input, adopts forward-propagating to calculate the output of neutral net output layer, generates formulation parameter by output layer.
Output layer neuron that neutral net is set up during according to above-mentioned simulation training and the relation between the output layer neuron automatically according to the raw material parameter of input layer, change the value of output parameter, by the neuron output formulation parameter of output layer.These output variables and required prescription are corresponding one by one, by the value of output layer variable, just can obtain the low rotating speed, height of stirring and stir rotating speed, a low grain rotating speed, high grain rotating speed, total time, height and stir that beginning, low grain beginning, high grain beginning, whitewashing duration, whitewashing begin, the value of terminal point electric current.
If input layer is a vector
, represent 7 kinds of raw material parameters respectively,
Then output layer is a vector
, represent the value of EAT in the formulation parameter, leaving air temp, temperature of charge, total time, blower fan frequency, valve area respectively.
Wherein i component of output layer is:
The formulation parameter that step 3.3 nerve network system output nerve network generates automatically is directed into the fluid bed control system, the running of control fluid bed.
Although content of the present invention has been done detailed introduction by above preferred embodiment, will be appreciated that above-mentioned description should not be considered to limitation of the present invention.After those skilled in the art have read foregoing, for multiple modification of the present invention with to substitute all will be conspicuous.Therefore, protection scope of the present invention should be limited to the appended claims.
Claims (6)
1. fluid bed prescription generation method based on nerve network system is characterized in that the method includes the steps of:
The collection of step 1 nerve network system is used for the training sample of the neutral net of fluid bed;
Step 2 nerve network system simulation training is used for the neutral net of fluid bed;
Step 3 Application of Neural Network generates the fluid bed prescription automatically.
2. the fluid bed prescription generation method based on nerve network system as claimed in claim 1 is characterized in that described step 1 comprises following steps:
The fluid bed dry run of step 1.1 reality;
The raw material parameter of step 1.2 default fluid bed operation;
Step 1.3 system provides and the corresponding formulation parameter of raw material parameter according to the running of fluid bed;
Step 1.4 nerve network system storage corresponding raw material parameter of fluid bed and formulation parameter, acquisition is used for the training sample of neural network training.
3. the fluid bed prescription generation method based on nerve network system as claimed in claim 2, it is characterized in that the raw material parameter described in the described step 1.3 comprises raw material type, material density, raw material granularity, slurry viscosity, slurries ratio, slurry temperature and hydrophily.
4. the fluid bed prescription generation method based on nerve network system as claimed in claim 2, it is characterized in that the formulation parameter described in the described step 1.3 comprises the low rotating speed, height of stirring and stirs rotating speed, a low grain rotating speed, high grain rotating speed, total time, height and stir that beginning, low grain beginning, high grain beginning, whitewashing duration, whitewashing begin, the terminal point electric current.
5. the fluid bed prescription generation method based on nerve network system as claimed in claim 1 is characterized in that described step 2 comprises following steps:
Step 2.1 nerve network system imports each the raw material parameter in the training sample, as the input layer of neutral net;
Step 2.2 nerve network system imports each formulation parameter in the training sample, as the output layer neuron of neutral net;
Step 2.3 nerve network system calculates the number of hidden neuron, hidden neuron number N=number of training/2+e, and e gets the integer value between 0 to 9 at random;
Step 2.4 nerve network system utilizes training sample to carry out simulation training, determines the weights and the threshold value of neutral net;
The training sample raw material parameter of step 2.5 nerve network system forward-propagating input layer input, and judge whether the output of output layer is consistent with the formulation parameter of the training sample that imports output layer, if, then jump to step 2.8, if not, then jump to step 2.6;
Step 2.6 nerve network system reverse propagated error, and learn, the weight and the threshold value of modification or each node of each layer of iteration reduce cost function;
Step 2.7 nerve network system judges whether cost function can reduce again, if, then finished the mapping of input layer and output layer, then jump to step 2.6, if not, and jump to step 2.8;
Step 2.8 nerve network system simulation training is finished, and sets up the relation between input layer and the output layer neuron.
