CN102937036A - Method and device for monitoring gas on basis of BP (back propagation) neural network - Google Patents

Abstract

The invention provides a method and a device for monitoring gas on the basis of a BP (back propagation) neural network. the method includes building an initial model on the basis of the BP neural network; setting parameters of the initial model; computing according to the initial model after the parameters of the initial model are set; determining connection weights in the model; and determining relations among the BP neural network model, input and output according to limit errors used as references. The initial model comprises an input layer, a hidden layer and an output layer. According to the scheme, the method and the device have the advantages that messages of similar sensors at different positions in an underground monitoring substation are fused, so that measurement errors are reduced to a great extent, and the accuracy of data is improved.

Description

A kind of gas monitor method and device based on the BP neutral net

Technical field

The present invention relates to sensor information and process and the safety of coal mines technical field, particularly relate to a kind of gas monitor method and device based on the BP neutral net.

Background technology

Safety of Coal Mine Production is a problem of paying much attention at present, increase along with the pit mining degree of depth, the complexity that varied, the gas of down-hole geological condition detects and the uncertainty of Gas Outburst are more remarkable, make Safety of Coal Mine Production have great potential safety hazard.Mine gas monitoring mainly is that the parameters such as fire damp and CO concentration, dust content, temperature, wind speed, negative pressure are detected (gas monitoring system of prior art as shown in Figure 1) in real time, needs to adopt a large amount of dissimilar sensors.But because the sensor measurement precision is limit and the interference of environmental factor, can make survey data and physical presence deviation, more seriously, the erosion of pernicious gas can make operative sensor forfeiture detectability in the complex environment of down-hole.Traditional a lot of solutions are that single-sensor is repeatedly measured or a plurality of sensor measurement data are averaged, although can improve to a certain extent like this reliability of TT﹠C system, but can't satisfy the requirement of system real time, inaccurate survey data still can affect final judged result.

Therefore, generally speaking, in present coal mine gas monitoring scheme, mostly also there be manual detection and the tour mode that falls behind, detect inaccurate, not in place, untimely phenomenon and problem.And some colliery utilizes the instrument of automation to carry out gas monitor, but normally comparatively simple single-sensor detection means exists detection unstable, insecure shortcoming.Although it is to use several sensors to carry out combine detection that some detecting instruments are also arranged, normally simple data reasoning is omitted potential safety hazard, does not take full advantage of data and carries out reasoning on the profound level.

In order to overcome above-mentioned the deficiencies in the prior art, the invention provides a kind of gas monitor method and device based on the BP neutral net, attempt proposing to utilize the BP neutral net downhole sensor information to be carried out the method for designing of fusion treatment, enable intelligence perfect, reflect that environmental characteristic has improved recognition speed and the reliability to target exactly, be one of moment sensor information processing and safety of coal mines technical field problem demanding prompt solution.

Summary of the invention

In view of this, the embodiment of the invention has proposed a kind of gas monitor method and device based on the BP neutral net, comprise input layer by foundation, the initial model based on the BP neutral net of hidden layer and output layer, then initial model is carried out the parameter setting, then the initial model after arranging according to parameter calculates, determine the connection weights in the model, be benchmark according to limit error finally, determine the relation of BP neural network model and input and outlet chamber, this programme has merged the information of the same type of sensor of diverse location in the underground monitoring substation, reduce to a great extent measure error, improved the accuracy of data.

For solving the problems of the technologies described above, the purpose of the embodiment of the invention is achieved through the following technical solutions:

A kind of gas monitor method based on the BP neutral net comprises:

Step 1, foundation comprise the initial model based on the BP neutral net of input layer, hidden layer and output layer;

Step 2, initial model is carried out the parameter setting;

Step 3, the initial model after arranging according to parameter calculate, and determine the connection weights in the model;

Step 4, be benchmark according to limit error, determine the relation of BP neural network model and input and outlet chamber.

Preferably, in the above-mentioned steps one, the nodes of input layer is 6.

Preferably, the node of above-mentioned input layer comprises gas, carbon monoxide, temperature, wind speed, dust and negative pressure data.

Preferably, in the above-mentioned steps one, safe condition information under the described output layer output well is divided into five grades, comprising: safe, safer, Generally Recognized as safe, dangerous, dangerous, so the output layer nodes is chosen as 5.

