CN102390096A - On-line automatic measurement method for Mooney viscosity of rubber - Google Patents

Abstract

The invention discloses an on-line automatic measurement method for the Mooney viscosity of rubber. The method comprises the following steps of: inputting rheological parameters of final mixed rubber which are obtained by a quality inspection system into an on-line Mooney viscosity prediction model, and automatically outputting a Mooney viscosity value ynew; feeding back the Mooney viscosity value ynew to the quality inspection system, judging whether the Mooney viscosity value ynew is within a preset range value of the Mooney viscosity by the quality inspection system, continuing the production and updating the on-line Mooney viscosity prediction model if the Mooney viscosity value ynew is within the preset range value of the Mooney viscosity, and giving an alarm, checking the production flow and correcting by operating personnel and abandoning the Mooney viscosity value ynew which exceeds the range if the Mooney viscosity value ynew is not within the preset range value of the Mooney viscosity; and repeating the operation, and finishing the flow until the rheological parameters of the final mixed rubber are not input. By the method, the stay time of the rubber is furthest reduced on the premise that sampling burden is not increased, so the continuity of the production is ensured to a certain extent; a real-time predicted value of the Mooney viscosity is calculated by using the on-line Mooney viscosity prediction model; meanwhile, the reliability of the model is ensured.

Description

A kind of on-line automatic measuring method of Mooney viscosity of rubber

Technical field

The present invention relates to rubber tyre and make the quality method for real-time monitoring in field, particularly a kind of on-line automatic measuring method of Mooney viscosity of rubber.

Background technology

Rubber industry is one of important foundation industry of national economy.It not only provides daily life indispensable light industry rubber product such as daily, medical for people, and to heavy industry such as digging, traffic, building, machinery, electronics and new industry various rubber system production equipments or rubber components is provided.Along with improving constantly of production technology level, market oriented management style is had higher requirement to the quality of rubber product.

Compounding rubber is with rubber mixing machine rubber or plasticate rubber and compounding ingredient to be smelt the technology of rubber, is the most important production technology of rubber processing.Essence is compounding ingredient homodisperse process in rubber, and granular compounding ingredient is decentralized photo, and rubber is continuous phase.The quality of the rubber mass behind the adding compounding ingredient all has decisive influence to half-finished processing performance and end product quality.It is particularly important that this point seems in the production process of rubber tyre.So guaranteeing the quality of rubber is the basic demand in the rubber processing process.Rubber mixing process has stronger time variation, non-linear, is typical industrial batch production process.Therefore, obtaining real-time, reliable rubber product qualitative data is to guarantee to produce successional key.Mooney viscosity is an important indicator of weighing rubber processing performance quality, but it has reflected many-sided performances such as viscosity characteristics processing characteristics and the calendering characteristic of sizing material.

In the current domestic rubber tyre production process, the detection of Mooney viscosity of rubber relies on mainly that quality inspection personnel is manually sampled, sample preparation and use pertinent instruments to measure, and mainly passes through following steps:

1, sizing material through banburying, extrude, roll, open operation such as refining after, through interleaving agent, blower fan cooling, the relevant train number information of air-cooled back lamination and record;

2, the rubber of folding need be parked certain hour (4～8 hours), makes its material characteristic stable, then by quality inspection personnel sampling censorship;

3, manually dash appearance in fast inspection chamber, prepare suitable sample;

4, adopt the Mooney viscosity and the record of Mooney viscosity appearance measuring samples.

It is thus clear that the sizing material quality index that obtains thus---Mooney viscosity obviously lags behind actual production, make the continuity of production process reduce greatly.Per car time rubber need be parked more than 4 hours at least; To be detectedly qualifiedly can carry out following process, only need 2～3 minutes and per car time sizing material is mixing, the restriction of technical merit; Make the verification and measurement ratio of Mooney viscosity of rubber usually less than 20%; So the serious lag effect causes production efficiency significantly to reduce, and is seriously restricting the popularization and application of various advanced control technologys and the further raising of product quality, makes the manufacturer of rubber tyre be faced with huge economic risk.In addition, because in the measuring process, cut-parts, sampling, dash work such as appearance and accomplish by manual work, increased the uncertainty of measurement data, further affect the quality of rubber, processing characteristics also can not get guaranteeing.And this detection method needs special staff, is equipped with many Mooney detecting instruments, has increased various manpowers, financial resources, material resources cost, has reduced the enterprise production benefit.Therefore, hysteresis quality and uncertainty that Mooney viscosity value detects are seriously restricting the development of rubber mixing process for a long time, are the improved bottleneck problems of rubber tyre production technology.

