CN108323797B - GPR (general purpose) model-based cigarette weight control system starting position prediction method and system - Google Patents
GPR (general purpose) model-based cigarette weight control system starting position prediction method and system Download PDFInfo
- Publication number
- CN108323797B CN108323797B CN201810134472.4A CN201810134472A CN108323797B CN 108323797 B CN108323797 B CN 108323797B CN 201810134472 A CN201810134472 A CN 201810134472A CN 108323797 B CN108323797 B CN 108323797B
- Authority
- CN
- China
- Prior art keywords
- starting position
- predicted value
- residual
- prediction
- model
- 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.)
- Active
Links
Images
Classifications
-
- A—HUMAN NECESSITIES
- A24—TOBACCO; CIGARS; CIGARETTES; SIMULATED SMOKING DEVICES; SMOKERS' REQUISITES
- A24C—MACHINES FOR MAKING CIGARS OR CIGARETTES
- A24C5/00—Making cigarettes; Making tipping materials for, or attaching filters or mouthpieces to, cigars or cigarettes
- A24C5/32—Separating, ordering, counting or examining cigarettes; Regulating the feeding of tobacco according to rod or cigarette condition
- A24C5/34—Examining cigarettes or the rod, e.g. for regulating the feeding of tobacco; Removing defective cigarettes
- A24C5/3424—Examining cigarettes or the rod, e.g. for regulating the feeding of tobacco; Removing defective cigarettes by weighing
Landscapes
- Feedback Control In General (AREA)
Abstract
The invention discloses a method and a system for predicting the starting position of a cigarette weight control system based on a GPR model, which comprises the steps of acquiring target position data of a collimator of the cigarette weight control system of a cigarette making machine before shutdown and target position data after starting, and respectively extracting a characteristic variable set and a target variable set from the target position data before shutdown and the target position data after starting; taking the characteristic variable set and the target variable set as training samples, and initializing and training a GPR model; obtaining a predicted value of a starting position through a constructed GPR model according to target position data before the shutdown, calculating a predicted residual of the starting position through the predicted value and an actual value of the starting position, and further constructing a residual prediction GPR model to obtain a predicted value of the residual; and correcting the predicted value of the starting position to obtain an output value of the starting position of the weight control system. The method has better robustness, and can effectively avoid the possible jump influence on the overall prediction result caused by a certain sudden change of the production state.
Description
Technical Field
The invention relates to the technical field of dynamic optimization of technological parameters of cigarette making machines, in particular to a method and a system for predicting the starting position of a cigarette weight control system based on a GPR model.
Background
At present, a cigarette weight control system of a cigarette making machine mainly comprises a cigarette density detector, a signal processing module, an industrial control computer, a leveling device (with two modes of cleaver movement and cut tobacco suction belt pinch roller movement), other transmission mechanisms, an adjusting mechanism and the like. Wherein, the leveler is the executive part of weight control, and the control of cigarette props up the lift through control leveler and realizes: when a cigarette passes through the detector, the average weight and the weight distribution condition of the cigarette can be obtained, a weight deviation signal is further obtained by comparing the average weight with the set cigarette weight, and then the weight deviation signal is fed back to the PLC and outputs a control pulse to drive the height adjusting motor of the leveler to drive the leveling device to lift. The leveler is in a predetermined activation position after the system is shut down or before the system is activated. When the cigarette making machine is started, the weight control system does not intervene in control before the cigarette making machine reaches a certain speed, the leveler is positioned at the starting position and does not participate in the regulation of the cigarette weight, and the cigarettes generally do not meet the weight requirement in the period of time; after the cigarette making machine reaches a certain speed, the weight control system is controlled in an intervention way, and the leveling device can regulate the cigarette weight and can ensure that the cigarette weight deviation is converged within a certain time.
At present, the setting of the starting position is mainly performed by manual experience, the initial debugging is generally adjusted according to a set value recommended by a manufacturer, and the initial debugging is generally set after certain compensation is performed on the basis of an average value of the target position of the leveler in a previous period of time in production. However, in actual production, the setting of the starting position is often kept unchanged for a long time, or the target position in the previous period fluctuates greatly, and the strategy of compensating according to the average value cannot really obtain a more ideal starting position. These experience-based approaches do not have strict experimental data or theoretical basis, have little effect on cigarette weight control at the starting stage of the cigarette making machine, and have poor objectivity. Meanwhile, the related process strategies can only be based on experience, and more cigarettes are removed in a conservative manner in the starting stage of the cigarette making machine to ensure the product quality, which causes considerable waste. Therefore, the starting position of the cigarette making machine is optimized, waste of fixed rejection in the starting stage can be reduced, and great economic benefits are achieved.
Because industrial processes often have multi-stage and time-varying dynamic characteristics, the prediction effect on process parameters cannot be ensured by adopting a general data-driven model. The GPR is a local nonparametric probability model which can be established by Gaussian process regression, and the obtained regression model can not only give a predicted value, but also obtain the confidence coefficient of the predicted value to the model. The method has a strict statistical learning theoretical basis, has good adaptability to processing complex problems of high dimension, small samples, nonlinearity and the like, and has strong generalization capability. Compared with models such as a neural network and a support vector machine, the GPR has the advantages of easiness in implementation, self-adaptive acquisition of hyper-parameters, flexibility in non-parameter inference, probability significance in output and the like.