6. the fluid bed prescription generation method based on nerve network system as claimed in claim 1 is characterized in that described step 3 method comprises following steps:
Step 3.1 nerve network system input raw material parameter is to the input layer of neutral net;
Step 3.2 nerve network system is handled the raw material parameter, and output layer generates formulation parameter;
The formulation parameter that step 3.3 nerve network system output nerve network generates, the running of control fluid bed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2011101386320A CN102258960A (en) | 2011-05-26 | 2011-05-26 | Method for automatically generating formula of fluidized bed based on neural network system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2011101386320A CN102258960A (en) | 2011-05-26 | 2011-05-26 | Method for automatically generating formula of fluidized bed based on neural network system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN102258960A true CN102258960A (en) | 2011-11-30 |
Family
ID=45005908
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2011101386320A Pending CN102258960A (en) | 2011-05-26 | 2011-05-26 | Method for automatically generating formula of fluidized bed based on neural network system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102258960A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104562194A (en) * | 2013-10-24 | 2015-04-29 | 上海西门子工业自动化有限公司 | Technical process control method |
| CN106682451A (en) * | 2016-12-08 | 2017-05-17 | 深圳先进技术研究院 | Formula proportion determining method for biological tissue simulation material and system |
| CN107544252A (en) * | 2017-09-19 | 2018-01-05 | 中国计量大学 | Vertical material blanking machine controller based on machine learning |
| CN109669349A (en) * | 2019-01-28 | 2019-04-23 | 上海锐驰通科技有限公司 | A kind of EMS electronic manufacture monitoring system |
-
2011
- 2011-05-26 CN CN2011101386320A patent/CN102258960A/en active Pending
Non-Patent Citations (3)
| Title |
|---|
| 叶海文: "燃煤循环流化床锅炉模型化的人工神经网络方法", 《清华大学学报.自然科学版》 * |
| 易文林 等: "BP神经网络在吴淞口潮位短期预报中的应用", 《东北水利水电》 * |
| 郑建荣 等: "流化床喷雾制粒神经网络模型及其应用", 《华东理工大学学报》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104562194A (en) * | 2013-10-24 | 2015-04-29 | 上海西门子工业自动化有限公司 | Technical process control method |
| CN104562194B (en) * | 2013-10-24 | 2017-05-31 | 西门子工厂自动化工程有限公司 | The temprature control method of polysilicon production process |
| CN106682451A (en) * | 2016-12-08 | 2017-05-17 | 深圳先进技术研究院 | Formula proportion determining method for biological tissue simulation material and system |
| CN106682451B (en) * | 2016-12-08 | 2019-05-14 | 深圳先进技术研究院 | A kind of formula rate of biological tissue's simulation material determines method and system |
| CN107544252A (en) * | 2017-09-19 | 2018-01-05 | 中国计量大学 | Vertical material blanking machine controller based on machine learning |
| CN107544252B (en) * | 2017-09-19 | 2020-08-11 | 中国计量大学 | Machine learning-based direct-falling material blanking machine controller |
| CN109669349A (en) * | 2019-01-28 | 2019-04-23 | 上海锐驰通科技有限公司 | A kind of EMS electronic manufacture monitoring system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102184454B (en) | Granulator formula generation method based on neural network system | |
| Ou et al. | Comparison between PSO and GA for parameters optimization of PID controller | |
| CN104834215B (en) | A kind of BP neural network pid control algorithm of mutation particle swarm optimization | |
| CN110568359B (en) | Lithium battery residual life prediction method | |
| CN104408518B (en) | Based on the neural network learning optimization method of particle swarm optimization algorithm | |
| CN105138717A (en) | Transformer state evaluation method by optimizing neural network with dynamic mutation particle swarm | |
| CN107239859A (en) | The heating load forecasting method of Recognition with Recurrent Neural Network is remembered based on series connection shot and long term | |
| CN112700060A (en) | Station terminal load prediction method and prediction device | |
| CN109800461A (en) | Crucial gabarit parameter optimization method and device for tire construction light-weight design | |
| CN117970783B (en) | Numerical control high-speed milling and drilling machine control method based on improved river horse algorithm | |
| CN108594927B (en) | A MPPT Control of Partially Shaded Photovoltaic Array Based on Random Frog Leaping Global Search Algorithm | |
| CN102258960A (en) | Method for automatically generating formula of fluidized bed based on neural network system | |
| CN110020724A (en) | A kind of neural network column Sparse methods based on weight conspicuousness | |
| CN103246934B (en) | Based on the electric system Equivalent Model parametric classification optimization method of trace sensitivity | |
| CN108182490A (en) | A Short-term Load Forecasting Method in Big Data Environment | |
| CN107634516A (en) | A Distribution Network State Estimation Method Based on Gray-Markov Chain | |
| CN107069708B (en) | Extreme learning machine-based transmission network line active safety correction method | |
| CN109711527A (en) | A kind of Robotic Manipulator method based on particle swarm optimization algorithm | |
| CN119203733A (en) | A smart emission reduction method and system based on big data | |
| CN108550180A (en) | Blood vessel modeling method based on interior point set region constraint and Gaussian process parameter optimization | |
| CN118625671A (en) | A control method of intelligent fitness equipment based on electronic load | |
| CN105069192B (en) | A kind of improved method that power of fan parameter of curve model is solved based on genetic algorithm | |
| CN116976192A (en) | JS-BP model-based die forging defect accurate repair process parameter decision method | |
| CN111279276A (en) | Randomized reinforcement learning for controlling complex systems | |
| Yue et al. | Power system short-term load forecasting based on neural network with artificial immune algorithm |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C12 | Rejection of a patent application after its publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20111130 |