Summary of the invention

The technical problem that the present invention will solve is to provide a kind of on-line automatic measuring method of Mooney viscosity of rubber, and this method can realize real-time measurement Mooney viscosity, measurement result accurately with become the product low cost and other advantages, see the hereinafter description for details:

A kind of on-line automatic measuring method of Mooney viscosity of rubber said method comprising the steps of:

(1) the finished composition rheological parameter input Mooney viscosity on-line prediction model that quality inspection system is obtained is exported Mooney viscosity value y automatically _New

(2) with said Mooney viscosity value y _NewFeed back to said quality inspection system, said quality inspection system is judged said Mooney viscosity value y _NewWhether in Mooney viscosity preset range value, if, continue to produce, upgrade said Mooney viscosity on-line prediction model; If not, give the alarm, check production procedure and proofread and correct by operating personnel, give up off-limits Mooney viscosity value y _New

(3) repeated execution of steps (1)-step (2), when no longer including the input of said finished composition rheological parameter, flow process finishes.

Said Mooney viscosity on-line prediction model is specially:

1) at first gathers said finished composition rheological parameter, set up original sample collection X _Old, to said original sample collection X _OldCarry out the multinomial transformation and obtain X _Extra, form X _New=[X _OldX _Extra];

2), extract X through offset minimum binary algorithm (PLS) _NewLatent variable u and the latent variable t of Y;

w＝X ^T×Y/(Y ^T×Y)；

w＝w/sqrt(w ^T×w)；

t＝X×w：

q＝Y ^T×t/(t ^T×t)；

u＝Y×q/(q ^T×q)；

Wherein, Y representes the Mooney viscosity value of normal range (NR), and w representes the weight of X, and q is the load vector of Y;

3) utilize Gaussian process to set up the regression relation of said latent variable u and said latent variable t;

4) calculated data matrix X _NewResidual error E and the residual error F of Y, be back to the 2nd) step, extract second couple of latent variable u and t once more;

5) repeating step 3) and step 4), less than convergency value, be convergence until the increment of prediction residual quadratic sum, obtain said Mooney viscosity on-line prediction model.

The on-line automatic measuring method of a kind of Mooney viscosity of rubber provided by the invention compared with prior art has following advantage:

According to verification and measurement ratio in the rubber mixing process is that 100% rheological parameter is predicted Mooney viscosity, under the prerequisite that does not increase the sampling burden, reduces the rubber storage period to greatest extent, guarantees the continuity of producing to a certain extent; Utilize Mooney viscosity on-line prediction Model Calculation to go out the real-time estimate value of Mooney viscosity; Guarantee the reliability of model simultaneously, promptly bring in constant renewal in Mooney viscosity on-line prediction model, rubber mixing process is monitored in real time, can reflect in real time and follow the tracks of production status, make Mooney viscosity on-line prediction model more can embody existing production characteristic.

Description of drawings

Fig. 1 is the sketch map of Mooney viscosity on-line measurement model provided by the invention;

Fig. 2 is the flow chart of the on-line automatic measuring method of a kind of Mooney viscosity of rubber provided by the invention.

The specific embodiment

For making the object of the invention, technical scheme and advantage clearer, will combine accompanying drawing that embodiment of the present invention is done to describe in detail further below.

Each item index that flow graph detects is the important component part of rubber quality system, and common detection time is shorter, only needs about 2 minutes, and need not park the back and detect.Investigation shows, exists between rheology index and the Mooney viscosity to be closely connected.So, estimate or predict that Mooney viscosity value is to solve the effective way that Mooney viscosity detects problem through rheological parameter.

Flow graph detects each parameter that obtains and has reflected that to a certain extent sizing material quality and processing characteristics are the important component parts of rubber quality system.Mainly comprise minimum torque, the highest moment of torsion and cure time parameter.The verification and measurement ratio of rheological parameter is 100%, only needs 2 minutes at every turn, and need not park the back and detect.Investigation shows, exists between rheology index and the Mooney viscosity to be closely connected.So adopt the method for data-driven, promptly estimate or predict that Mooney viscosity value is the effective way that solves Mooney viscosity detection problem through rheological parameter.