Furthermore, the predictive performance of the static model is not satisfactory due to the time-varying nature of the production process and the diversity of the plant characteristics. In order to avoid the problems of poor model robustness, low prediction precision adaptability and the like, the prediction model needs to be continuously updated on line. The online learning-based method can well process the time-varying characteristic of the process and improve the dynamic performance of the prediction model. Compared with the traditional global model established offline, the model established by the online learning method has a local dynamic structure, is online and can better track the current state of the process.
Disclosure of Invention
The invention provides a method and a system for predicting the starting position of a cigarette weight control system based on a GPR model, aiming at the defects in the prior art.
In order to solve the technical problem, the invention is solved by the following technical scheme:
the present invention discloses:
a method for predicting the starting position of a cigarette weight control system based on a GPR model comprises the following steps:
acquiring target position data of a leveler of a weight control system of the cigarette making machine before shutdown and target position data after startup, and respectively extracting a characteristic variable set and a target variable set from the target position data before shutdown and the target position data after startup;
taking the characteristic variable set and the target variable set as training samples, and initializing and training a GPR model;
obtaining a predicted value of a starting position through a constructed GPR model according to target position data before the shutdown, calculating a predicted residual of the starting position through the predicted value and an actual value of the starting position, and further constructing a residual prediction GPR model to obtain a predicted value of the residual;
and correcting the predicted value of the starting position by using the predicted value of the residual error to obtain the output value of the starting position of the weight control system.
As an implementation manner, taking the feature variable set and the target variable set as training samples, initializing and training a GPR model specifically includes:
an initialization process: setting a covariance function and a hyper-parameter initial value by adopting a square exponential covariance function;
training process: establishing a negative log-likelihood function of the training sample conditional probability, and solving a partial derivative of the hyperparameter according to the negative log-likelihood function; and (3) minimizing the partial derivative by adopting a conjugate gradient method to obtain an optimal solution of the hyperparameter of the GPR model, and finally establishing the GPR model.
As an implementation manner, the specific steps of obtaining the predicted value of the residual error by obtaining the predicted value of the start position through the constructed GPR model according to the target position data before the current shutdown, obtaining the predicted residual error through calculating the predicted value and the actual value of the start position, and further constructing the residual error prediction GPR model include:
substituting target position data before the shutdown into the constructed GPR model to obtain a predicted value of a starting position;
obtaining a residual error of each prediction by using the predicted value and the actual value of the starting position, establishing a residual error set, and establishing a first characteristic variable set and a first target variable set by using the residual error set;
constructing a residual prediction GPR model for residual prediction according to the first characteristic variable set and the first target variable set;
and substituting the predicted value of the starting position into the residual prediction GPR model to obtain the predicted value of the residual prediction at this time.
As an implementation, the process of establishing the first characteristic variable set and the first target variable set by using the residual sets is:
and sequentially taking out continuous j residual errors from the first element in the residual error set as characteristic variables, correspondingly taking out the j +1 th residual error as a target variable, and repeating the steps to build l-j characteristic variable sets and l-j target variable sets.
As an implementation manner, the predicted value of the starting position is corrected by using the predicted value of the residual error to obtain the output value of the starting position of the weight control system, specifically,
and correcting the predicted value of the starting position by utilizing residual prediction, wherein the corrected predicted value of the starting position is as follows: and the corrected value is the predicted value of the starting position plus the predicted value of the residual error, and the corrected value is the output value of the starting position.
As an implementation mode, each halt trains a GPR model and a residual prediction GPR model according to data of the previous halts, and the GPR model and the residual prediction GPR model are respectively subjected to online learning.
The invention also discloses:
a cigarette weight control system starting position prediction system based on a GPR model comprises a data acquisition and processing module, a model construction module, a residual prediction model construction module and a correction output module;
the data acquisition and processing module is used for acquiring target position data of a leveler of the weight control system of the cigarette making machine before shutdown and target position data after startup, and extracting a characteristic variable set and a target variable set from the target position data before shutdown and the target position data after startup respectively;
the model building module is used for initializing and training a GPR model by taking the characteristic variable set and the target variable set as training samples;
the residual prediction model building module is used for obtaining a predicted value of a starting position through a built GPR model according to target position data before the shutdown, obtaining a predicted residual through calculation of the predicted value and an actual value of the starting position, and further building a residual prediction GPR model to obtain a predicted value of the residual;
and the correction output module is used for correcting the predicted value of the starting position by using the predicted value of the residual error to obtain the output value of the starting position of the weight control system.
