CN116817306A - Gas equipment capable of monitoring and controlling gas flow - Google Patents
Gas equipment capable of monitoring and controlling gas flow Download PDFInfo
- Publication number
- CN116817306A CN116817306A CN202311090625.7A CN202311090625A CN116817306A CN 116817306 A CN116817306 A CN 116817306A CN 202311090625 A CN202311090625 A CN 202311090625A CN 116817306 A CN116817306 A CN 116817306A
- Authority
- CN
- China
- Prior art keywords
- gas
- gas flow
- monitoring
- neural network
- controller module
- 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.)
- Granted
Links
- 238000012544 monitoring process Methods 0.000 title claims abstract description 89
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 52
- 230000000007 visual effect Effects 0.000 claims abstract description 25
- 238000000605 extraction Methods 0.000 claims abstract description 22
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 17
- 238000012423 maintenance Methods 0.000 claims abstract description 13
- 230000010354 integration Effects 0.000 claims abstract description 12
- 230000008439 repair process Effects 0.000 claims abstract description 7
- 230000008859 change Effects 0.000 claims description 7
- 239000000446 fuel Substances 0.000 claims description 4
- 230000007246 mechanism Effects 0.000 claims description 3
- 238000012546 transfer Methods 0.000 claims description 2
- 210000005036 nerve Anatomy 0.000 abstract 1
- 230000006870 function Effects 0.000 description 46
- 238000000034 method Methods 0.000 description 36
- 238000011176 pooling Methods 0.000 description 21
- 238000013528 artificial neural network Methods 0.000 description 19
- 230000004913 activation Effects 0.000 description 18
- 230000008569 process Effects 0.000 description 18
- 210000002569 neuron Anatomy 0.000 description 17
- 239000013598 vector Substances 0.000 description 13
- 238000003745 diagnosis Methods 0.000 description 9
- 238000012800 visualization Methods 0.000 description 8
- 238000012549 training Methods 0.000 description 7
- 238000004364 calculation method Methods 0.000 description 6
- 238000005070 sampling Methods 0.000 description 6
- 238000009826 distribution Methods 0.000 description 5
- 230000008901 benefit Effects 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 230000009466 transformation Effects 0.000 description 4
- 230000005540 biological transmission Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 2
- 239000003990 capacitor Substances 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000013016 damping Methods 0.000 description 2
- 238000007405 data analysis Methods 0.000 description 2
- 230000001419 dependent effect Effects 0.000 description 2
- 238000009795 derivation Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000002360 preparation method Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000013139 quantization Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 238000012935 Averaging Methods 0.000 description 1
- 238000007476 Maximum Likelihood Methods 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000003213 activating effect Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000003321 amplification Effects 0.000 description 1
- 230000002238 attenuated effect Effects 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000012886 linear function Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000007620 mathematical function Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000003199 nucleic acid amplification method Methods 0.000 description 1
- 230000001902 propagating effect Effects 0.000 description 1
- 238000009877 rendering Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000000630 rising effect Effects 0.000 description 1
- 229920006395 saturated elastomer Polymers 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
Classifications
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F23—COMBUSTION APPARATUS; COMBUSTION PROCESSES
- F23N—REGULATING OR CONTROLLING COMBUSTION
- F23N1/00—Regulating fuel supply
- F23N1/005—Regulating fuel supply using electrical or electromechanical means
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F23—COMBUSTION APPARATUS; COMBUSTION PROCESSES
- F23N—REGULATING OR CONTROLLING COMBUSTION
- F23N5/00—Systems for controlling combustion
- F23N5/24—Preventing development of abnormal or undesired conditions, i.e. safety arrangements
- F23N5/245—Preventing development of abnormal or undesired conditions, i.e. safety arrangements using electrical or electromechanical means
Landscapes
- Engineering & Computer Science (AREA)
- Chemical & Material Sciences (AREA)
- Combustion & Propulsion (AREA)
- Mechanical Engineering (AREA)
- General Engineering & Computer Science (AREA)
- Feedback Control In General (AREA)
Abstract
The invention discloses a gas device capable of monitoring and controlling gas flow, which comprises a flow sensor, a gas flow controller module, a gas electromagnetic valve, a manual key module, a pressure transmitter, a microcomputer additional A/D converter, a convolutional neural network monitoring system and a display module which are sequentially connected through matched pipelines; the gas electromagnetic valve and the gas flow controller module are connected with a pressure transmitter, the pressure transmitter is connected with the gas electromagnetic valve and the gas flow controller module to convert pressure into an electric signal, and the gas electromagnetic valve and the gas flow controller module are connected with the convolutional neural network monitoring integration platform; the monitoring integrated platform comprises a microcomputer, an additional A/D converter and a convolutional nerve monitoring system, the gas quantity acquisition value is monitored through a feature extraction unit and a classification layer unit in a platform algorithm, and finally, the monitoring data quantity is displayed through a display module to form a visual chart, so that gas pipeline maintenance and operation staff can conveniently check and repair the numerical range of the gas flow controller module according to the monitoring result.
Description
Technical Field
The invention relates to the technical field of gas monitoring and control, and provides a gas device capable of monitoring and controlling gas flow.
Background
In the prior art, a high-pressure gas valve or a low-pressure gas valve is manually controlled to limit the pressure of the gas inlet to control the gas flow, and the method has the defects of complex operation, low instantaneity, poor flow control precision and the like, and is difficult to quickly obtain accurate control of the gas flow to achieve proper air-fuel ratio.
For the traditional gas monitoring system, a relation function between two node voltages is used as a fault characteristic, and the method is generally applicable to an improved fault method for diagnosing faults of an analog circuit; because the integration level and complexity of modern analog circuits are continuously increased, most fault characteristics cannot be recorded one by one, intelligent fault monitoring mainly depends on artificial intelligence technology, and fault diagnosis monitoring of the analog circuits is further derived.