The development of Along with computer technology and industrial automation has all obtained extensive use in a lot of fields based on the method for data-driven, comprises process industrial, commerce and financial circles etc.Usually, these methods, are analyzed and the extraction relevant information, to tackle the current or following decision-making needs by statistical means according to historical data.But, many times, though historical data is more, the information that comprises lacks relatively, and the development that this has also promoted statistical method to a certain extent impels people to seek more effectively, mathematical tool solves relevant issues more reliably.For many years; Scientists has proposed many mathematical methods based on statistics, like PCA (Principal Component Analysis, principal component analysis), PLS (Partial Least Squares; Partial least square), ANN (Artificial Neural Networks; Artificial neural network), SVM (Support Vector Machine, SVMs) and GP (Gaussian Process, Gaussian process) or the like.Wherein, linear method early such as PCA, PLS, convenient and reliable owing to it should be readily appreciated that, still be widely used so far.They are through extracting latent variable, and overcome the dimension that correlation between the variable reduces variable, improve computational efficiency, are not only applicable to the small sample data set, also are fit to the big-sample data collection.Yet these methods just have been not enough to the correlation between decryption information and the data when changing in the face of complex process.What many times, the industrial process of complicacy showed is non-linear stronger.Between the quality index or a certain particular demands index and various measurable variable like product, owing to influenced by complicated external environment, or the change of properties of itself, present complicated non-linear relation.At this moment just need reasonable nonlinear method to solve corresponding problem.

GP is a kind of new nonlinear algorithm that last decade cause science researcher proposes, and it is non-linear, exhibits excellent on the small sample data set.Be a kind of machine learning algorithm of nuclear study of probability meaning, it adopts the method for iteration to optimize learning parameter.But GP also can meet difficulty when setting up Mathematical Modeling in some aspects.Such as, when the input variable dimension than higher the time, need more time to come parameters optimization, significantly increased and assessed the cost.A kind of effective instrument is PCA, reduces the dimension of input variable.Use PLS perhaps more effective, because it has considered influencing each other between the input and output, and PCA does not consider this point.Because its prerequisite hypothesis of the method for many data-drivens is exactly the sample Gaussian distributed.In fact, for rubber mixing process, because various noise of instrument or measurement The noise, measurement result satisfies on GP distribution this point hypothesis.So, in conjunction with the advantage of PLS and GP, can obtain more effective Mathematical Modeling, come the data of description relation.

Because PLS is linear, in order to obtain better model accuracy, science researchers have developed some nonlinear PLS algorithms; Like Kernel PLS (KPLS), Neural Network PLS (NNPLS), Generalized PLS (GPLS); Wherein, GPLS is a kind of PLS method that adopts multinomial to the input data transaction, and definition original input data matrix is X, X=(x ₁, x ₂..., x _l) ^T, so new input variable is X _New=(x ₁, x ₂..., x _l..., x _L+s).Just: X _New=[X _OldX _Extra].

Wherein s=l+ (l-1)+(l-2)+...+1, X _OldBe original input variable X, X _ExtraBe the transformation variable that adds, X _ExtraEach variable be respectively x _Ijx _Ik, x _Ijx _Ik, x _Im, wherein (i=1,2..., n; J, k, m=1,2 ..., l).Other steps are identical with common PLS.

In recent years, Gaussian process has attracted many researchers' attention as a kind of effective modeling tool, and it not only can solve regression problem, also can solve classification problem, and many research work show that it is more effective under the part situation than ANN and SVM.It is a kind of probability nuclear learning machine based on bayesian theory.Generally, think that Gaussian process is the set of stochastic variable, the associating Gaussian distribution is obeyed in the combination of any limited stochastic variable.Gaussian process can be definite fully by a mean value function and a covariance function, generally speaking, gets 0 as its mean value function.

f(x)～GP(0，C)

Wherein C is N rank covariance matrixes, and following covariance function form has been proved to be in most of the cases and is suitable for and does well:

$C(x_i,x_j)=v_0\exp \{-\frac{1}{2}\underset{l=1}{\overset{d}{\mathrm{\Σ}}}{w}_l{(x_{\mathrm{il}}-x_{\mathrm{jl}})}^2\}+a_0+a_1\underset{l=1}{\overset{d}{\mathrm{\Σ}}}{x}_{\mathrm{il}}{x}_{\mathrm{jl}}+v_1{\mathrm{\δ}}_{\mathrm{ij}}$