Preferably, in the above-mentioned steps one, described the number of hidden nodes is taken as 10.

Preferably, in the above-mentioned steps two, initial model is carried out the parameter setting comprise that initial weight gets the random number between (1,1), the η learning rate gets 0.1, and anticipation error gets 0.001, and maximum step number is made as 1000.

Preferably, in the above-mentioned steps two, further comprise hidden layer and output layer activation primitive and choose continuously differentiable unipolarity sigmoid function, namely

w _KiBe the be connected weights of hidden node k with input layer i; v _JkBe the be connected weights of output layer node j with hidden node k.

Preferably, in the above-mentioned steps four, further comprise, the difference between calculating output and the sample object output is fed back to neutral net, each connection weights of network constantly are adjusted to till the relation that can simulate sample input and output in limit error.

A kind of gas monitor device based on the BP neutral net, comprise initialization unit, parameter set unit, computing unit and performance element, the initial model based on the BP neutral net that comprises input layer, hidden layer and output layer by foundation, then initial model is carried out the parameter setting, then the initial model after arranging according to parameter calculates, determining the connection weights in the model, is benchmark according to limit error finally, determines the relation of BP neural network model and input and outlet chamber.

Preferably, above-mentioned initialization unit is used for setting up the initial model based on the BP neutral net that comprises input layer, hidden layer and output layer.

Preferably, the above-mentioned parameter setting unit is used for initial model is carried out the parameter setting.

Preferably, above-mentioned computing unit calculates for the initial model after arranging according to parameter, determines the connection weights in the model.

Preferably, it is benchmark that above-mentioned performance element is used for according to limit error, determines the relation of BP neural network model and input and outlet chamber.

In sum, the invention provides a kind of gas monitor method and device based on the BP neutral net, comprise input layer by foundation, the initial model based on the BP neutral net of hidden layer and output layer, then initial model is carried out the parameter setting, then the initial model after arranging according to parameter calculates, determine the connection weights in the model, be benchmark according to limit error finally, determine the relation of BP neural network model and input and outlet chamber, this programme has merged the information of the same type of sensor of diverse location in the underground monitoring substation, reduce to a great extent measure error, improved the accuracy of data.

Description of drawings

Fig. 1 is the gas monitoring system schematic diagram of prior art;

Fig. 2 is a kind of gas monitor method schematic diagram based on the BP neutral net;

Fig. 3 is the network model schematic diagram of the embodiment of the invention;

Fig. 4 is the training error curve synoptic diagram of the embodiment of the invention;

Fig. 5 is the network integration result schematic diagram of the embodiment of the invention;

Fig. 6 is a kind of gas monitor device schematic diagram based on the BP neutral net.

The specific embodiment

A kind of gas monitor method and device based on the BP neutral net that the embodiment of the invention provides, comprise input layer by foundation, the initial model based on the BP neutral net of hidden layer and output layer, then initial model is carried out the parameter setting, then the initial model after arranging according to parameter calculates, determine the connection weights in the model, be benchmark according to limit error finally, determine the relation of BP neural network model and input and outlet chamber, this programme has merged the information of the same type of sensor of diverse location in the underground monitoring substation, reduce to a great extent measure error, improved the accuracy of data.

Main thought of the present invention is: utilize the BP neutral net downhole sensor information to be carried out the method for designing of fusion treatment, process from the visual angle of many information, comprehensively, obtain inner link and the rule of various information, reject useless and wrong composition, keep correct and useful composition, the intelligent optimization of final realization information, enable intelligence perfect, reflect that environmental characteristic has improved recognition speed and the reliability to target exactly, have redundancy, complementarity and real-time through the sensor information after merging.

For making purpose of the present invention, technical scheme and advantage clearer, the embodiment that develops simultaneously with reference to the accompanying drawings, the present invention is described in more detail.

The embodiment of the invention provides a kind of gas monitor method based on the BP neutral net, and as shown in Figure 2, concrete steps comprise:

Step 1, foundation comprise the initial model based on the BP neutral net of input layer, hidden layer and output layer;

Particularly, in embodiments of the present invention, be to adopt the BP neutral net to the distribution information such as gas are carried out Intelligent treatment, finally draw subsurface environment information and make man-rate, Comprehensive the subsurface environment parameter, real-time, objective, provide the evaluation of the safe status exactly.