As an implementation, the model building module comprises an initialization unit and a hyper-parameter determination unit;
the initialization unit is used for initializing: setting a covariance function and a hyper-parameter initial value by adopting a square exponential covariance function;
the hyper-parameter determining unit is used for establishing a negative log-likelihood function of the conditional probability of the training sample and solving a partial derivative of the hyper-parameter according to the negative log-likelihood function; and minimizing the partial derivative by adopting a conjugate gradient method to obtain an optimal solution of the hyperparameter of the GPR model, and establishing the GPR model through the optimal solution of the hyperparameter.
As an implementation manner, the residual prediction model building module comprises a prediction result obtaining unit, a training sample building unit, a prediction model building unit and a calculating unit;
the prediction result acquisition unit is used for substituting target position data before the shutdown into the constructed GPR model to obtain a prediction value of a starting position;
the training sample establishing unit is used for obtaining a residual error of each prediction by using the predicted value and the actual value of the starting position, establishing a residual error set, and establishing a first characteristic variable set and a first target variable set by using the residual error set;
the built prediction model unit is used for building a residual prediction GPR model for residual prediction according to the first characteristic variable set and the first target variable set;
and the computing unit is used for substituting the predicted value of the starting position into the residual prediction GPR model to obtain the predicted value of the residual prediction at this time.
The invention also discloses:
a computer readable storage medium storing a computer program which, when executed by a processor, performs the steps of a method and system for GPR model based cigarette weight control system start-up position prediction.
Due to the adoption of the technical scheme, the invention has the remarkable technical effects that:
the invention has the beneficial effects that: aiming at the problem that cigarette waste is caused by unreasonable manual experience setting method of starting position parameters of a cigarette making machine weight control system, a prediction model is trained according to data of previous shutdown times during shutdown each time, the model is subjected to online learning, the shutdown data is substituted into the model, and a recommended target position for next startup is quickly given, so that the target position can effectively reduce convergence time of cigarette weight deviation of the cigarette making machine in the startup stage, and waste of finished cigarettes is reduced.
The method has self-growing property, the state of the system can be optimized firstly after the prediction model is applied, and the optimized system can feed back and improve the model, so that the system can achieve a virtuous circle state along with the continuous online learning and optimization of system data accumulation. The method has better robustness, and can effectively avoid the possible jump influence on the overall prediction result caused by a certain sudden change of the production state.
The method and the system for predicting the starting position of the cigarette weight control system based on the GPR model can solve the pain point problem of waste of finished cigarettes in the existing cigarette industry, set the starting position of a cigarette making machine from the current extensive, experience dependence and manual intervention states into the data-driven, modeled and predictive states, and have immeasurable economic benefits and social benefits.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
FIG. 1 is a schematic overall flow diagram of the present invention;
FIG. 2 is a detailed flow diagram of FIG. 1;
FIG. 3 is a graph of the starting position of a cigarette making machine weight control system predicted by a GPR model in comparison to the actual starting position in accordance with one embodiment of the present invention;
FIG. 4 is a diagram illustrating a residue error of a cigarette weight control system using GPR model prediction according to an embodiment of the present invention;
fig. 5 is a schematic diagram of the overall system architecture of the present invention.
Reference numbers in the drawings: 100. a data acquisition processing module; 200. a model building module; 300. a residual prediction model construction module; 400. a correction output module; 210. an initialization unit; 220. a hyper-parameter determination unit; 310. a prediction result acquisition unit; 320. establishing a training sample unit; 330. constructing a prediction model unit; 340. a calculation unit;
Detailed Description
The present invention will be described in further detail with reference to examples, which are illustrative of the present invention and are not to be construed as being limited thereto.
The invention discloses:
a method for predicting the starting position of a cigarette weight control system based on a GPR model comprises the following steps as shown in figure 1:
s100, acquiring target position data of a leveler of a weight control system of the cigarette making machine before shutdown and target position data after startup, and respectively extracting a characteristic variable set and a target variable set from the target position data before shutdown and the target position data after startup;
s200, taking the characteristic variable set and the target variable set as training samples, and initializing and training a GPR model;
s300, obtaining a predicted value of a starting position through a constructed GPR model according to target position data before the shutdown, calculating a predicted residual error of the starting position through the predicted value and an actual value of the starting position, and further constructing a residual error prediction GPR model to obtain a predicted value of the residual error;
s400, correcting the predicted value of the starting position by using the predicted value of the residual error to obtain the output value of the starting position of the weight control system.
In step S200, the specific steps of initializing and training a GPR model using the feature variable set and the target variable set as training samples include:
s210, an initialization process: setting a covariance function and a hyper-parameter initial value by adopting a square exponential covariance function;
s220, establishing a negative log-likelihood function of the training sample conditional probability, and solving a partial derivative of the hyperparameter according to the negative log-likelihood function; and minimizing the partial derivative by adopting a conjugate gradient method to obtain an optimal solution of the hyperparameter of the GPR model, and establishing the GPR model through the optimal solution of the hyperparameter.