X wherein _iBe i variable, and when i=j δ _Ij=1, θ=log (v ₀, v ₁, w ₁..., w _d, a ₀, a ₁) be the ultra parameter of model

For a new tested point, its output distributes and to remain Gauss, and its average and variance are respectively:

${\hat{y}}^{*}\left(x^{*}\right)=k^T\left(x^{*}\right)K^{-1}y$

${\mathrm{\σ}}_{{\hat{y}}^{*}}^2\left(x^{*}\right)=C(x^{*},x^{*})-k^T\left(x^{*}\right)K^{-1}k\left(x^{*}\right)$

Wherein, k (x)=(C (x ^*, x ₁) ..., C (x ^*, x _n)) ^T, K _Ij=C (x _i, x _j).

* number expression new samples in several in the above formulas.

Through following likelihood function, use the method for maximum a posteriori estimation or Markov Chain Monte Carlo, can obtain the ultra parameter of optimum of model,

$L=-\frac{1}{2}\log \det C-\frac{1}{2}{t}^T{C}^{-1}t-\frac{n}{2}\log 2\mathrm{\π}$

Describe the on-line automatic measuring method of a kind of Mooney viscosity of rubber that the embodiment of the invention provides below in detail by the practical implementation process.

101:, export Mooney viscosity value y automatically with the finished composition rheological parameter input Mooney viscosity on-line prediction model that quality inspection system obtains _New

Wherein, this step is specially: after current train number mixing process finishes, quality inspection system is detected the finished composition rheological parameter information input Mooney viscosity on-line prediction model that obtains, export Mooney viscosity value y automatically _New

Wherein, The process of setting up of this Mooney viscosity on-line prediction model is: according to two main quality index---the connecting each other between Mooney viscosity and the rheological parameter of rubber; Set up Mooney viscosity on-line prediction model GPLS-GP in conjunction with improved partial least-square regression method GPLS and Gaussian process GP according to rheological parameter, obtain final predicted value as the Mooney viscosity reference value.

The embodiment of the invention is based on Analysis on Mechanism and lot of experiment validation; Set up Mooney viscosity on-line prediction model GPLS-GP thus; Utilize rheological parameter to dope Mooney viscosity value, model description is following: set up initial model according to existing historical creation data, the initial model sample number is 30.Produce just often, every acquisition one train number stream of rubber variable element then substitutes the sample in the archetype with corresponding Mooney viscosity value, guarantees that total sample number is no less than 30 in the model.

1) at first gathers the finished composition rheological parameter, set up original sample collection X _Old, to original sample collection X _OldCarry out the multinomial transformation and obtain X _Extra, form X _New=[X _OldX _Extra];

Wherein, original sample collection X _OldIn the sample point number surpass 30 and upgrade original sample collection X _Old, raw sample data matrix X is obtained new data matrix X through transforming _NewThereby, guaranteeing the non-linear of model, this is the GPLS condition.

2), extract X through offset minimum binary algorithm (PLS) _NewLatent variable u and the latent variable t of Y;

w＝X ^T×Y/(Y ^T×Y)；

w＝w/sqrt(w ^T×w)；

t＝X×w；

q＝Y ^T×t/(t ^T×t)；

u＝Y×q/(q ^T×q)；

Wherein, Y representes the Mooney viscosity value of normal range (NR), is the historical experience data before the modeling, then substitutes with normal predicted value after the modeling, and w representes the weight of X, and q is the load vector of Y.

3) utilize Gaussian process to set up the regression relation of u and t;

4) calculated data matrix X _NewResidual error E and the residual error F of Y, be back to the 2nd) step, extract second couple of latent variable u and t once more;

5) repeating step 3) and step 4), the increment of prediction residual quadratic sum is convergence less than convergency value, obtains Mooney viscosity on-line prediction model.

So far, Mooney viscosity on-line prediction model is set up and is finished.

Wherein, with PRESS _(k)The prediction residual quadratic sum of representing n the sample in the k time extraction back, promptly

${\mathrm{PRESS}}_{\left(k\right)}=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(y_i-{\hat{y}}_i)}^2$

If PRESS _(k)-PRESS _(k-1)＜ε, wherein, convergency value ε sets according to the needs in the practical application, and for example: convergency value is slightly larger than 0 positive number for generally getting, as 10 ^-6

102: with Mooney viscosity value y _NewFeed back to quality inspection system, quality inspection system is judged Mooney viscosity value y _NewWhether in Mooney viscosity preset range value, if, continue to produce, upgrade Mooney viscosity on-line prediction model; If not, give the alarm, check production procedure and proofread and correct by operating personnel, and give up off-limits Mooney viscosity value y _New

Wherein,, producing timing, needing adjustment rubber to extrude parameter usually, calendering parameter, sizing material proportioning etc. according to actual conditions.