Disclosure of Invention
The present invention is directed to a gas device capable of monitoring and controlling the gas flow rate, so as to solve the above-mentioned problems in the prior art.
In order to achieve the above purpose, the present invention provides the following technical solutions: the gas equipment capable of monitoring and controlling the gas flow comprises a flow sensor, a gas flow controller module, a gas electromagnetic valve, a manual key module, a pressure transmitter, a microcomputer additional A/D converter, a convolutional neural network monitoring integration platform and a display module which are sequentially connected through matched pipelines; the gas solenoid valve and the gas flow controller module are connected with the pressure transmitter, the pressure is converted into an electric signal, the electric signal is connected with the microcomputer additional A/D converter and the convolutional neural network monitoring integrated platform through the convolutional neural network, the characteristic extraction unit and the classification layer unit of the convolutional neural network monitoring integrated platform monitor the gas quantity, finally the display module displays the monitored data quantity and forms a visual chart, the gas pipeline maintenance and operation staff can conveniently repair the numerical range of the gas flow controller module manually through the manual key module according to the monitoring result.
Further, a gas device capable of monitoring and controlling gas flow is provided, and the flow sensor is used for detecting the real-time air quantity and the gas quantity in the transmission pipeline, so that the air-fuel ratio can be controlled to an optimal value, and the gas flow is conveniently controlled by the gas electromagnetic valve and the gas flow controller module.
Further, the gas flow controller module is connected with the flow sensor and the pressure gauge and controls the gas electromagnetic valve to adjust the output gas flow; the gas flow controller module is mutually connected and matched with the flow sensor, and meanwhile, the PD principle is added to the gas flow controller module to improve the control mechanism.
Further, the gas flow controller module adopts PD control combination, P is proportional control, D is differential control according to pipeline gas conditions acquired by the flow sensor, and the gas flow controller module is used for controlling the gas electromagnetic valve according to the real-time flow value of the gas so as to adjust the output gas flow.
Further, the pressure transmitter is arranged on the matched pipeline and is used for receiving and converting the control result generated after the pressure gauge is connected with the gas flow controller module.
Further, the pressure transmitter converts the measured pressure into an electric signal in the range of the measured pressure, and the electric signal is amplified by the pressure transmitter due to the small change of the electric signal converted from the measured pressure, so that the microcomputer is convenient for the additional A/D converter to receive the signal.
Further, the microcomputer is attached with an A/D converter to convert the electric signal transmitted by the pressure transmitter into a digital signal recognizable by the computer network.
Further, the convolutional neural network monitoring integrated platform comprises a microcomputer additional A/D converter and a convolutional neural network monitoring system; the integrated platform is connected through the convolutional neural network monitoring, the microcomputer is additionally provided with an A/D converter for acquiring an electric signal of the pressure transmitter, converting the electric signal into a digital signal quantity which can be identified by the convolutional neural network monitoring integrated platform, and monitoring the gas quantity; meanwhile, the convolutional neural network monitoring integration platform integrates a matched algorithm idea, monitors and diagnoses the input digital signal quantity through a feature extraction layer and a classification layer in an algorithm model, and transmits a monitoring result to a display module.
The convolution layer module is used for receiving the electric signal information, selecting and adjusting the electric signal information according to actual conditions through the pooling layer module, preventing the phenomenon of over fitting or under fitting, and effectively reducing the size of the parameter matrix, thereby reducing the number of parameters in the final connection layer.
Further, the display module displays the monitored data amount and forms a visual chart.
Further, the manual key module can facilitate operation and maintenance personnel to manually repair the numerical range of the gas flow controller module according to the monitoring result.
Compared with the prior art, the invention has the following advantages:
when the device disclosed by the invention works, the actual gas flow in the gas pipeline is detected in real time through the gas flow controller module, if the actual gas flow value is different from the pipeline gas threshold ratio acquired by the flow sensor, the gas flow controller module controls the gas solenoid valve to adjust the gas flow in the input pipeline, the gas flow and the pressure can be controlled, the gas with continuously variable flow can be output, and the gas flow controller module has high automatic control degree through the combined control principle of the combined flow sensor and the PD.
The gas flow controller module adopts the combined control principle of the flow sensor and the PD to regulate the gas electromagnetic valve to control the gas flow, in order to ensure the real-time property of the gas flow, the control structure of the gas flow controller module is single, the proportion parameters of the valve value of the pipeline are fixed, the problems that the traditional simulation circuit fault diagnosis method is seriously dependent on feature extraction and selection and the like can be overcome after the convolutional neural network monitoring system is introduced,
the method has the advantages that the deep confidence network is utilized for fault diagnosis and monitoring, the convolutional neural network model can effectively classify long sequence data in the signals of the output circuit of the additional A/D converter of the microcomputer, the characteristic extraction layer of the convolutional neural network algorithm directly performs characteristic extraction on the acquired data, the labor cost of fault diagnosis on the characteristic extraction is reduced, and the universality of a gas flow monitoring system is effectively improved; according to the invention, a convolutional neural network monitoring integrated platform is added to gas control equipment, so that the original sampling data is directly subjected to feature extraction and fault classification monitoring; the convolutional neural network monitoring integrated platform is provided, the characteristic extraction process of an original signal is simplified, fault signals output by the microcomputer additional A/D converter are transmitted to the convolutional neural network monitoring system after being processed by the convolutional neural network monitoring integrated platform, and the processed data result is converted into a visual chart through the platform visual model, so that the gas pipeline maintenance and operation and maintenance personnel can repair the numerical range of the gas flow controller module manually according to the monitoring result.
Drawings
FIG. 1 is a schematic diagram of a gas flow monitoring and control apparatus according to the present invention.