More specifically, the specific process of this embodiment is as follows:
in the initialization process, a square exponential covariance function is adopted, and a covariance function and a hyper-parameter initial value are set; variable Tp,TqThe squared exponential covariance function of (a) is:
in the formula: t isp,TqAny element in the characteristic variable set T obtained in the step;
is the data variance; m ═ diag (b)2) Is a diagonal matrix, the order of which is TpB is a variance scale;
the hyper-parameters of the model are:
in the formula
To observe the noise variance.
Establishing a negative log-likelihood function of the conditional probability of the training sample:
in the formula:
Kn＝(kpq) Covariance matrix, K, which is positive definite for n x n order symmetrypq＝k(TP,Tq) Is TP,TqSquared exponential covariance function of (a); f is the target variable set.
Solving a partial derivative of the hyper-parameter theta by using a negative log-likelihood function L (theta);
in the formula: α ═ C-1(ii) a tr () represents the trace of the matrix;
partial derivative by conjugate gradient method
Performing minimization to obtain hyper-parameters
The optimal solution of (1).
In step S300, the specific steps of obtaining the predicted value of the start position through the constructed GPR model according to the target position data before the current shutdown, obtaining the predicted residual thereof through calculation of the predicted value and the actual value of the start position, and further constructing a residual prediction GPR model to obtain the predicted value of the residual are:
s310, substituting target position data before the shutdown into the constructed GPR model to obtain a predicted value of a starting position;
s320, obtaining a residual error of each prediction by using the predicted value and the actual value of the starting position, establishing a residual error set, and establishing a first characteristic variable set and a first target variable set by using the residual error set, wherein the process of establishing the first characteristic variable set and the first target variable set by using the residual error set comprises the following steps: sequentially taking out continuous j residual errors from the first element in the residual error set as characteristic variables, correspondingly taking out the j +1 th residual error as a target variable, and repeating the steps to build l-j characteristic variable sets and l-j target variable sets;
s330, constructing a residual prediction GPR model for residual prediction according to the first characteristic variable set and the first target variable set;
and S340, substituting the predicted value of the starting position into the residual prediction GPR model to obtain the predicted value of the residual prediction at this time.
In this embodiment, assume that the residual set of the first 10 prediction results is taken
Taking out continuous 5 residual errors from the first element in the residual error set E as characteristic variables, correspondingly taking out the 6 th residual error as a target variable, and so on to construct 5 characteristic variable sets E in totalT＝{ETk＝{ek,ek+1...ek+41,2.. 5} and 5 sets of target variables EF＝{e6,e7...e10}；
Using a set of characteristic variables ETWith a target set of variables EFEstablishing a residual prediction GPR model for residual prediction according to the step S200;
set of residuals { e }6,e7,e8,e9,e10Substituting the predicted residual error into the established residual error prediction GPR model as a characteristic variable to obtain a predicted value e of the predicted residual error*；
Substituting target position data before the shutdown as system observation measurement T into a GPR model to obtain a predicted value f of the starting position:
in the formula: k (T)*，T)＝{k(T*T) | k ═ 1,2,.., n } is T*And a covariance matrix of order 1 xn between T; k (T)*，Tk) Is T*And TkThe covariance function of (a); k (T)*，Tk) The expression of (1) is a square exponential covariance function expression; kn＝(kpq) A covariance matrix which is n multiplied by n order symmetric positive definite; k is a radical ofpq＝k(Tp，Tq) Is Tp，TqSquared exponential covariance function of (a); i isnIs an n-dimensional identity matrix.
In step S400, the utilizing the predicted value of the residual error to the residual error is performedCorrecting the predicted value of the starting position to obtain an output value of the starting position of the weight control system, specifically, correcting the predicted value of the starting position by using residual prediction, wherein the corrected predicted value of the starting position is as follows: and the corrected value is the predicted value of the starting position plus the predicted value of the residual error, and the corrected value is the output value of the starting position. Here, the predicted value formulated as the starting position is corrected to
f*To predict value, e*In order to be a residual prediction value,
i.e. the final weight control system start position output value given.
The specific embodiment is as follows:
for the widely applied model ZJ112 cigarette making machine set, verification is performed by combining the above method, as shown in fig. 2, which is a flowchart of the cigarette weight control system start position prediction based on the online learning GPR model provided in this embodiment of the present invention, and includes the specific implementation steps of the method applied in this embodiment:
after the cigarette making machine is stopped, reading target position data of a leveling device before the cigarette making machine is stopped and target position data after the cigarette making machine is started from previous n-5 times of stop data;
standardizing the target position data of the leveler before the shutdown, intercepting the data of 3 minutes before the shutdown, and supplementing the data of less than 3 minutes by using the previous data, so as to construct a target variable set T (T) by using the target position of the leveler before the shutdownk|k＝1,2,…,5}；
Calculating the average value of the target positions in the first 3 seconds after starting to obtain a target variable set F ═ Fk|k＝1,2,…,5}；
Constructing a GPR model:
initializing a GPR model: a square exponential covariance function can be adopted, and a covariance function and a hyper-parameter initial value are set; variable Tp,TqThe squared exponential covariance function of (a) is:
in the formula: t isp,TqAny element in the characteristic variable set T obtained in the step;
is the data variance; m ═ diag (b)2) Is a diagonal matrix, the order of which is TpB is a variance scale;
the hyper-parameters of the GPR model are:
in the formula
To observe the noise variance.