Therefore, in this programme, adopt three layers of BP neutral net, particularly, comprise input layer, hidden layer and output layer.Input layer receives gas, the CO(carbon monoxide that is obtained by the one-level data fusion), temperature, wind speed, dust and negative pressure data, so the nodes that input layer connects is 6.Safe condition information under the output layer output well is divided into five grades, comprising: safe, safer, Generally Recognized as safe, dangerous, dangerous, so the output layer nodes is chosen as 5.In addition, the number of hidden nodes of this programme is taken as 10.Network model as shown in Figure 3.

Step 2, initial model is carried out the parameter setting;

Particularly, in embodiments of the present invention, the BP neural network algorithm is a kind of feed-forward type algorithm, therefore in the initial model based on the BP neutral net, we carry out as giving a definition and the parameter setting: initial weight is got the random number between (1,1), and the η learning rate gets 0.1, anticipation error gets 0.001, and maximum step number is made as 1000.Choose the actual measurement environmental data sample of 5 mines and after one-level merges, train as fan-in network, and:

Hidden layer and output layer activation primitive are chosen continuously differentiable unipolarity sigmoid function, namely $f\left(x\right)=\frac{1}{1+e^{-x}}$

Therefore, hidden layer k(k=1,2 ..., 10) and total input and output of individual node are

${\mathrm{net}}_k=\underset{i=1}{\overset{6}{\mathrm{\Σ}}}{w}_{\mathrm{ki}}{X}_i---\left(1\right)$

$O_k=f\left({\mathrm{net}}_k\right)=\frac{1}{1+e^{{-\mathrm{net}}_k}}---\left(2\right)$

In the formula, w _KiBe the be connected weights of hidden node k with input layer i.

Output layer j(j=1,2 ..., 5) and total input and output of individual node are:

${\mathrm{net}}_j=\underset{k=1}{\overset{10}{\mathrm{\Σ}}}{v}_{\mathrm{jk}}{O}_k---\left(3\right)$

$Y_j=f\left({\mathrm{net}}_j\right)=\frac{1}{1+e^{{-\mathrm{net}}_j}}---\left(4\right)$

In the formula, v _JkBe the be connected weights of output layer node j with hidden node k.

Output error is $E=\frac{1}{2}\underset{j=1}{\overset{6}{\mathrm{\Σ}}}{(d_j-Y_j)}^2$

Begin network training after giving initial weight and input vector, namely by repeatedly calculating, ask for error, adjust weights, the weights adjustment amount should be directly proportional with the Gradient Descent of error, namely

${\mathrm{\Δw}}_{\mathrm{ki}}=-\mathrm{\η}\frac{\∂E}{{\∂w}_{\mathrm{ki}}}=-\mathrm{\η}\frac{\∂E}{{\∂\mathrm{net}}_k}\frac{{\∂\mathrm{net}}_k}{{\∂w}_{\mathrm{ki}}}---\left(5\right)$

${\mathrm{\Δv}}_{\mathrm{jk}}=-\mathrm{\η}\frac{\∂E}{{\∂v}_{\mathrm{jk}}}=-\mathrm{\η}\frac{\∂E}{{\∂\mathrm{net}}_j}\frac{{\∂\mathrm{net}}_j}{{\∂v}_{\mathrm{jk}}}---\left(6\right)$

In the formula: η is learning rate.

Step 3, the initial model after arranging according to parameter calculate, and determine the connection weights in the model;

Particularly, in embodiments of the present invention, the BP neural network algorithm is a kind of feed-forward type algorithm, the difference that it will calculate between output and the sample object output feeds back to neutral net, and each connection weights of network constantly are adjusted to till the relation that can simulate sample input and output in limit error.Algorithm comprises 3 parts such as forward direction computing, reverse computing and weights adjustment.So, in this programme, will calculate according to the initial model after parameter arranges in the above-mentioned steps two, determine the connection weights in the model.In step 2, in the initial model based on the BP neutral net, we carry out as giving a definition and the parameter setting: initial weight is got the random number between (1,1), and the η learning rate is got, and anticipation error gets 0.001, and maximum step number is made as 1000.Choosing the actual measurement environmental data sample of 5 mines trains as fan-in network after one-level merges.