Fig. 2 is a schematic block diagram of PD control in the flow control module of the present invention.
FIG. 3 is a schematic diagram of an algorithm model of the convolutional neural network monitoring integration platform of the present invention.
Detailed Description
The preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.
As shown in FIG. 1, the gas equipment capable of monitoring and controlling the gas flow comprises a flow sensor, a gas flow controller module, a gas electromagnetic valve, a manual key module, a pressure transmitter, a microcomputer additional A/D converter, a convolutional neural network monitoring integrated platform and a display module which are sequentially connected through matched pipelines; the gas solenoid valve and the gas flow controller module are connected with the pressure transmitter, the pressure is converted into an electric signal, the electric signal is connected with the microcomputer additional A/D converter and the convolutional neural network monitoring integrated platform through the convolutional neural network, the characteristic extraction unit and the classification layer unit of the convolutional neural network monitoring integrated platform monitor the gas quantity, finally the display module displays the monitored data quantity and forms a visual chart, the gas pipeline maintenance and operation staff can conveniently repair the numerical range of the gas flow controller module manually through the manual key module according to the monitoring result.
The flow sensor, the gas flow controller module and the gas electromagnetic valve are arranged on the back surface of the gas control mounting plate, and the manual key module and the pressure transmitter are arranged on the front surface of the gas control mounting plate; the integrated platform is monitored by linking a microcomputer with an additional A/D converter and a convolutional neural network through the convolutional neural network.
The flow sensor is used for detecting the real-time air quantity and the gas quantity in the transmission pipeline, so that the air-fuel ratio can be controlled to be at an optimal value, and the gas flow is conveniently controlled by the gas electromagnetic valve and the gas flow controller module.
The gas flow controller module is connected with the flow sensor and the pressure gauge and used for controlling the gas electromagnetic valve to adjust the output gas flow; the gas flow controller module is mutually connected and matched with the flow sensor, and meanwhile, the PD principle is added to the gas flow controller module to improve the control mechanism.
The PD principle is that Proportional is Proportional control abbreviated as P, derivative is differential control abbreviated as D, on the basis of the existing control, the Integral in PDI control is changed into a flow sensor according to the actual gas equipment control scene, the flow sensor can control the system to be in a steady state 0 type and set a constant flow value, then an Integral control link is not required to be introduced, and steady state errors including step signals and disturbance of constant input signals can be eliminated;
proportional (P) -Integral (I) -Derivative (D) Control, is actually three feedback controls: proportional control, integral control and differential control are collectively called as a proportional control, an integral control and a differential control; according to the control object and the actual application scene condition, the partial combination of the three controls, namely P control, PI control and PD control or the combination of the three controls, namely PID control in the true sense, can be generally called as PID control law;
example 1:
the scaling factor P in the PD principle adopted by the gas flow controller module is that the larger the scaling factor k is used in the linear magnification factor k passing through the coordinate point (0, 0), the larger the slope of the linear is, so that the scaling factor k is used in y=k×x, where k is the scaling factor P, and is called Kp for short, so that y=kp×x is changed, x is the difference between the current value currentValue and the target value totalValue, and error err for short, y is the corresponding output value U of the gas flow controller module, so that the corresponding output value u=kp (currentValue-totalValue) of the gas flow controller module is obtained;
the output value corresponding to the gas flow controller module at the current 1 st adjustment is u1=kp (curentValue 1-totvalue 1);
the output value corresponding to the gas flow controller module in the 2 nd adjustment is U2 = Kp (currentValue 2-total Value 2);
the application of the scaling factor P, i.e. scaling; the scaling is based on the difference between the current value and the target value, multiplied by a factor of Kp to obtain an output value that directly affects the next current value change.
The differential coefficient D in the PD principle is actually the differential of the error, and the addition of the error 1 is err (1). Error 2 is err (2), the differential of error err is err2-err1, multiplied by differential coefficient D, generally called KD, K is amplification factor, when the actuator has error of 1 st time after 1 st time adjustment, has error of 2 nd time after 2 nd time adjustment, and combines P coefficient, PD combines, according to the experience calculation of error value during each time adjustment, the coefficient of D can be selected; if the error is smaller and smaller, then the post-differentiation must be a negative value; the positive value of the negative value is smaller than the value obtained by simply using the proportional adjustment after multiplying the negative value by a D coefficient and adding the proportional adjustment, so that the damping effect is started, the system area is stable due to the damping effect, and the PD combined formula is as follows after the analysis:
the output value of the middle U gas flow controller module is t times of the output value of the gas flow controller module; KP is K and is the magnification, P is the proportionality coefficient; />Where err is the error and t is the number of errors; />Where K is the magnification and D is the differential coefficient; />Indicating the rate of change of the error over time.
The gas flow controller module adopts PD control combination, P is proportional control, D is differential control according to pipeline gas conditions acquired by the flow sensor, and is used for controlling the gas electromagnetic valve according to the real-time flow value of the gas so as to adjust the output gas flow.
The PD controlling means can in the gas flow controller module, can real-time supervision gas's flow condition, can be according to if actual gas flow value and flow sensor collection's pipeline gas threshold ratio are different, control gas solenoid valve adjusts the gas flow in the input pipeline, in case the abnormal conditions is found, gas flow controller control gas solenoid valve action, the opening degree through the gas solenoid valve comes the hole to the flow of getting into the gas in the combustor, plays the effect of guarantee safe handling.
The matched pipeline is provided with a pressure transmitter which is used for receiving and converting pressure electric signals and is connected with the gas flow controller module.