Establishing a negative log-likelihood function of the conditional probability of the training sample:
in the formula:
Kn＝(kpq) Covariance matrix, K, which is positive definite for n x n order symmetrypq＝k(TP,Tq) Is TP,TqSquared exponential covariance function of (a); f is the target variable set.
Solving a partial derivative of the hyper-parameter theta by using a negative log-likelihood function L (theta);
wherein: α ═ C-1(ii) a tr () represents the trace of the matrix;
partial derivative by conjugate gradient method
Performing minimization to obtain hyper-parameters
The optimal solution of (1).
Taking the target position data 3 minutes before the shutdown as the system observation T*Substituting into GPR model to obtain the predicted value f of starting position*：
In the formula: k (T)*，T)＝{k(T*T) | k ═ 1,2,.., n } is T*And a covariance matrix of order 1 xn between T; k (T)*，Tk) Is T*And TkThe covariance function of (a); k (T)*，Tk) The expression of (2) is a square exponential covariance function expression; kn＝(kpq) Covariance matrix, k, which is positive definite for n x n order symmetrypq＝k(Tp，Tq) Is Tp，TqSquared exponential covariance function of InIs an n-dimensional identity matrix.
Residual prediction:
taking the residual set of the first 10 prediction results
Taking out continuous 5 residual errors from the first element in the residual error set E as characteristic variables, correspondingly taking out the 6 th residual error as a target variable, and so on to construct 5 characteristic variable sets E in totalT＝{ETk＝{ek,ek+1...ek+41,2.. 5} and 5 sets of target variables EF＝{e6,e7...e10}；
Using 5 sets of characteristic variables ETAnd 5 target variable sets EFEstablishing a residual prediction GPR model for residual prediction by adopting the method of the step S200;
substituting the residual set as a characteristic variable into the established residual prediction GPR model to obtain a predicted value e of the predicted residual*。
And (4) outputting a prediction result: using the obtained prediction value f*And the obtained residual error predicted value e*Correcting the predicted value of the starting position to
I.e. the final weight control system start position output value given.
After the data verification of 115 times of shutdown of a certain ZJ112 type cigarette making and tipping machine set, the cigarette weight control system based on the GPR model is adopted to start the position prediction method and system, the prediction result of the starting position is shown in figure 3, the comparison error between the prediction result and the actual starting position is small, as shown in figure 4, the root mean square error is 0.41, the accuracy is very ideal, and the method is applied to production and can greatly reduce the cigarette loss of a cigarette making machine in the starting stage.
The invention also discloses:
a cigarette weight control system starting position prediction system based on an online learning GPR model is shown in FIG. 5 and comprises a data acquisition and processing module 100, a model construction module 200, a residual prediction model construction module 300 and a correction output module 400;
the data acquisition and processing module 100 is used for acquiring target position data of a leveler of the weight control system of the cigarette making machine before shutdown and target position data after startup, and extracting a characteristic variable set and a target variable set from the target position data before shutdown and the target position data after startup respectively;
the model building module 200 is configured to initialize and train a GPR model by using the feature variable set and the target variable set as training samples;
the residual prediction model building module 300 is configured to obtain a predicted value of a starting position through a built GPR model according to target position data before the current shutdown, obtain a predicted residual thereof through calculation of the predicted value and an actual value of the starting position, and further build a residual prediction GPR model to obtain a predicted value of the residual;
and the correction output module 400 is configured to correct the predicted value of the start position by using the predicted value of the residual error, so as to obtain an output value of the start position of the weight control system.
Further, the model building module 200 includes an initialization unit 210 and a hyper-parameter determination unit 220;
the initialization unit 210 is configured to initialize: setting a covariance function and a hyper-parameter initial value by adopting a square exponential covariance function;
the hyper-parameter determining unit 220 is configured to establish a negative log-likelihood function of the conditional probability of the training sample, and calculate a partial derivative of the hyper-parameter according to the negative log-likelihood function; and minimizing the partial derivative by adopting a conjugate gradient method to obtain an optimal solution of the hyperparameter of the GPR model, and establishing the GPR model through the optimal solution of the hyperparameter.
In other embodiments, the residual prediction model building module 300 includes a prediction result obtaining unit 310, a training sample building unit 320, a prediction model building unit 330, and a calculating unit 340;
the prediction result obtaining unit 310 is configured to substitute target position data before the current shutdown into the constructed GPR model to obtain a prediction value of a starting position;
the training sample establishing unit 320 is configured to obtain a residual error of each prediction by using the predicted value and the actual value of the starting position, establish a residual error set, and establish a first characteristic variable set and a first target variable set by using the residual error set;
the build prediction model unit 330 is configured to build a residual prediction GPR model for performing residual prediction according to the first characteristic variable set and the first target variable set;
and the calculating unit 340 is configured to substitute the predicted value of the starting position into the residual prediction GPR model to obtain a predicted value of the current prediction residual.