Wherein, Mooney viscosity preset range value is set according to the needs in the practical application, and when specifically realizing, the embodiment of the invention does not limit this.

103: repeated execution of steps 101-102, when no longer including the input of finished composition rheological parameter, flow process finishes.

In sum; The embodiment of the invention provides a kind of on-line automatic measuring method of Mooney viscosity of rubber, and the embodiment of the invention dopes Mooney viscosity value according to the rheological parameter with rubber, significantly reduces the hysteresis quality of measurement; Realize the online detection of Mooney viscosity; So that control the quality of elastomeric compound in real time, for the quality that further guarantees rubber lays the first stone, and the production of high-quality rubber has also brought higher economic interests for manufacturer; Reduce the production cost of manufacturer: saved a large amount of expenses of buying and safeguarding the Mooney appearance; Avoid the required cost that a large amount of manpower and materials consumed of traditional measurement Mooney appearance method, can reduce the cost input of manufacturer greatly, improved factory's interests; This method has been considered the non-linear relation between the variable fully, makes the Mooney viscosity on-line prediction model of setting up more can reflect the relation between the parameter, and prediction data more accurately and reliably.Simultaneously because the timely replacement of Mooney viscosity on-line prediction model, so it can reflect in real time and follow the tracks of production status, make Mooney viscosity on-line prediction model more can embody and have the production characteristic now.The proposition of this method is a successful Application of advanced control strategy, for huge contribution has been made in the development of the Based Intelligent Control of rubber production; Improve production automation level; More the development of enterprise provides huge help, practices thrift great amount of cost, creates more profit.

It will be appreciated by those skilled in the art that accompanying drawing is the sketch map of a preferred embodiment, the invention described above embodiment sequence number is not represented the quality of embodiment just to description.

The above is merely preferred embodiment of the present invention, and is in order to restriction the present invention, not all within spirit of the present invention and principle, any modification of being done, is equal to replacement, improvement etc., all should be included within protection scope of the present invention.

Claims (2)

1. the on-line automatic measuring method of a Mooney viscosity of rubber is characterized in that, said method comprising the steps of:

(1) the finished composition rheological parameter input Mooney viscosity on-line prediction model that quality inspection system is obtained is exported Mooney viscosity value y automatically _New

(2) with said Mooney viscosity value y _NewFeed back to said quality inspection system, said quality inspection system is judged said Mooney viscosity value y _NewWhether in Mooney viscosity preset range value, if, continue to produce, upgrade said Mooney viscosity on-line prediction model; If not, give the alarm, check production procedure and proofread and correct by operating personnel, give up off-limits Mooney viscosity value y _New

(3) repeated execution of steps (1)-step (2), when no longer including the input of said finished composition rheological parameter, flow process finishes.

2. the on-line automatic measuring method of a kind of Mooney viscosity of rubber according to claim 1 is characterized in that, said Mooney viscosity on-line prediction model is specially:

1) at first gathers said finished composition rheological parameter, set up original sample collection X _Old, to said original sample collection X _OldCarry out the multinomial transformation and obtain X _Extra, form X _New=[X _OldX _Extra];

2), extract X through the offset minimum binary algorithm _NewLatent variable u and the latent variable t of Y;

w＝X ^T×Y/(Y ^T×Y)；

w＝w/sqrt(w ^T×w)；

t＝X×w；

q＝Y ^T×t/(t ^T×t)；

u＝Y×q/(q ^T×q)；

Wherein, Y representes the Mooney viscosity value of normal range (NR), and w representes the weight of X, and q is the load vector of Y;

3) utilize Gaussian process to set up the regression relation of said latent variable u and said latent variable t;

4) calculated data matrix X _NewResidual error E and the residual error F of Y, be back to the 2nd) step, extract second couple of latent variable u and t once more;

5) repeating step 3) and step 4), less than convergency value, obtain said Mooney viscosity on-line prediction model until the increment of prediction residual quadratic sum.

Images