The control device is connected with the gas flow controller module, the gas enters the gas flow controller module through the flow sensor and each matched pipeline, the pressure transmitter works at the moment, the gas value of the gas flow controller module in real time is monitored through the displayed pressure value, when the gas value in the gas flow controller module is different from the preset gas value, the pressure transmitter immediately measures and converts the gas value into a pressure value and an electric signal to display, and meanwhile, the pressure transmitter transmits the pressure electric signal to the microcomputer additional A/D converter.
The pressure transmitter consists of a sensing portion, measurement circuitry and process connection, the capacitance of the capacitor being determined by the size, shape, relative position of its two plates and the dielectric constant of the dielectric medium between the plates; if one plate is fixed and the other plate changes its position with pressure, the capacitance of the capacitor changes with pressure, and the change of capacitance is converted into current or voltage output signal by measuring circuit, and sent to related unit to realize display and control.
The pressure transmitter converts the measured pressure into an electric signal in the range of the measured pressure, and the change of the electric signal converted by the measured pressure is tiny, so that the electric signal is amplified by the pressure transmitter, and the microcomputer is convenient for the additional A/D converter to receive the signal.
The pressure transmitter converts weak non-electric pressure-pressure variable sensed by the sensor into a transmissible standardized signal to be output, and the electric signal is attenuated in the transmission process, so that the transmitted electric signal is required to be amplified, the converted electric signal and the pressure variable have a certain continuous functional relation comprising a linear function, the pressure variable comprises positive gauge pressure, negative gauge pressure, differential pressure and pressure, and the pressure transmitter can be divided into two types of a common pressure transmitter of 0.001-20MPa and a micro differential pressure transmitter of 0-1.5kpa according to the pressure measuring range, so as to supply an additional A/D converter of a microcomputer for connection measurement and process adjustment.
The microcomputer is additionally provided with an A/D converter which converts the electric signals transmitted by the pressure transmitter into digital signals which can be identified by the convolutional neural network monitoring integrated platform.
The process of a/D conversion has four stages, namely sample, hold, quantization and encoding.
Sampling is the process of changing a continuous time signal into a discrete time signal; the analog signal which is sampled, time-continuous and value-continuous becomes a signal with discrete time and value-continuous, which is called a sampling signal; the sampling circuit is equivalent to an analog switch, and the analog switch works periodically; theoretically, the closing time of the analog switch approaches 0 in each period; the moment when the analog switch is closed includes the sampling moment, we "take" a "sample" of the analog signal; the holding is a process of changing a time-discrete, numerical value-continuous signal into a time-continuous, numerical value-discrete signal; during quantization and encoding, the holding circuit acts as a constant voltage source that "holds" the signal voltage at the sampling instant at the input of the quantizer.
The neural network monitoring integrated platform comprises a microcomputer additional A/D converter, a convolutional neural network monitoring system connected with the microcomputer additional A/D converter through signals, and a characteristic extraction unit and a layering unit, wherein the characteristic extraction unit comprises a convolutional layer module and a pooling layer module;
the convolution layer module in the characteristic extraction unit is used for receiving the electric signal information, selecting and adjusting the electric signal information according to actual conditions through the pooling layer module, preventing the phenomenon of over fitting or under fitting, and effectively reducing the size of a parameter matrix so as to reduce the number of parameters in a final connection layer; the adjusting unit is used for adjusting the electric signal according to the received compensation signal;
it should be noted that, the gas flow controller module of the invention adopts the flow sensor and PD combined control principle to regulate the gas electromagnetic valve to control the gas flow, in order to ensure the real-time performance of the gas flow, the control structure of the gas flow controller module is single, the threshold ratio parameter of the pipeline gas is fixed, after introducing the convolutional neural network monitoring system, the problems that the traditional analog circuit fault diagnosis method is seriously dependent on feature extraction and selection can be overcome, and the like, the fault diagnosis monitoring is carried out by utilizing a deep confidence network, the network model can effectively classify long sequence data in the signals of the additional A/D converter output circuit of the microcomputer, the convolutional neural network directly carries out feature extraction on the acquired data, the labor cost of the feature extraction of the fault monitoring diagnosis is reduced, and the universality of the gas flow monitoring system is effectively improved;
the microcomputer is additionally provided with an A/D converter which converts the electric signal transmitted by the pressure transmitter into a digital signal which can be identified by a computer network;
example 2:
the method comprises the steps that a convolution neural network monitoring integration platform based on integration of a microcomputer additional A/D, a convolution neural network and a gigabit network interface is built, the microcomputer additional A/D transmits collected data to a convolution neural network system monitoring system through the gigabit network interface, the convolution neural network monitoring integration platform carries out model training and diagnosis on the received data, and diagnosis results are displayed on a display module interface through the convolution neural network monitoring system;
the feature extraction unit in the convolutional neural network monitoring integrated platform is used for receiving the electric signal information, and because the circuit response signal generated by the gas equipment for controlling the gas flow is sampled and preprocessed to form one-dimensional data, the one-dimensional convolutional neural network is selected when the convolutional neural network system monitoring system model is designed, the algorithm idea of the convolutional neural network monitoring integrated platform combines a feature extraction layer module and a classification layer module into a whole, the feature extraction layer comprises a convolutional layer and a downsampling layer, and the classification layer module mainly comprises a full connection layer:
(1) The convolution layer is the most core module of the convolution neural network, and compared with the full connection layer, the convolution layer adopts a non-full connection type weight sharing calculation mode, when an input sample is one-dimensional characteristic data, the convolution core adopts a one-dimensional structure to carry out convolution operation from left to right on each sample data, and as the weight of the convolution core is kept unchanged, parameters in the neural network are reduced in a local connection mode, the calculation complexity is reduced, and then the characteristic vector is subjected to nonlinear transformation output through an activation function.