The invention also discloses:
a computer readable storage medium storing a computer program which, when executed by a processor, performs the steps of a method and system for GPR model based cigarette weight control system start-up position prediction.
For the device embodiment, since it is basically similar to the method embodiment, the description is simple, and for the relevant points, refer to the partial description of the method embodiment.
The embodiments in the present specification are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, apparatus, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention has been described with reference to flowchart illustrations and/or block diagrams of methods, terminal devices (systems), and computer program products according to the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing terminal to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing terminal to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing terminal to cause a series of operational steps to be performed on the computer or other programmable terminal to produce a computer implemented process such that the instructions which execute on the computer or other programmable terminal provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
It should be noted that:
reference in the specification to "one embodiment" or "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the invention. Thus, the appearances of the phrase "one embodiment" or "an embodiment" in various places throughout this specification are not necessarily all referring to the same embodiment.
While preferred embodiments of the present invention have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, it is intended that the appended claims be interpreted as including preferred embodiments and all such alterations and modifications as fall within the scope of the invention.
In addition, it should be noted that the specific embodiments described in the present specification may differ in the shape of the components, the names of the components, and the like. All equivalent or simple changes of the structure, the characteristics and the principle of the invention which are described in the patent conception of the invention are included in the protection scope of the patent of the invention. Various modifications, additions and substitutions for the specific embodiments described may be made by those skilled in the art without departing from the scope of the invention as defined in the accompanying claims.
Claims (10)
1. A method for predicting the starting position of a cigarette weight control system based on a GPR model is characterized by comprising the following steps:
acquiring target position data of a leveler of a weight control system of the cigarette making machine before shutdown and target position data after startup, and respectively extracting a characteristic variable set and a target variable set from the target position data before shutdown and the target position data after startup;
taking the characteristic variable set and the target variable set as training samples, and initializing and training a GPR model;
obtaining a predicted value of a starting position through a constructed GPR model according to target position data before the shutdown, calculating a predicted residual of the starting position through the predicted value and an actual value of the starting position, and further constructing a residual prediction GPR model to obtain a predicted value of the residual;
and correcting the predicted value of the starting position by using the predicted value of the residual error to obtain the output value of the starting position of the weight control system.
2. The method for predicting the starting position of the cigarette weight control system based on the GPR model as claimed in claim 1, wherein the specific steps of initializing and training the GPR model by using the feature variable set and the target variable set as training samples are as follows:
an initialization process: setting a covariance function and a hyper-parameter initial value by adopting a square exponential covariance function;
training process: establishing a negative log-likelihood function of the training sample conditional probability, and solving a partial derivative of the hyperparameter according to the negative log-likelihood function; and (3) minimizing the partial derivative by adopting a conjugate gradient method to obtain an optimal solution of the hyperparameter of the GPR model, and finally establishing the GPR model.
3. The method for predicting the starting position of the cigarette weight control system based on the GPR model as claimed in claim 2, wherein the method comprises the specific steps of obtaining a predicted value of the starting position through the constructed GPR model according to target position data before the halt, obtaining a predicted residual error through calculation of the predicted value and an actual value of the starting position, and further constructing a residual error prediction GPR model, wherein the predicted value of the residual error is obtained through the specific steps of:
substituting target position data before the shutdown into the constructed GPR model to obtain a predicted value of a starting position;
obtaining a residual error of each prediction by using the predicted value and the actual value of the starting position, establishing a residual error set, and establishing a first characteristic variable set and a first target variable set by using the residual error set;
constructing a residual prediction GPR model for residual prediction according to the first characteristic variable set and the first target variable set;
and substituting the predicted value of the starting position into the residual prediction GPR model to obtain the predicted value of the residual prediction at this time.
4. The method for predicting the starting position of the cigarette weight control system based on the GPR model as claimed in claim 3, wherein the process for establishing the first characteristic variable set and the first target variable set by using the residual sets comprises the following steps:
and sequentially taking out continuous j residual errors from the first element in the residual error set as characteristic variables, correspondingly taking out the j +1 th residual error as a target variable, and repeating the steps to build l-j characteristic variable sets and l-j target variable sets.
5. The method for predicting the starting position of the cigarette weight control system based on the GPR model as claimed in claim 3, wherein the predicted value of the starting position is corrected by using the predicted value of the residual error to obtain the output value of the starting position of the weight control system,
and correcting the predicted value of the starting position by utilizing residual prediction, wherein the corrected predicted value of the starting position is as follows: and the corrected value is the predicted value of the starting position plus the predicted value of the residual error, and the corrected value is the output value of the starting position.
6. The method for predicting the starting position of the cigarette weight control system based on the GPR model is characterized in that each halt is implemented by training the GPR model and the residual prediction GPR model according to data of the first few halts, and the GPR model and the residual prediction GPR model are respectively subjected to online learning.