In this programme the connection weights comprise w _KiBe the be connected weights of hidden node k with input layer i; And v _JkBe the be connected weights of output layer node j with hidden node k.By network training error is constantly reduced, until reach desired value.The training error curve as shown in Figure 4, network integration result is as shown in Figure 5.

Step 4, be benchmark according to limit error, determine the relation of BP neural network model and input and outlet chamber.

Particularly, in embodiments of the present invention, the BP neural network algorithm is a kind of feed-forward type algorithm, the difference that it will calculate between output and the sample object output feeds back to neutral net, and each connection weights of network constantly are adjusted to till the relation that can simulate sample input and output in limit error.In this programme, the training error curve as shown in Figure 4, network integration result is as shown in Figure 5.As seen, the output data can reflect the subsurface environment safe condition exactly.

Further, in this programme, be to adopt three layers of BP neutral net, particularly, comprise input layer, hidden layer and output layer.Input layer receives gas, CO, temperature, wind speed, dust and the negative pressure data that obtained by the one-level data fusion, so the nodes that input layer connects is 6.Safe condition information under the output layer output well is divided into five grades, comprising: safe, safer, Generally Recognized as safe, dangerous, dangerous, so the output layer nodes is chosen as 5.In addition, the number of hidden nodes of this programme is taken as 10.Network model as shown in Figure 3.

Therefore, in embodiments of the present invention, for the deficiencies in the prior art, in coal-mine gas monitoring system comprising, utilize Intelligent Multi layer BP neutral net downhole sensor information to be carried out the method for designing of information fusion, process from the visual angle of many information, comprehensively, obtain inner link and the rule of various information, reject useless and wrong composition, keep correct and useful composition, the optimization of final realization information enables perfect, reflect that exactly environmental characteristic has improved recognition speed and the reliability to target, have redundancy through the sensor information after merging, reliability and real-time.

In addition, in this programme, adopt the method for Intelligent Multi layer BP neutral net that downhole sensor data is carried out information fusion, merged the information of the same type of sensor of diverse location in the underground monitoring substation, reduce to a great extent measure error, improved the accuracy of data.This method Comprehensive the subsurface environment parameter, real-time, objective, provide the evaluation of the safe status exactly.The simulation experiment result shows that this method is practicable, enable perfect, reflect Gas Distribution situation under the coal mine exactly, improve speed and reliability to gas monitor, had redundancy, complementarity and real-time through the sensor information after merging.Coal mine gas monitoring and safety in production had very high use value.

In addition, the embodiment of the invention also provides a kind of gas monitor device based on the BP neutral net.As shown in Figure 2, a kind of gas monitor device schematic diagram based on the BP neutral net that provides for the embodiment of the invention.

A kind of gas monitor device based on the BP neutral net comprises initialization unit 11, parameter set unit 22, computing unit 33 and performance element 44.

Initialization unit 11 is used for setting up the initial model based on the BP neutral net that comprises input layer, hidden layer and output layer;

Particularly, in embodiments of the present invention, be to adopt the BP neutral net to the distribution information such as gas are carried out Intelligent treatment, finally draw subsurface environment information and make man-rate, Comprehensive the subsurface environment parameter, real-time, objective, provide the evaluation of the safe status exactly.

Therefore, in this programme, adopt three layers of BP neutral net, particularly, comprise input layer, hidden layer and output layer.Input layer receives gas, CO, temperature, wind speed, dust and the negative pressure data that obtained by the one-level data fusion, so the nodes that input layer connects is 6.Safe condition information under the output layer output well is divided into five grades, comprising: safe, safer, Generally Recognized as safe, dangerous, dangerous, so the output layer nodes is chosen as 5.In addition, the number of hidden nodes of this programme is taken as 10.Network model as shown in Figure 3.

Parameter set unit 22 is used for initial model is carried out the parameter setting;

Particularly, in embodiments of the present invention, the BP neural network algorithm is a kind of feed-forward type algorithm, therefore in the initial model based on the BP neutral net, we carry out as giving a definition and the parameter setting: initial weight is got the random number between (1,1), gets 0.1, anticipation error gets 0.001, and maximum step number is made as 1000.Choose the actual measurement environmental data sample of 5 mines and after one-level merges, train as fan-in network, and:

Hidden layer and output layer activation primitive are chosen continuously differentiable unipolarity sigmoid function, namely $f\left(x\right)=\frac{1}{1+e^{-x}}$

Therefore, hidden layer k(k=1,2 ..., 10) and total input and output of individual node are

${\mathrm{net}}_k=\underset{i=1}{\overset{6}{\mathrm{\Σ}}}{w}_{\mathrm{ki}}{X}_i---\left(1\right)$

$O_k=f\left({\mathrm{net}}_k\right)=\frac{1}{1+e^{{-\mathrm{net}}_k}}---\left(2\right)$

In the formula, w _KiBe the be connected weights of hidden node k with input layer i.

Output layer j(j=1,2 ..., 5) and total input and output of individual node are:

${\mathrm{net}}_j=\underset{k=1}{\overset{10}{\mathrm{\Σ}}}{v}_{\mathrm{jk}}{O}_k---\left(3\right)$

$Y_j=f\left({\mathrm{net}}_j\right)=\frac{1}{1+e^{{-\mathrm{net}}_j}}---\left(4\right)$

In the formula, v _JkBe the be connected weights of output layer node j with hidden node k.

Output error is $E=\frac{1}{2}\underset{j=1}{\overset{6}{\mathrm{\Σ}}}{(d_j-Y_j)}^2$

Begin network training after giving initial weight and input vector, namely by repeatedly calculating, ask for error, adjust weights, the weights adjustment amount should be directly proportional with the Gradient Descent of error, namely

${\mathrm{\Δw}}_{\mathrm{ki}}=-\mathrm{\η}\frac{\∂E}{{\∂w}_{\mathrm{ki}}}=-\mathrm{\η}\frac{\∂E}{{\∂\mathrm{net}}_k}\frac{{\∂\mathrm{net}}_k}{{\∂w}_{\mathrm{ki}}}---\left(5\right)$

${\mathrm{\Δv}}_{\mathrm{jk}}=-\mathrm{\η}\frac{\∂E}{{\∂v}_{\mathrm{jk}}}=-\mathrm{\η}\frac{\∂E}{{\∂\mathrm{net}}_j}\frac{{\∂\mathrm{net}}_j}{{\∂v}_{\mathrm{jk}}}---\left(6\right)$

In the formula: η is learning rate.

Computing unit 33 calculates for the initial model after arranging according to parameter, determines the connection weights in the model;

Particularly, in embodiments of the present invention, the BP neural network algorithm is a kind of feed-forward type algorithm, the difference that it will calculate between output and the sample object output feeds back to neutral net, and each connection weights of network constantly are adjusted to till the relation that can simulate sample input and output in limit error.Algorithm comprises 3 parts such as forward direction computing, reverse computing and weights adjustment.So, in this programme, will calculate according to the initial model after parameter arranges in the above-mentioned steps two, determine the connection weights in the model.In step 2, in the initial model based on the BP neutral net, we carry out as giving a definition and the parameter setting: initial weight is got the random number between (1,1), gets 0.1, and anticipation error gets 0.001, and maximum step number is made as 1000.Choosing the actual measurement environmental data sample of 5 mines trains as fan-in network after one-level merges.

In this programme the connection weights comprise w _KiBe the be connected weights of hidden node k with input layer i; And v _JkBe the be connected weights of output layer node j with hidden node k.By network training error is constantly reduced, until reach desired value.The training error curve as shown in Figure 4, network integration result is as shown in Figure 5.

Performance element 44, being used for according to limit error is benchmark, determines the relation of BP neural network model and input and outlet chamber.

Particularly, in embodiments of the present invention, the BP neural network algorithm is a kind of feed-forward type algorithm, the difference that it will calculate between output and the sample object output feeds back to neutral net, and each connection weights of network constantly are adjusted to till the relation that can simulate sample input and output in limit error.In this programme, the training error curve as shown in Figure 4, network integration result is as shown in Figure 5.As seen, the output data can reflect the subsurface environment safe condition exactly.

Further, in this programme, be to adopt three layers of BP neutral net, particularly, comprise input layer, hidden layer and output layer.Input layer receives gas, CO, temperature, wind speed, dust and the negative pressure data that obtained by the one-level data fusion, so the nodes that input layer connects is 6.Safe condition information under the output layer output well is divided into five grades, comprising: safe, safer, Generally Recognized as safe, dangerous, dangerous, so the output layer nodes is chosen as 5.In addition, the number of hidden nodes of this programme is taken as 10.Network model as shown in Figure 3.

One of ordinary skill in the art will appreciate that and realize that all or part of step that above-described embodiment method is carried is to come the relevant hardware of instruction to finish by program, described program can be stored in a kind of computer-readable recording medium, this program comprises step of embodiment of the method one or a combination set of when carrying out.

In addition, each functional unit in each embodiment of the present invention can be integrated in the processing module, also can be that the independent physics of unit exists, and also can be integrated in the module two or more unit.Above-mentioned integrated module both can adopt the form of hardware to realize, also can adopt the form of software function module to realize.If described integrated module realizes with the form of software function module and during as independently production marketing or use, also can be stored in the computer read/write memory medium.

In sum, this paper provides a kind of gas monitor method and device based on the BP neutral net, comprise input layer by foundation, the initial model based on the BP neutral net of hidden layer and output layer, then initial model is carried out the parameter setting, then the initial model after arranging according to parameter calculates, determine the connection weights in the model, be benchmark according to limit error finally, determine the relation of BP neural network model and input and outlet chamber, this programme has merged the information of the same type of sensor of diverse location in the underground monitoring substation, reduce to a great extent measure error, improved the accuracy of data.

More than a kind of gas monitor method and device based on the BP neutral net provided by the present invention is described in detail, used specific case herein principle of the present invention and embodiment are set forth, the explanation of above embodiment just is used for helping to understand the solution of the present invention; Simultaneously, for one of ordinary skill in the art, according to thought of the present invention, all will change in specific embodiments and applications, in sum, this description should not be construed as limitation of the present invention.

Claims (13)

1. gas monitor method based on the BP neutral net is characterized in that described method comprises:

Step 1, foundation comprise the initial model based on the BP neutral net of input layer, hidden layer and output layer;

Step 2, initial model is carried out the parameter setting;

Step 3, the initial model after arranging according to parameter calculate, and determine the connection weights in the model;

Step 4, be benchmark according to limit error, determine the relation of BP neural network model and input and outlet chamber.

2. method according to claim 1 is characterized in that, in the described step 1, the nodes of input layer is 6.

3. method according to claim 2 is characterized in that, the node of described input layer comprises gas, carbon monoxide, temperature, wind speed, dust and negative pressure data.

4. method according to claim 1 is characterized in that, in the described step 1, safe condition information under the described output layer output well, be divided into five grades, comprise: safe, safer, Generally Recognized as safe, dangerous, dangerous, so the output layer nodes is chosen as 5.

5. method according to claim 1 is characterized in that, in the described step 1, described the number of hidden nodes is taken as 10.

6. method according to claim 1 is characterized in that, in the described step 2, initial model is carried out the parameter setting comprise that initial weight gets the random number between (1,1), and the η learning rate gets 0.1, and anticipation error gets 0.001, and maximum step number is made as 1000.

7. method according to claim 1 is characterized in that, in the described step 2, further comprises hidden layer and output layer activation primitive and chooses continuously differentiable unipolarity sigmoid function, namely

w _KiBe the be connected weights of hidden node k with input layer i; v _JkBe the be connected weights of output layer node j with hidden node k.

8. method according to claim 1, it is characterized in that, in the described step 4, further comprise, difference between calculating output and the sample object output is fed back to neutral net, each connection weights of network constantly are adjusted to till the relation that can simulate sample input and output in limit error.

9. gas monitor device based on the BP neutral net, it is characterized in that, described device comprises initialization unit, parameter set unit, computing unit and performance element, the initial model based on the BP neutral net that comprises input layer, hidden layer and output layer by foundation, then initial model is carried out the parameter setting, then the initial model after arranging according to parameter calculates, determine the connection weights in the model, be benchmark according to limit error finally, determine the relation of BP neural network model and input and outlet chamber.

10. device according to claim 9 is characterized in that, described initialization unit is used for setting up the initial model based on the BP neutral net that comprises input layer, hidden layer and output layer.

11. device according to claim 9 is characterized in that, described parameter set unit is used for initial model is carried out the parameter setting.

12. device according to claim 9 is characterized in that, described computing unit calculates for the initial model after arranging according to parameter, determines the connection weights in the model.

13. device according to claim 9 is characterized in that, it is benchmark that described performance element is used for according to limit error, determines the relation of BP neural network model and input and outlet chamber.

Images