Assume that the input of the convolutional layer isWhere k represents a single sample, m represents the total number of samples, n represents the data dimension of each sample,/->An output vector which is the j-th feature vector in the t-1 th layer,/th feature vector>For the ith convolution kernel of the jth feature vector corresponding to the last layer in the t-th layer,/->Representing the output threshold of the convolution layer, and then completing the calculation of the convolution layer by activating the function, the mathematical model can be described as:
)
on the upper partThe j-th eigenvector at layer t is represented, the convolution operation is represented, f () represents the activation function, the convolution kernel size is 3*3, and the step size is 1.
(2) The pooling layer is also called as a downsampling layer and is used for retaining the characteristics extracted by the convolution layer and reducing the operation complexity; the common operation methods of the pooling layer are max-pooling and mean-pooling; maximum pooling reduces the number of model parameters by preserving significant features ignoring non-significant features, i.e., selecting the maximum value of an image region as the value after pooling of that region; the averaging pooling retains a large amount of information of the 'secondary importance' elements, but easily causes blurring of the feature map; the convolution neural network system monitoring adopts maximum pooling to downsample the output characteristics of the convolution layer, and the formula is as follows:
in the method, in the process of the invention,(k) In the j-th eigenvector represented in layer t, the pooling width is the maximum value of neuron k in v,/>The maximum pooling results obtained at the t+1 layer neurons are shown.
(3) An activation function, wherein the activation function in the neural network is a nonlinear mathematical function, which increases the ability to learn and solve complex problems, thus making the expression ability of the deep neural network more powerful; the main purpose of the activation function is to convert the input signal of a node in the CNN into an output signal, and then take the output signal as the input of the next layer; because the Sigmoid and Tanh functions have a common problem, namely when the independent variable is large or small, gradient disappearance is easy to cause, so that the network convergence speed is slow; in recent years, relu functions are more used, and compared with other activation functions, the Relu functions are selected as the activation functions of convolution layer operation because the Relu functions effectively solve the problem of overfitting in the training process and have simple mathematical operation, and the mathematical expression is as follows
，
The expression x is an output value after the linear transformation of the neural network, the ReLU converts the result of the linear transformation into a nonlinear value, and the nonlinear function is obviously obtained after the function mapping is activated, so that the nonlinear classification is easy to realize.
Observing function curves of sigmoid and tanh, when the input of the sigmoid and tanh is between [ -1,1], the function value is sensitive to change, and once the function value approaches or exceeds an interval, the function curve loses sensitivity and is in a saturated state, so that the accuracy value of the neural network prediction is affected; the output and input of tanh can keep nonlinear monotonic rising and falling relation, data can be concentrated around zero, and if the activation value is large or small, the gradient of the activation function in the area is small, so that the training speed is slow.
The ReLu function, which is called Rectified Linear Unit in full, the Chinese name is a linear rectification function, and is an activation function commonly used in a neural network; in a general sense, the method refers to a ramp function in mathematics, and a ReLu activation function is used as an output result obtained by nonlinear transformation in a convolutional neural network;
the ReLU activation function is proposed to solve the gradient vanishing problem, and the gradient of the ReLU can only take two values: 0 or 1, when the input is less than 0, the gradient is 0; when the input is greater than 0, the gradient is 1; the advantages are that: the running of the gradient of ReLU does not converge to 0, and the result of the running can only take two values: 0 or 1; if the value is 1, the gradient keeps the value unchanged for forward propagation; if the value is 0, the gradient stops propagating forward from that location.
(4) Before entering the full-connection layer, the extracted features are input into a flattening layer to be flattened to form one-dimensional feature vectors, the one-dimensional feature vectors are used as input of the full-connection layer, and then the full-connection layer performs feature extraction and classification on the one-dimensional vectors again and outputs classification results through a softmax layer; layer t of full connection layerjThe weighted inputs of the individual nodes are:
in the method, in the process of the invention,representing weights between the kth neuron of the t-1 th layer and the jth neuron of the t layer; />An output value representing the t-1 layer; />Representing bias of all neurons of t-1 layer to j-th neurons of t layer, when the layer is used as output layer and the activation function adopts softmax, the activation value of j-th node is->,
In the middle ofThe values are all elements of the input vector of the softmax function, they can be any numerical value; />The standard exponential function is applied to each element of the input vector, not fixed within the (0, 1) range,/for>The term at the bottom of the formula is a normalized term that ensures that all output values of the function are neutralized to 1 and that each value is in the (0, 1) range, thus constituting an effective probability distribution.
The essence of the softmax function is to convert a vector into a probability distribution; specifically, for each element in the vector, softmax converts it to a real value between 0-1, while ensuring that the sum of all elements is equal to 1; all outputs of the softmax layer are positive numbers and sum to 1, and thus can be seen as
Probability distribution; because the output of the softmax layer neurons is equal to the ratio of the input of each neuron to the sum of the inputs of all neurons of the layer, the larger the output value of a certain neuron is, the higher the possibility that the corresponding class of the neuron is the true class is; the class corresponding to the neuron with the largest output value can be selected as the prediction result.
To train the neural network, a cross entropy loss function is typically added after the softmax layer; it may measure the gap between the actual output probability distribution and the target probability distribution,
(5) The loss function is used for evaluating the difference between the predicted value and the true value of the network model, and the better the convergence effect of the loss function is, the better the performance of the model is in general; the common calculation methods include a least square method, a maximum likelihood estimation method and cross entropy, and the cross entropy error is used as a loss function in a deep neural network more:
in the method, in the process of the invention,is a cross entropy error, m represents the input sample batch size; />Representing sample x at output layer nodejIs a true value of (2); />Representing sample x at nodejCan be represented by a softmax layer activation value;
the above-described embodiment 2 completes the overall process setup of the forward propagation of the convolutional neural network algorithm, in which the neural network weights and inputs the activation function for each neuron, producing outputs that are the inputs to the next layer until the output to the output layer is ultimately produced.