7. A cigarette weight control system starting position prediction system based on a GPR model is characterized by comprising a data acquisition and processing module, a model construction module, a residual prediction model construction module and a correction output module;
the data acquisition and processing module is used for acquiring target position data of a leveler of the weight control system of the cigarette making machine before shutdown and target position data after startup, and extracting a characteristic variable set and a target variable set from the target position data before shutdown and the target position data after startup respectively;
the model building module is used for initializing and training a GPR model by taking the characteristic variable set and the target variable set as training samples;
the residual prediction model building module is used for obtaining a predicted value of a starting position through a built GPR model according to target position data before the shutdown, obtaining a predicted residual through calculation of the predicted value and an actual value of the starting position, and further building a residual prediction GPR model to obtain a predicted value of the residual;
and the correction output module is used for correcting the predicted value of the starting position by using the predicted value of the residual error to obtain the output value of the starting position of the weight control system.
8. The GPR model-based cigarette weight control system starting position prediction system as claimed in claim 7, wherein the model building module comprises an initialization unit and a hyper-parameter determination unit;
the initialization unit is used for initializing: setting a covariance function and a hyper-parameter initial value by adopting a square exponential covariance function;
the hyper-parameter determining unit is used for establishing a negative log-likelihood function of the conditional probability of the training sample and solving a partial derivative of the hyper-parameter according to the negative log-likelihood function; and minimizing the partial derivative by adopting a conjugate gradient method to obtain an optimal solution of the hyperparameter of the GPR model, and establishing the GPR model through the optimal solution of the hyperparameter.
9. The GPR model-based cigarette weight control system starting position prediction system as recited in claim 8, wherein the residual prediction model building module comprises a prediction result obtaining unit, a training sample building unit, a prediction model building unit and a calculation unit;
the prediction result acquisition unit is used for substituting target position data before the shutdown into the constructed GPR model to obtain a prediction value of a starting position;
the training sample establishing unit is used for obtaining a residual error of each prediction by using the predicted value and the actual value of the starting position, establishing a residual error set, and establishing a first characteristic variable set and a first target variable set by using the residual error set;
the built prediction model unit is used for building a residual prediction GPR model for residual prediction according to the first characteristic variable set and the first target variable set;
and the computing unit is used for substituting the predicted value of the starting position into the residual prediction GPR model to obtain the predicted value of the residual prediction at this time.
10. A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out the steps of the method according to any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810134472.4A CN108323797B (en) | 2018-02-09 | 2018-02-09 | GPR (general purpose) model-based cigarette weight control system starting position prediction method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810134472.4A CN108323797B (en) | 2018-02-09 | 2018-02-09 | GPR (general purpose) model-based cigarette weight control system starting position prediction method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108323797A CN108323797A (en) | 2018-07-27 |
| CN108323797B true CN108323797B (en) | 2020-11-24 |
Family
ID=62927369
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810134472.4A Active CN108323797B (en) | 2018-02-09 | 2018-02-09 | GPR (general purpose) model-based cigarette weight control system starting position prediction method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108323797B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109330018B (en) * | 2018-10-30 | 2021-05-18 | 浙江中烟工业有限责任公司 | Method for setting starting position of cut tobacco suction belt of cigarette weight control system of cigarette making machine |
| CN110037336B (en) * | 2019-04-19 | 2021-09-03 | 浙江中烟工业有限责任公司 | Method for predicting position of actuating mechanism of cigarette circumference control system |
| CN111562574B (en) * | 2020-05-22 | 2022-08-16 | 中国科学院空天信息创新研究院 | MIMO ground penetrating radar three-dimensional imaging method based on backward projection |
| CN111932037A (en) * | 2020-09-23 | 2020-11-13 | 浙江创泰科技有限公司 | Parking space state prediction method and system based on machine learning |
| CN113313106A (en) * | 2021-04-14 | 2021-08-27 | 深圳市睿达科技有限公司 | Feeding deviation rectifying method and device, computer equipment and storage medium |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US3146910A (en) * | 1960-11-07 | 1964-09-01 | Industrial Nucleonics Corp | Control system |
| GB1313861A (en) * | 1969-06-02 | 1973-04-18 | Industrial Nucleonics Corp | System and method for optimizing processor or equipment profit |