In convolutional neural networks, "forward" generally refers to forward propagation, also known as feed forward transfer, which is a fundamental operation of a neural network to process and transform input data in the network to ultimately yield an output result.
During the forward computation, the weight and bias of each neuron are fixed, so that the forward computation can be performed efficiently; at the same time, forward computation is also the basis of the back-propagation algorithm, since back-propagation requires the use of intermediate results in the forward computation process to update the weights and biases of neurons.
Example 3:
in the forward propagation process of the neural network, input data is transformed through a series of operations and parameters to finally generate an output result; in the back propagation process, calculating the influence of each parameter on the loss function through a chain rule; specifically, the gradient of the output result of the loss function is calculated first, then the gradient of each operation and parameter is calculated layer by layer, and finally the gradient of each parameter is obtained.
In convolutional neural networks, the backward, i.e., back propagation algorithm, refers to computing gradients for each parameter in the neural network model, as opposed to the forward propagation process, which are used to update the parameters of the model during training to minimize the loss function and improve model performance.
The method comprises the steps of constructing a convolutional neural network monitoring integrated platform based on microcomputer additional A/D, convolutional neural network and gigabit network interface integration, wherein a classification layer unit of the monitoring integrated platform comprises a full-connection layer and an output layer and consists of a counter propagation model training algorithm; the process firstly derives the output value of the loss function, then solves the gradient of the weight and bias of the full-connection layer, the pooling layer, the convolution layer according to the chain rule, and then carries out optimization training on the gradient through an optimization algorithm.
(1) And calculating the gradient of the full connection layer, and deriving the loss function output value of a single sample to obtain:
due toRepresenting the true value, which is usually assumed to be 1 in the case of multiple classifications, the derivative of the fully connected layer with respect to the weight of the loss function can be expressed as,
in the middle ofIs cross entropy error, < >>Represents the t th layer and the k th layerjWeights between neurons, +.>Layer t, thjWeighted inputs of the individual nodes.
Due to the fact that in the pairConsideration is needed in derivationjAnd (3) withkWhether or not to be equal whenj＝kThis can be expressed as a time-scale representation,
when (when)j≠kThis can be expressed as a time-scale representation,
the derivative of the final full connection layer bias can be expressed as
Similarly, the derivative of the loss function with respect to bias can be derived, i.e
(2) Pooled layer gradient calculation
Since the forward propagation process in the convolutional neural network uses maximum pooling to downsample the feature vector, the pooling layer needs to be upsampled in the backward propagation operation so as to restore the feature region in front of the pooling layer; however, in the operation of the pooling layer, the pooling core does not relate to operation parameters, so that only the derivative of the loss function on the input value of the pooling layer is needed to be calculated;
in the middle ofIs cross entropy error, < >>Representing the true value, K tableShowing the number of times; />The maximum pooling results obtained at the t+1 layer neurons are shown.
(3) Convolution layer gradient computation
The backward propagation process of the convolution layer mainly comprises two parts, wherein the first part is required to conduct derivation on an activation function used in forward propagation; the second part solves the derivative of the weight and the bias of the convolution layer according to the value obtained after deriving the activation function;
(4) The convolutional neural network monitoring Adam optimizing algorithm carries out optimizing training on the weight and bias of the model, and the method solves the problem that the learning rate in the SGD algorithm is difficult to select by adopting an adaptive learning rate, and also solves the problem that the learning rate in the SGD cannot be adaptively adjusted by calculating the accumulated gradient by combining a method of exponentially weighted moving average; the optimization algorithm used to replace random gradient descent in the present algorithm build model,
the display module displays the monitored data amount and forms a visual chart.
The visual model is integrated in the convolutional neural monitoring system, so that the gas flow monitoring data are converted into a graphical form, and gas pipeline maintenance and operation maintenance personnel can more intuitively understand and analyze the data; this conversion typically includes the following steps: data preparation: in the visualization model, data needs to be prepared firstly, including collecting data, cleaning the data, converting the data format and the like; and (3) data processing: on the basis of data preparation, certain processing is needed to be carried out on the data so as to be convenient for better visualization;
visual model selection: selecting a proper visual model according to the characteristics and the requirements of the data; common visual models include histograms, line graphs, scatter plots, pie charts, radar charts, and the like;
visual design: after the visual model is selected, the visualization needs to be designed, including chart colors, fonts, coordinate axes and the like; the design aims to make the visualization more attractive and readable; visual presentation: the final step is to present the visualization to the user, typically in the form of a graphical interface or web page.
The visual model has wide application in the fields of data analysis, machine learning, artificial intelligence and the like, and is mainly characterized in the following aspects of data exploration: the visual model can help the user to understand the data more intuitively, explore the characteristics, rules and anomalies of the data and provide basis for subsequent analysis and modeling; data analysis: the visual model can help the user to analyze the data more deeply, reveal potential rules and correlations in the data, and find hidden information in the data; decision support: the visual model can present the data in a graphical form, so that a user can more easily understand and compare different data, and decision making and prediction are supported; interactive display: the visualization model may support the user to interactively explore the data, and in addition, the visualization model has the following advantages: visual and understandable, the visual model can convert abstract data into visual graphic form, so that a user can understand and master the data more easily and feed back the data quickly: the visual model can quickly feed back the characteristics and rules of the data, so that a user can quickly explore the data and make corresponding decisions; the method can be customized: the visual model can be customized according to different requirements and data characteristics, so that a user can explore data more flexibly; multidimensional rendering: the visualization model may present data in multiple dimensions, such as time, geographic location, classification, etc., at the same time, allowing the user to better understand aspects of the data.
The manual key module can facilitate the maintenance of the gas pipeline, and an operation and maintenance personnel to manually repair the numerical range of the gas flow controller module according to the monitoring result.
The preferred embodiments of the invention disclosed above are merely intended to help illustrate the invention; the preferred embodiments are not intended to be exhaustive or to limit the invention to the precise form disclosed; obviously, many modifications and variations are possible in light of the above teachings; the embodiments were chosen and described in order to best explain the principles of the invention and the practical application, to thereby enable others skilled in the art to best understand and utilize the invention; the invention is limited only by the claims and the full scope and equivalents thereof.
Claims (10)
1. The gas equipment capable of monitoring and controlling the gas flow is characterized by comprising a flow sensor, a gas flow controller module, a gas electromagnetic valve, a manual key module, a pressure transmitter, a microcomputer additional A/D converter, a convolutional neural network monitoring integrated platform and a display module which are sequentially connected through matched pipelines; the gas solenoid valve and the gas flow controller module are connected with the pressure transmitter, the pressure is converted into an electric signal, the electric signal is connected with the microcomputer additional A/D converter and the convolutional neural network monitoring integrated platform through the convolutional neural network, the characteristic extraction unit and the classification layer unit of the convolutional neural network monitoring integrated platform monitor the gas quantity, finally the display module displays the monitored data quantity and forms a visual chart, the gas pipeline maintenance and operation staff can conveniently repair the numerical range of the gas flow controller module manually through the manual key module according to the monitoring result.
2. The gas appliance for monitoring and controlling gas flow according to claim 1, wherein the flow sensor is adapted to detect the amount of air and the amount of gas in the transfer duct in real time, thereby controlling the air-fuel ratio to an optimum value, and facilitating the control of the gas flow by the gas solenoid valve and the gas flow controller module.
3. The gas appliance for monitoring and controlling gas flow according to claim 1, wherein the gas flow controller module is connected to the flow sensor and the pressure transmitter, and controls the gas solenoid valve to adjust the output gas flow; the gas flow controller module is mutually connected and matched with the flow sensor, and meanwhile, the PD principle is added to the gas flow controller module to improve the control mechanism.
4. A gas appliance for monitoring and controlling gas flow according to claim 1 or 3, wherein the gas flow controller module employs PD control combination, i.e. proportional control, D differential control, according to the pipeline gas conditions collected by the flow sensor, for controlling the gas solenoid valve to regulate the output gas flow according to the real-time flow value of the gas.
5. A gas appliance for monitoring and controlling gas flow as claimed in claim 1 wherein the associated piping is provided with a pressure transducer for receiving and converting electrical pressure signals for connection to the gas flow controller module.
6. The gas device for monitoring and controlling gas flow according to claim 5, wherein the pressure transmitter converts the measured pressure into an electrical signal in the range of the measured pressure, and the pressure transmitter is required to amplify the electrical signal because the change of the measured pressure into the electrical signal is small, so that the microcomputer can receive the signal by adding the a/D converter.
7. A gas appliance for monitoring and controlling gas flow according to claim 1 or 6, wherein the microcomputer is additionally provided with an a/D converter for converting the electrical signal transmitted by the pressure transmitter into a digital signal recognizable by the convolutional neural network monitoring integration platform.
8. The gas device capable of monitoring and controlling gas flow according to claim 1, wherein the convolutional neural network monitoring integration platform comprises a microcomputer additional A/D converter and a convolutional neural network monitoring system; the integrated platform is connected through the convolutional neural network monitoring, the microcomputer is additionally provided with an A/D converter for acquiring an electric signal of the pressure transmitter, converting the electric signal into a digital signal quantity which can be identified by the convolutional neural network monitoring integrated platform, and monitoring the gas quantity; meanwhile, the convolutional neural network monitoring integration platform integrates a matched algorithm idea, monitors and diagnoses the input digital signal quantity through a feature extraction layer and a classification layer in an algorithm model, and transmits a monitoring result to a display module.
9. A gas appliance for monitoring and controlling gas flow according to claim 1 or 8, characterized in that the display module presents the monitored data quantity and forms a visual chart.
10. The gas appliance for monitoring and controlling gas flow according to claim 1, wherein the manual key module is capable of facilitating pipeline maintenance and maintenance personnel to manually restore the numerical range of the gas flow controller module according to the monitoring result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311090625.7A CN116817306B (en) | 2023-08-29 | 2023-08-29 | Gas equipment capable of monitoring and controlling gas flow |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311090625.7A CN116817306B (en) | 2023-08-29 | 2023-08-29 | Gas equipment capable of monitoring and controlling gas flow |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116817306A true CN116817306A (en) | 2023-09-29 |
| CN116817306B CN116817306B (en) | 2023-12-01 |
Family
ID=88116948
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311090625.7A Active CN116817306B (en) | 2023-08-29 | 2023-08-29 | Gas equipment capable of monitoring and controlling gas flow |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116817306B (en) |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090005952A1 (en) * | 2007-06-26 | 2009-01-01 | General Electric Company | Systems and Methods for Using a Combustion Dynamics Tuning Algorithm with a Multi-Can Combustor |
| CN106292785A (en) * | 2015-05-18 | 2017-01-04 | 广东兴发铝业有限公司 | Aluminum-bar heating furnace ignition temperature automaton based on neutral net |
| CN208804710U (en) * | 2018-12-28 | 2019-04-30 | 广东省特种设备检测研究院佛山检测院 | Gas fired-boiler gas regulating device |
| US20210038321A1 (en) * | 2018-03-12 | 2021-02-11 | Koninklijke Philips N.V. | Ultrasound imaging dataset acquisition for neural network training and associated devices, systems, and methods |
| CN112966547A (en) * | 2021-01-05 | 2021-06-15 | 北京市燃气集团有限责任公司 | Neural network-based gas field abnormal behavior recognition early warning method, system, terminal and storage medium |
| CN113790447A (en) * | 2021-11-02 | 2021-12-14 | 上海倍卡福智控技术有限公司 | Gas flow control device and method for combustor |
| CN115047848A (en) * | 2022-06-10 | 2022-09-13 | 北卡科技有限公司 | Industrial control system anomaly detection method based on PID neural network |
| DE102021206638A1 (en) * | 2021-05-10 | 2022-11-10 | Vitesco Technologies GmbH | Computer-implemented method and control apparatus for controlling a powertrain of a vehicle using a convolutional neural network. |
| CN115875665A (en) * | 2022-08-31 | 2023-03-31 | 上海明华电力科技有限公司 | On-line monitoring method for temperature of smoke at furnace outlet of coal-fired power station boiler |
| CN115879026A (en) * | 2022-12-07 | 2023-03-31 | 吉林大学 | Rotary drilling rig hydraulic system fault diagnosis method based on convolutional neural network-integrated classifier |
| KR20230112308A (en) * | 2022-01-20 | 2023-07-27 | 성균관대학교산학협력단 | Method and device for monitoring combustion of combustors using machine learning |
-
2023
- 2023-08-29 CN CN202311090625.7A patent/CN116817306B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090005952A1 (en) * | 2007-06-26 | 2009-01-01 | General Electric Company | Systems and Methods for Using a Combustion Dynamics Tuning Algorithm with a Multi-Can Combustor |
| CN106292785A (en) * | 2015-05-18 | 2017-01-04 | 广东兴发铝业有限公司 | Aluminum-bar heating furnace ignition temperature automaton based on neutral net |
| US20210038321A1 (en) * | 2018-03-12 | 2021-02-11 | Koninklijke Philips N.V. | Ultrasound imaging dataset acquisition for neural network training and associated devices, systems, and methods |
| CN208804710U (en) * | 2018-12-28 | 2019-04-30 | 广东省特种设备检测研究院佛山检测院 | Gas fired-boiler gas regulating device |
| CN112966547A (en) * | 2021-01-05 | 2021-06-15 | 北京市燃气集团有限责任公司 | Neural network-based gas field abnormal behavior recognition early warning method, system, terminal and storage medium |
| DE102021206638A1 (en) * | 2021-05-10 | 2022-11-10 | Vitesco Technologies GmbH | Computer-implemented method and control apparatus for controlling a powertrain of a vehicle using a convolutional neural network. |
| CN113790447A (en) * | 2021-11-02 | 2021-12-14 | 上海倍卡福智控技术有限公司 | Gas flow control device and method for combustor |
| KR20230112308A (en) * | 2022-01-20 | 2023-07-27 | 성균관대학교산학협력단 | Method and device for monitoring combustion of combustors using machine learning |
| CN115047848A (en) * | 2022-06-10 | 2022-09-13 | 北卡科技有限公司 | Industrial control system anomaly detection method based on PID neural network |
| CN115875665A (en) * | 2022-08-31 | 2023-03-31 | 上海明华电力科技有限公司 | On-line monitoring method for temperature of smoke at furnace outlet of coal-fired power station boiler |
| CN115879026A (en) * | 2022-12-07 | 2023-03-31 | 吉林大学 | Rotary drilling rig hydraulic system fault diagnosis method based on convolutional neural network-integrated classifier |
Non-Patent Citations (1)
| Title |
|---|
| 李之涵: "基于卷积神经网络的工业互联网入侵检测方法研究", 硕士学位论文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116817306B (en) | 2023-12-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106644162B (en) | Ring main unit wire core temperature soft measurement method based on neighborhood preserving embedding regression algorithm | |
| CN106526112B (en) | A kind of water quality toxicity detection method based on fish activity analysis | |
| CN111539132B (en) | Dynamic load time domain identification method based on convolutional neural network | |
| KR102104026B1 (en) | Real-time Solenoid Valve Monitoring System and Method Using Artificial Neural Network | |
| CN112132430B (en) | Reliability evaluation method and system for distributed state sensor of power distribution main equipment | |
| CN113743016B (en) | Engine residual life prediction method based on self-encoder and echo state network | |
| CN109145516A (en) | A kind of analog circuit fault recognition methods based on modified extreme learning machine | |
| Wang et al. | Prediction model of natural gas pipeline crack evolution based on optimized DCNN-LSTM | |
| CN114397043B (en) | Multi-point temperature intelligent detection system | |
| CN114548311B (en) | Hydraulic equipment intelligent control system based on artificial intelligence | |
| CN112903929A (en) | Food quality detection system | |
| CN116817306B (en) | Gas equipment capable of monitoring and controlling gas flow | |
| CN117114913A (en) | Intelligent agricultural data acquisition system based on big data | |
| CN112836719B (en) | Indicator diagram similarity detection method integrating two classifications and triplets | |
| CN111814403B (en) | Reliability assessment method for distributed state sensor of distribution main equipment | |
| CN117574275A (en) | Abnormality detection and classification method for power system | |
| CN114970745B (en) | Intelligent security and environment big data system of Internet of things | |
| CN116757078A (en) | Method and system for measuring flow velocity of pulverized coal based on acting force | |
| CN114399024B (en) | Oil gas concentration big data intelligent detection system | |
| CN115062764B (en) | Intelligent illuminance adjustment and environmental parameter Internet of things big data system | |
| CN114417242B (en) | Big data detection system for livestock and poultry activity information | |
| CN113758627B (en) | Water supply network transient flow event detection method | |
| CN113988210A (en) | Method and device for restoring distorted data of structure monitoring sensor network and storage medium | |
| Tsang | Sensor data validation using gray models | |
| CN110045716B (en) | Method and system for detecting and diagnosing early fault of closed-loop control system |
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 |