| CN201640440U (en) * | 2010-03-19 | 2010-11-24 | 河南中烟工业有限责任公司 | Cigarette weight control system of cigarette making machine |
| CN104048675A (en) * | 2014-06-26 | 2014-09-17 | 东南大学 | Integrated navigation system fault diagnosis method based on Gaussian process regression |
| CN105105326A (en) * | 2015-07-28 | 2015-12-02 | 郑州长河电子工程有限公司 | Industrial control computer (IPC)-based cigarette making machine tobacco rod weight control and quality detection integrated device and method |
| CN205176622U (en) * | 2015-12-11 | 2016-04-20 | 成都博发控制技术有限责任公司 | Stable and weight optimal control system of a cigarette feed |
| CN205337572U (en) * | 2016-01-20 | 2016-06-29 | 河南中烟工业有限责任公司 | Cigarette machine induced draft room and cigarette machine |
| CN205390286U (en) * | 2016-02-29 | 2016-07-27 | 宝鸡市信诚电子有限公司 | Novel cigarette zhi chongliang control structure |
| CN106056127A (en) * | 2016-04-07 | 2016-10-26 | 江南大学 | GPR (gaussian process regression) online soft measurement method with model updating |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE19919207A1 (en) * | 1999-04-28 | 2000-11-02 | Focke & Co | Method and device for testing cigarettes |
| JP3466989B2 (en) * | 2000-05-24 | 2003-11-17 | 日本たばこ産業株式会社 | Combustion test equipment |
| CN100444153C (en) * | 2005-11-28 | 2008-12-17 | 颐中烟草(集团)有限公司 | Cigarette internal quality index extimating method based on regression function estimating SVM |
| CN102406234B (en) * | 2011-07-13 | 2013-09-11 | 常德烟草机械有限责任公司 | Cigarette position signal generation and quality detection rejection method |
| CN103385539B (en) * | 2013-08-02 | 2015-04-15 | 南京文采科技有限责任公司 | Single cigarette empty head detection method based on machine vision and special equipment |
-
2018
- 2018-02-09 CN CN201810134472.4A patent/CN108323797B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US3146910A (en) * | 1960-11-07 | 1964-09-01 | Industrial Nucleonics Corp | Control system |
| GB1313861A (en) * | 1969-06-02 | 1973-04-18 | Industrial Nucleonics Corp | System and method for optimizing processor or equipment profit |
| CN201640440U (en) * | 2010-03-19 | 2010-11-24 | 河南中烟工业有限责任公司 | Cigarette weight control system of cigarette making machine |
| CN104048675A (en) * | 2014-06-26 | 2014-09-17 | 东南大学 | Integrated navigation system fault diagnosis method based on Gaussian process regression |
| CN105105326A (en) * | 2015-07-28 | 2015-12-02 | 郑州长河电子工程有限公司 | Industrial control computer (IPC)-based cigarette making machine tobacco rod weight control and quality detection integrated device and method |
| CN205176622U (en) * | 2015-12-11 | 2016-04-20 | 成都博发控制技术有限责任公司 | Stable and weight optimal control system of a cigarette feed |
| CN205337572U (en) * | 2016-01-20 | 2016-06-29 | 河南中烟工业有限责任公司 | Cigarette machine induced draft room and cigarette machine |
| CN205390286U (en) * | 2016-02-29 | 2016-07-27 | 宝鸡市信诚电子有限公司 | Novel cigarette zhi chongliang control structure |
| CN106056127A (en) * | 2016-04-07 | 2016-10-26 | 江南大学 | GPR (gaussian process regression) online soft measurement method with model updating |
Non-Patent Citations (1)
| Title |
|---|
| 卷烟机数字PID重量控制***的实现;仲伟庆;《烟草科技》;20050630(第6期);第18-20页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108323797A (en) | 2018-07-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108323797B (en) | GPR (general purpose) model-based cigarette weight control system starting position prediction method and system | |
| CN109382838B (en) | Control device and machine learning device | |
| Deshpande et al. | Intelligent state estimation for fault tolerant nonlinear predictive control | |
| US11156988B2 (en) | Drive apparatus and machine learning apparatus | |
| CN110471276B (en) | Apparatus for creating model functions for physical systems | |
| CN103472723A (en) | Predictive control method and system based on multi-model generalized predictive controller | |
| CN102985884A (en) | A method and system for updating a model in a model predictive controller | |
| CN111433689B (en) | Generation of control systems for target systems | |
| CN101390024A (en) | Operation control method, operation control device and operation control system | |
| JP7151547B2 (en) | Predictive control development device, predictive control development method, and predictive control development program | |
| CN105807615A (en) | Fuzzy feedforward-feedback controller | |
| Hua | Reinforcement Learning Aided Performance Optimization of Feedback Control Systems | |
| CN112819224B (en) | Unit output prediction and confidence evaluation method based on deep learning fusion model | |
| CN116880191A (en) | Intelligent control method of process industrial production system based on time sequence prediction | |
| CN117001965A (en) | Constraint tracking control method and system for batch injection molding process | |
| CN116520703A (en) | Reinforced learning intermittent process control method based on improved AC algorithm | |
| Engin et al. | Auto-tuning of PID parameters with programmable logic controller | |
| CN109635465B (en) | Batch operation track adjusting method based on TPLS model | |
| CN114442557A (en) | Method and system for quickly identifying temperature field of machine tool | |
| CN110037336B (en) | Method for predicting position of actuating mechanism of cigarette circumference control system | |
| CN116954156B (en) | Numerical control processing process route planning method, device, equipment and medium | |
| CN115877700A (en) | Process control method and device | |
| WO2021245916A1 (en) | Servo control device | |
| CN115877811B (en) | Flow process treatment method, device and equipment | |
| CN111752152B (en) | LSTM neural network-based parameter self-tuning method for SISO (SISO) compact format model-free controller |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |