CN103302777B - Neural-network-based energy consumption anomaly detection method of tire vulcanizing machine - Google Patents

Neural-network-based energy consumption anomaly detection method of tire vulcanizing machine Download PDF

Info

Publication number
CN103302777B
CN103302777B CN201310261390.3A CN201310261390A CN103302777B CN 103302777 B CN103302777 B CN 103302777B CN 201310261390 A CN201310261390 A CN 201310261390A CN 103302777 B CN103302777 B CN 103302777B
Authority
CN
China
Prior art keywords
energy consumption
data
neutral net
net
node
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201310261390.3A
Other languages
Chinese (zh)
Other versions
CN103302777A (en
Inventor
杨海东
刘国胜
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guangdong University of Technology
Original Assignee
Guangdong University of Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Guangdong University of Technology filed Critical Guangdong University of Technology
Priority to CN201310261390.3A priority Critical patent/CN103302777B/en
Publication of CN103302777A publication Critical patent/CN103302777A/en
Application granted granted Critical
Publication of CN103302777B publication Critical patent/CN103302777B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Abstract

The invention discloses a neural-network-based energy consumption anomaly detection method of a tire vulcanizing machine. The energy consumption anomaly detection method comprises the steps of collecting technical data, calculating theoretical energy consumption, collecting energy consumption data, calculating actual energy consumption, calculating energy source efficiency, collecting product parameters, collecting environment parameters, and establishing a neural network for predicating. According to the energy consumption anomaly detection method disclosed by the invention, temperature signals, pressure signals and steam flow signals in a tire vulcanizing process can be continuously collected to be taken as basic data; then, based on a physical process and a chemical change of vulcanizing, the effective energy consumption and energy efficiency in the vulcanizing process can be accurately calculated; finally, specification parameters, performance parameters and environment parameters of a tire are taken as input vectors to carry out neural network study, and after the parameter learning of a plurality of sets of standard workpieces and a processing process, a connection weight in the neural network is determined; after a fluctuation section threshold value, which is allowed by the energy consumption of a product, is determined, the trained neural network can carry out real-time predication and judgment on an energy consumption condition of the current vulcanizing machine to find energy consumption anomaly information in time.

Description

A kind of tyre vulcanizer energy consumption method for detecting abnormality based on neutral net
Technical field
The present invention relates to a kind of method for detecting abnormality of tire manufacturing art, is a kind of tyre vulcanizer based on neutral net energy consumption method for detecting abnormality in process of production specifically.
Background technology
China is manufacturing powerhouse, and the energy that manufacturing industry consumes every year accounts for the over half of domestic total amount, and therefore, manufacturing industry of making greater efforts to promote energy-saving and emission-reduction are inevitable choices that China realizes industrial upgrading transition and development low-carbon economy.Tire production industry is one of highly energy-consuming trade, its main production equipments is tyre vulcanizer, and along with the development of science and technology, the vulcanizer of computer-aided control is progressively applied, supplementary controlled system completes predetermined vulcanisation operation according to the operational order of setting, substantially increases the operational efficiency of equipment.But, tire vulcanization process comprises complicated chemical reaction and physical reactions, the factors such as the temperature of sulfuration, pressure and time all produce the quality of vulcanizate and determine impact, although the automatization level of equipment can be improved by computer-aided equipment, be difficult to carry out effective monitoring to the quality of production and anomalous event.
Vulcanizer is generally made up of four parts: clamp system, control system, pressure system and heating system.Clamp system is generally made up of frame and bolt, and control system is made up of electric cabinet and a secondary wiring, and pressure system is made up of hydraulic pressure plate and hydraulic test pump, and heating system is made up of heating plate and thermal insulation board.Before beginning sulfuration, first capturing embryo in mould by clamp system, carrying out pre-setting, then matched moulds by being filled with compressed air in capsule.
As shown in Figure 1, in sulfidation, metal pattern 1 temperature remains at 147 DEG C (wherein, metal pattern 1 is contained in heat insulation layer 4, and gusset forms air layer 5 by air between metal film 1 and heat insulation layer 4), be filled with 190 DEG C ~ 210 DEG C high-pressure saturated steams to vulcanizer capsule 2 simultaneously, capsule 2 is rapidly heated, then (vapor recycle is moved towards to stop high-temperature steam entering, be condensed water time out), then the high pressure of 2.0MPa ~ 2.6MPa is filled with to capsule 2, high-purity nitrogen, and keep stable high voltage, with the conditions of vulcanization of the high pressure that reaches a high temperature, this process will continue 20 ~ 30 minutes.After sulfuration terminates, be filled with cooling water and make it reduce to room temperature in capsule 2, then the air extracted out in capsule 2 carries out the demoulding (referring to metal pattern 1), removes the tire 3 vulcanizated.
Summary of the invention
For the problems referred to above, the invention provides a kind of tyre vulcanizer energy consumption method for detecting abnormality based on neutral net, it is on the physical reactions of tyre vulcanization and the theoretical foundation of chemical reaction, according to tyre vulcanizer operational factor, Product Process parameter and workshop condition parameter, with tyre vulcanizer energy use efficiency for monitoring objective, the change of neural network model to energy efficiency is adopted to monitor, with the production abnormal conditions in Timeliness coverage tire vulcanization process.
The present invention is achieved in that a kind of tyre vulcanizer energy consumption method for detecting abnormality based on neutral net, and it comprises the steps:
Step one: collection technology data, described process data comprises vapor (steam) temperature in capsule and pressure, and the vapor (steam) temperature collected by the predetermined sampling period and pressure mark by sequence respectively:
T=T 1,T 2,T 3,……,T n, (1);
P=P 1,P 2,P 3,……,P h, (2);
Wherein, T irepresent the capsule temperature in i-th sampling period, P irepresent the capsule pressure in i-th sampling period, 1≤i≤h, h is sampling number, △ T=T 2-T 1for the sampling period, metal outer mold keeps steady temperature by electric-heating-wire-heating, uses T 0mark;
Step 2: calculate theoretical energy consumption,
The heat of steam conduction in (a) capsule and pressure, divide period treatment by vapor (steam) temperature and pressure, the account form within each sampling period is as follows:
Wherein, n is evaporative substance amount, n ≈ 55.56m 1, m 1be reaction vapor quality, R=8.314J/(Kmol) be gas constant, c ≈ 4.2KJ/(Kg DEG C) be the specific heat capacity of steam;
B the heat of () metal outer mold conduction, in sulfidation, the temperature of metal outer mold remains at, and supposes that tire embryo reaches the reaction temperature of specifying by metal outer mold heat transfer, then the heat of its conduction is: W theoretical=cm 2(T 0-25), (4);
Wherein, c ≈ 1.7KJ/(Kg DEG C) be the specific heat capacity of rubber, m 2for the quality of tire embryo;
Step 3: gather energy consumption data, gathers the actual energy resource consumption in sulfidation, comprises vulcanizer power consumption W 1, compressor power consumption W 2, use quantity of steam m 3;
Step 4: calculate actual consumption, actual consumption is primarily of vulcanizer electric energy, compressor electric energy and the boiler energy consumption composition producing high temperature and high pressure steam, and the gross energy of actual consumption is: W always=W 1+ W 2+ m 32796KJ/Kg/Kg, (5);
Wherein, m 32796KJ/Kg/Kg is steam contained energy;
Step 5: calculate energy efficiency,
Step 6: gather product parameters, comprises specifications parameter and performance is joined, and described specifications parameter comprises multiple appearance and size parameters of embryo, uses x respectively 1, x 2..., x prepresent, described performance parameter comprises the multiple physical characteristics required by target tire product, uses y respectively 1, y 2..., y qrepresent;
Step 7: gather ambient parameter, it is the ambient parameter relevant to tyre vulcanizing, uses z respectively 1, z 2..., z krepresent;
Step 8: set up neural network prediction, adopt three layers of BP neutral net in the object module chosen, using the specifications parameter of product, performance parameter and ambient parameter as input variable, comprise p+q+k node altogether, intermediate layer is hidden layer, comprises m altogether 4individual node, target output layer comprises individual node, is the predicted value of energy efficiency, and the weight parameter between model interior joint adopts standard historical data to train; At the input layer of described BP neutral net, its input signal is specifications parameter x 1~ x p, performance parameter x p+1~ x p+qwith ambient parameter x p+q+1~ x p+q+k, adopt regular function to convert in input node, be converted into the value y of [-1,1] interval range i:
y i = 2 × x i - x min x max - x min - 1 , - - - ( 7 ) ;
Wherein, x maxfor inputting the upper bound of data, x minfor inputting the lower bound of data, y is the value after regularization; At the hidden layer of described BP neutral net, its excitation function adopts hyperbolic sigmoid function z j:
z j = φ ( net 1 ) = a tanh ( b × net 1 ) = a [ 1 - exp ( b × net 1 ) 1 + exp ( b × net 1 ) ] , - - - ( 8 ) ;
Wherein, net 1for the input signal of hidden layer, numerically equal weighted average and this Node B threshold sum of input layer output signal w ijfor the connection weight between input layer i and hidden layer node j, h jfor the threshold value of hidden layer node j, a, b are the parameter of sigmoid function; At the output layer of described neutral net, its excitation function adopts jump function J: J = ψ ( net 2 ) = 1 net 2 > 0 0 net 2 ≤ 0 , - - - ( 9 ) ;
Wherein, net 2for the input signal of output layer, numerically equal the weighted average of hidden layer output signal and the difference of this Node B threshold w jfor the connection weight between hidden layer node j and output layer node, e is the threshold value of output layer node;
Step 9: detect energy consumption abnormal, neutral net after having trained utilizing historical standard floor data, by importing new data sample in real time, can analyze the energy consumption of new samples manufacture process, wherein J=0 represents without exception, and contrary J=1 indicates abnormal generation.
As the further improvement of such scheme, described specifications parameter comprises weight, surface area, thickness three appearance and size parameters of embryo.
As the further improvement of such scheme, described performance parameter comprises hardness required by target tire product, breaking tenacity, percentage elongation, expansion, elasticity five physical characteristics.
As the further improvement of such scheme, the ambient parameter relevant to tyre vulcanizing comprises temperature and humidity.
As the further improvement of such scheme, the vapor (steam) temperature T=T in acquisition capsule 1, T 2, T 3..., T nwith pressure data P=P 1, P 2, P 3..., P hafter, for making calculating more accurate, first carry out interpolation fitting to data, described interpolation method adopts Lagrange quadratic interpolation formula:
L ( i + x ) = 1 2 ( x - 1 ) ( x - 2 ) T i - x ( x - 2 ) T i + 1 + 1 2 x ( x - 1 ) T i + 2 , - - - ( 10 ) ;
Wherein, the observation T of continuous three times is utilized i, T i+1, T i+2or P h, P h+1, P h+2carry out interpolation strengthening to sensor signal, will shorten in the sampling period further, improve the accuracy of data, after carrying out interpolation processing, the formula (3) in employing step 2 and formula (4) calculate the steam energy consumption Q in capsule respectively theoreticalwith metal outer mold energy consumption W theoretical.
As the further improvement of such scheme, described BP nerve network system error is defined as:
E = 1 2 Σ i = 1 n ( J ^ i - J i ) 2 , - - - ( 11 ) ;
The correction △ w of output layer weights is revised successively according to error gradient descent method j, the correction △ e of output layer threshold value, the correction △ w of hidden layer weights ij, the correction △ h of hidden layer threshold value j, more new formula is as follows for corresponding improve parameter unification:
α ← α + Δα = α - η ∂ E ∂ α , - - - ( 12 ) .
Preferably, node in hidden layer turns cross-validation method by wheel and determines: training data is divided into N group, each will wherein N-1 as training data, BP neural network model is trained, remain 1 group as test data, to judge the precision of prediction of institute's training BP neutral net, like this wheel turns the consensus forecast precision of N as neutral net, finally choose consensus forecast precision maximum time node in hidden layer as optimal node number.
Again preferably, in the formula (8) of described step 8, getting a=1.176, b=2/3, thus ensure φ (0) ≈ 0.5, be approximately linear, and the extreme value of second dervative is at ± 2 places between [-1,1].
Further, train BP neutral net under standard condition after, be applied in the energy consumption abnormality detection in daily tyre vulcanizing: if actual efficiency and BP neural network prediction efficiency error are within the scope of reasonable interval:
| e ^ - e | ≤ ϵ , - - - ( 13 ) , Then think without exception;
Otherwise then there is abnormal generation.
As the further improvement of such scheme, sampling period △ T is 1 second.
The present invention by temperature signal, pressure signal and the steam flow signal in continuous acquisition tire vulcanization process, based on data; Then based on physical process and the chemical change of sulfuration, the available energy dissipation in sulfidation and efficiency is accurately calculated; Finally the specifications parameter (weight, surface area, thickness) of tire, performance parameter (hardness, breaking tenacity, percentage elongation, expansion, elasticity) and ambient parameter (temperature, humidity) are carried out neural network learning as input vector, after the parameter learning through too much organizing standard workpiece and process, determine the connection weight in neutral net; After determining the waving interval threshold value that product energy consumption allows, real-time estimate judgement can be carried out to the energy consumption operating mode of current vulcanizer according to the neutral net trained, Timeliness coverage energy consumption abnormal information.
The useful effect of the inventive method shows as:
(1) the present invention sets up tyre vulcanization efficiency computational methods, the method using the ratio of technology theory energy consumption and actual consumption as energy efficiency indexes, with the existing energy valid value calculated by input and output mode, this method is more scientific and reasonable, and be associated with actual process process and technological parameter, energy efficiency indexes can reflect the use energy situation of each link;
(2) adopt neutral net to describe the complex relationship of product parameters, ambient parameter and energy efficiency indexes, be convenient to enterprise and formulate energy use planning, improve enterprise energy and adjust the degree of accuracy;
(3) method being detected the unusual condition in tyre production technology by energy Significant Change is proposed, compared with setting most high energy consumption exceptional value with tradition, this method detects the use energy situation of links and the equipment considering tire vulcanization process, more accurately and reliably, the timely and effective anomalous event of excavating in production of enterprise is helped.
Accompanying drawing explanation
Fig. 1 is the machined structure schematic diagram of namely institute of the present invention applicable object at present.
Fig. 2 is the flow process frame diagram implementing the inventive method.
Fig. 3 is the schematic diagram of BP neural network structure in the present invention.
Fig. 4 is the tire specification parameter schematic diagram selected by the present invention.
Fig. 5 is the another specifications parameter schematic diagram of tire selected in Fig. 4.
Detailed description of the invention
In order to make object of the present invention, technical scheme and advantage clearly understand, below in conjunction with drawings and Examples, the present invention is further elaborated.Should be appreciated that specific embodiment described herein only in order to explain the present invention, be not intended to limit the present invention.
The present invention utilizes the method for supervising of information-based means fortified tyre manufacture process energy consumption exception, build the simulation and prediction model based on neutral net, excavation tire product and environmental factor are to the internal association relation manufacturing energy consumption, the neutral net detection model of having been trained by the manufaturing data under standard condition, can carry out accurate forecast to the energy consumption abnormal conditions in tire production.
As shown in Figure 2, the tyre vulcanizer energy consumption method for detecting abnormality based on neutral net of the present invention comprises the following steps.
Step one: collection technology data, i.e. signals collecting.
Described process data comprises vapor (steam) temperature and pressure in capsule, the actual acquisition value of the technological factors such as the temperature of metal pattern, by sensor according to the burst in the sampling period Real-time Collection tire vulcanization process preset, in capsule, vapor temperature signal sequence T mark, pressure signal sequence P mark:
T i=T 1,T 2,T 3,……,T n(1)
P=P 1,P 2,P 3,……,P h(2)
Wherein, T irepresent the capsule temperature in i-th sampling period, P irepresent the capsule pressure in i-th sampling period, 1≤i≤h, h is sampling number, △ T=T 2-T 1for the sampling period, be taken as 1 second in the present embodiment.
In addition, metal outer mold keeps steady temperature by electric-heating-wire-heating, uses T 0mark.
Step 2: calculate theoretical energy consumption, namely theoretical energy consumption calculates.
Tire vulcanization process needs to carry out in the environment of HTHP, and its direct energy source derives from capsule and metal pattern.
The heat of steam conduction in (a) capsule and pressure
Generally, the temperature and pressure in capsule changes greatly at initial vulcanization step, rises fast afterwards, is finally progressively stabilized to preset value, and in capsule, the state change process of steam is comparatively complicated.For the theoretical energy consumption of steam change procedure in accurate Calculation capsule, divide period treatment by temperature and pressure, with first law of thermodynamics formulae discovery energy ezpenditure within each sampling period, its account form is as follows:
Wherein, n is evaporative substance amount, n ≈ 55.56m 1, m 1be reaction vapor quality, R=8.314J/(Kmol) be gas constant, c ≈ 4.2KJ/(Kg DEG C) be the specific heat capacity of steam.P iand T ifor the pressure signal of collection in step one and the sequential value of temperature signal.
The heat of (b) metal outer mold conduction
In sulfidation, the temperature of metal outer mold remains at T 0, suppose that tire embryo reaches the reaction temperature of specifying by metal outer mold heat transfer, then the heat of its conduction is:
W theoretical=cm 2(T 0-25) (4)
Wherein, c ≈ 1.7KJ/(Kg DEG C) be the specific heat capacity of rubber (main component of tire), m is the quality of tire embryo.
Step 3: gather energy consumption data, gathers the actual energy resource consumption in sulfidation by intelligence instrument, comprises vulcanizer power consumption W 1, compressor power consumption W 2, use quantity of steam m 3.
Step 4: calculate actual consumption, namely actual consumption calculates.
Actual consumption is primarily of vulcanizer electric energy, compressor electric energy and the boiler energy consumption composition producing high temperature and high pressure steam.Electric energy directly gathers by intelligent electric meter, and boiler energy consumption is then comparatively difficult to obtain, and calculates in this empirical equation with steam contained energy.The gross energy of actual consumption is:
W always=W 1+ W 2+ m 32796KJ/Kg/Kg (5)
Wherein last is steam contained energy, and the approximate enthalpy being multiplied by steam by steam consumption calculates.
Step 5: calculate energy efficiency, namely energy efficiency calculates.
Energy efficiency is determined by the ratio of theoretical energy consumption and actual total energy consumption, and its computing formula is as follows:
Step 6: gather product parameters.
Gather product parameters and comprise specifications parameter and performance parameter.Specifications parameter forms primarily of the appearance and size such as weight, surface area, thickness of embryo, uses x respectively 1, x 2..., x prepresent.Performance parameter, primarily of the physical characteristic such as hardness, breaking tenacity, percentage elongation, expansion, the elasticity composition required by target tire product, uses y respectively 1, y 2..., y qrepresent.Specifications parameter in this step and performance parameter are the technical standards that target product will reach, and by order, business formulates, and its data can obtain in advance from the ERP system or form ordering system of tire manufacturing shop.
Step 7: gather ambient parameter, the ambient parameter relevant to tyre vulcanizing mainly comprises the information such as temperature, humidity in workshop, uses z respectively 1, z 2..., z krepresent, its data are by being arranged on thermometer meter in workshop and humidometer measurement measures.
Step 8: set up neural network prediction, i.e. neural network prediction.
As shown in Figure 3, adopt three layers of BP neutral net in the object module chosen, using product specification parameter, performance parameter and ambient parameter as input variable, comprise p+q+k node altogether, intermediate layer is hidden layer, comprises m altogether 4individual node, target output layer comprises individual node, is the predicted value of energy efficiency.Weight parameter between model interior joint adopts standard historical data to train.
At the input layer of described neutral net, its input signal is specifications parameter x 1~ x p, performance parameter x p+1~ x p+qwith ambient parameter x p+q+1~ x p+q+k, adopt regular function to convert in input node, be converted into the value of [-1,1] interval range:
y i = 2 × x i - x min x max - x min - 1 - - - ( 7 )
Wherein, x maxfor inputting the upper bound of data, x minfor inputting the lower bound of data, y is the value after regularization.
At the hidden layer of described neutral net, its excitation function adopts hyperbolic sigmoid function:
z j = φ ( net 1 ) = a tanh ( b × net 1 ) = a [ 1 - exp ( b × net 1 ) 1 + exp ( b × net 1 ) ] - - - ( 8 )
Wherein, net 1for the input signal of hidden layer, numerically equal weighted average and this Node B threshold sum of input layer output signal w ijfor the connection weight between input layer i and hidden layer node j, h jfor the threshold value of hidden layer node j, a, b are the parameter of sigmoid function.
At the output layer of described neutral net, its excitation function adopts jump function J:
J = ψ ( net 2 ) = 1 net 2 > 0 0 net 2 ≤ 0 - - - ( 9 )
Wherein, net 2for the input signal of output layer, numerically equal the weighted average of hidden layer output signal and the difference of this Node B threshold w jfor the connection weight between hidden layer node j and output layer node, e is the threshold value of output layer node.
Step 9: detect energy consumption abnormal, i.e. energy consumption anomaly analysis.
Neutral net after having trained utilizing historical standard floor data, by importing new data sample in real time, can analyze the energy consumption of new samples manufacture process, wherein J=0 represents without exception, and contrary J=1 indicates abnormal generation.
In sum, a kind of tyre vulcanizer energy consumption method for detecting abnormality based on neutral net of the present invention, its specific implementation process comprises the content of four aspects: data acquisition, efficiency calculating, neural network learning and energy consumption abnormality detection.
(1) mode of data acquisition
According to Data Source, data involved in the present invention are divided into sensing data and product data, and sensing data obtains primarily of the physical sensors be arranged in workshop and equipment, and product data obtain from the product database server.
Described sensor comprises temperature sensor, humidity sensor in workshop, pressure sensor, capsule temperature sensor and metal pattern temperature sensor in the steam pressure sensor be connected with vulcanizer, steam flow sensor, capsule, as described in Figure 1.Sensing data is sent to data acquisition module in server by building on the wireless network be connected in workshop, with server, and wireless network adopts Zigbee standard agreement, and sample frequency is between 0.5Hz ~ 2Hz.
Described product data comprise the specifications parameter of tire, and (outer diameter D, internal diameter d, width B and thickness A, the difference between outer diameter D and internal diameter d is 2H, as shown in Figures 4 and 5) and performance parameter (hardness, breaking tenacity, percentage elongation, expansion rate, coefficient of elasticity).
(2) mode of efficiency calculating
Efficiency relates to theoretical energy consumption and calculates and actual consumption calculating, using ratio between the two as energy efficiency indexes in calculating.
Described theoretical energy consumption comprises steam energy consumption and metal outer mold energy consumption in capsule, and wherein steam energy consumption is mainly the energy consumption needed at the vulcanization reaction environment of capsule generation HTHP, and this part energy provides primarily of the steam injected in capsule.
Obtaining temperature and pressure data T=T in capsule by sensor 1, T 2, T 3..., T n; P=P 1, P 2, P 3..., P hafter, for making calculating more accurate, first carry out interpolation fitting to data, described interpolation method adopts Lagrange quadratic interpolation formula:
L ( i + x ) = 1 2 ( x - 1 ) ( x - 2 ) T i - x ( x - 2 ) T i + 1 + 1 2 x ( x - 1 ) T i + 2 - - - ( 10 )
The observation T of continuous three times is utilized in above formula i, T i+1, T i+2or P h, P h+1, P h+2interpolation strengthening is carried out to sensor signal, can further the sampling period be shortened, improve the accuracy of data.
After carrying out interpolation processing, the formula (3) in employing step 2 and formula (4) calculate the steam energy consumption Q in capsule respectively theoreticalwith metal outer mold energy consumption W theoretical.
(3) mode of neural network learning
In the present invention, neutral net, by learning the data sample under standard condition, obtains the connection weight in network between each node layer and threshold value.Described standard condition comprises the process of the qualified tire product produced by standard procedure under all size product and various workshop condition.
In described BP neutral net, by the error in parameter learning through successively backpropagation, namely first by output layer, the neuronic output error of each layer is successively calculated, then regulate the weights and threshold of each layer according to error gradient descent method, make the final output of amended network can close to desired value.Described BP nerve network system error is defined as:
E = 1 2 Σ i = 1 n ( J ^ i - J i ) 2 - - - ( 11 )
The correction △ w of output layer weights is revised successively according to error gradient descent method j, the correction △ e of output layer threshold value, the correction △ w of hidden layer weights ij, the correction △ h of hidden layer threshold value j, more new formula is as follows for corresponding improve parameter unification:
α ← α + Δα = α - η ∂ E ∂ α - - - ( 12 )
In described step 8, for ensureing various signal data equal importance, all kinds of input data, at the same order of magnitude, therefore need to carry out classification regularization to the data of three types.
In described step 8, select suitable node in hidden layer m 4seem most important, best node in hidden layer in the present invention turns cross-validation method by wheel and determines, its specific practice is: training data is divided into N group, each wherein N-1 that incites somebody to action, as training data, train described neural network model, remain 1 group as test data, to judge the precision of prediction of the neutral net of training, wheel like this turns N, as the consensus forecast precision of neutral net, finally choose consensus forecast precision maximum time node in hidden layer as optimal node number.
In the formula (8) of described step 8, generally getting a=1.176, b=2/3, thus ensure φ (0) ≈ 0.5, be approximately linear, and the extreme value of second dervative is about ± 2 places between [-1,1].
(4) mode of energy consumption predicting abnormality
Train neutral net under standard condition after, can be applied in the energy consumption abnormality detection in daily tyre vulcanizing, concrete mode is: if actual efficiency and neural network prediction efficiency error are within the scope of reasonable interval:
| e ^ - e | ≤ ϵ - - - ( 13 )
Then think without exception, otherwise then have abnormal generation.
The advantage of the inventive method is as follows:
(1) the present invention sets up tyre vulcanization efficiency computational methods, the method using the ratio of technology theory energy consumption and actual consumption as energy efficiency indexes, with the existing energy valid value calculated by input and output mode, this method is more scientific and reasonable, and be associated with actual process process and technological parameter, energy efficiency indexes can reflect the use energy situation of each link;
(2) adopt neutral net to describe the complex relationship of product parameters, ambient parameter and energy efficiency indexes, be convenient to enterprise and formulate energy use planning, improve enterprise energy and adjust the degree of accuracy;
(3) method being detected the unusual condition in tyre production technology by energy Significant Change is proposed, compared with setting most high energy consumption exceptional value with tradition, this method detects the use energy situation of links and the equipment considering tire vulcanization process, more accurately and reliably, the timely and effective anomalous event of excavating in production of enterprise is helped.The foregoing is only preferred embodiment of the present invention, not in order to limit the present invention, all any amendments done within the spirit and principles in the present invention, equivalent replacement and improvement etc., all should be included within protection scope of the present invention.

Claims (10)

1., based on a tyre vulcanizer energy consumption method for detecting abnormality for neutral net, it is characterized in that: it comprises the steps:
Step one: collection technology data, described process data comprises vapor (steam) temperature in capsule and pressure, and the vapor (steam) temperature collected by the predetermined sampling period and pressure mark by sequence respectively:
T=T 1,T 2,T 3,……,T n, (1);
P=P 1,P 2,P 3,……,P h, (2);
Wherein, T irepresent the capsule temperature in i-th sampling period, P irepresent the capsule pressure in i-th sampling period, 1≤i≤h, h is sampling number, △ T=T 2-T 1for the sampling period, metal outer mold keeps steady temperature by electric-heating-wire-heating, uses T 0mark;
Step 2: calculate theoretical energy consumption,
The heat of steam conduction in (a) capsule and pressure, divide period treatment by vapor (steam) temperature and pressure, the account form within each sampling period is as follows:
Wherein, n is evaporative substance amount, n ≈ 55.56m 1, m 1be reaction vapor quality, R=8.314J/ (Kmol) is gas constant, and c ≈ 4.2KJ/ (Kg DEG C) is the specific heat capacity of steam;
B the heat of () metal outer mold conduction, in sulfidation, the temperature of metal outer mold remains at, and supposes that tire embryo reaches the reaction temperature of specifying by metal outer mold heat transfer, then the heat of its conduction is: W theoretical=cm 2(T 0-25), (4);
Wherein, the specific heat capacity that c ≈ 1.7KJ/ (Kg DEG C) is rubber, m 2for the quality of tire embryo;
Step 3: gather energy consumption data, gathers the actual energy resource consumption in sulfidation, comprises vulcanizer power consumption W 1, compressor power consumption W 2, use quantity of steam m 3;
Step 4: calculate actual consumption, actual consumption is primarily of vulcanizer electric energy, compressor electric energy and the boiler energy consumption composition producing high temperature and high pressure steam, and the gross energy of actual consumption is: W always=W 1+ W 2+ m 32796KJ/Kg, (5);
Wherein, m 32796KJ/Kg is steam contained energy;
Step 5: calculate energy efficiency,
Step 6: gather product parameters, comprises specifications parameter and performance is joined, and described specifications parameter comprises multiple appearance and size parameters of embryo, uses x respectively 1, x 2..., x prepresent, described performance parameter comprises the multiple physical characteristics required by target tire product, uses y respectively 1, y 2..., y qrepresent;
Step 7: gather ambient parameter, it is the ambient parameter relevant to tyre vulcanizing, uses z respectively 1, z 2..., z krepresent;
Step 8: set up neural network prediction, adopt three layers of BP neutral net in the object module chosen, using the specifications parameter of product, performance parameter and ambient parameter as input variable, comprise p+q+k node altogether, intermediate layer is hidden layer, comprises m altogether 4individual node, target output layer comprises individual node, is the predicted value of energy efficiency, and the weight parameter between model interior joint adopts standard historical data to train; At the input layer of described BP neutral net, its input signal is specifications parameter x 1~ x p, performance parameter x p+1~ x p+qwith ambient parameter x p+q+1~ x p+q+k, adopt regular function to convert in input node, be converted into the value y of [-1,1] interval range i: y i = 2 × x i - x min x max - x min - 1 , - - - ( 7 ) ;
Wherein, x maxfor inputting the upper bound of data, x minfor inputting the lower bound of data, y is the value after regularization; At the hidden layer of described BP neutral net, its excitation function adopts hyperbolic sigmoid function z j:
z j = φ ( net 1 ) = a tanh ( b × net 1 ) = a [ 1 - exp ( b × net 1 ) 1 + exp ( b × net 1 ) ] , - - - ( 8 ) ;
Wherein, net 1for the input signal of hidden layer, numerically equal weighted average and this Node B threshold sum of input layer output signal w ijfor the connection weight between input layer i and hidden layer node j, h jfor the threshold value of hidden layer node j, a, b are the parameter of sigmoid function; At the output layer of described neutral net, its excitation function adopts jump function J: J = ψ ( net 2 ) = 1 net 2 > 0 0 net 2 ≤ 0 - - - ( 9 ) ;
Wherein, net 2for the input signal of output layer, numerically equal the weighted average of hidden layer output signal and the difference of this Node B threshold w jfor the connection weight between hidden layer node j and output layer node, e is the threshold value of output layer node;
Step 9: detect energy consumption abnormal, neutral net after having trained utilizing historical standard floor data, by importing new data sample in real time, can analyze the energy consumption of new samples manufacture process, wherein J=0 represents without exception, and contrary J=1 indicates abnormal generation.
2., as claimed in claim 1 based on the tyre vulcanizer energy consumption method for detecting abnormality of neutral net, it is characterized in that: described specifications parameter comprises weight, surface area, thickness three appearance and size parameters of embryo.
3. as claimed in claim 1 based on the tyre vulcanizer energy consumption method for detecting abnormality of neutral net, it is characterized in that: described performance parameter comprises hardness required by target tire product, breaking tenacity, percentage elongation, expansion, elasticity five physical characteristics.
4., as claimed in claim 1 based on the tyre vulcanizer energy consumption method for detecting abnormality of neutral net, it is characterized in that: the ambient parameter relevant to tyre vulcanizing comprises temperature and humidity.
5. as claimed in claim 1 based on the tyre vulcanizer energy consumption method for detecting abnormality of neutral net, it is characterized in that: the vapor (steam) temperature T=T in acquisition capsule 1, T 2, T 3..., T nwith pressure data P=P 1, P 2, P 3..., P hafter, for making calculating more accurate, first carry out interpolation fitting to data, described interpolation method adopts Lagrange quadratic interpolation formula:
L ( i + x ) = 1 2 ( x - 1 ) ( x - 2 ) T i - x ( x - 2 ) T i + 1 + 1 2 x ( x - 1 ) T i + 2 , - - - ( 10 ) ;
Wherein, the observation T of continuous three times is utilized i,t i+1, T i+2or P h,p h+1, P h+2carry out interpolation strengthening to sensor signal, will shorten in the sampling period further, improve the accuracy of data, after carrying out interpolation processing, the formula (3) in employing step 2 and formula (4) calculate the steam energy consumption Q in capsule respectively theoreticalwith metal outer mold energy consumption W theoretical.
6., as claimed in claim 1 based on the tyre vulcanizer energy consumption method for detecting abnormality of neutral net, it is characterized in that: described BP nerve network system error is defined as:
E = 1 2 Σ i = 1 n ( J ^ i - J i ) 2 , - - - ( 11 ) ;
The correction △ w of output layer weights is revised successively according to error gradient descent method j, the correction △ e of output layer threshold value, the correction △ w of hidden layer weights ij, the correction △ h of hidden layer threshold value j, more new formula is as follows for corresponding improve parameter unification:
α ← α + Δα = α - η ∂ E ∂ α , - - - ( 12 ) .
7. as claimed in claim 6 based on the tyre vulcanizer energy consumption method for detecting abnormality of neutral net, it is characterized in that: node in hidden layer turns cross-validation method by wheel and determines: training data is divided into N group, each will wherein N-1 as training data, BP neural network model is trained, remain 1 group as test data, with judge the precision of prediction of training BP neutral net, wheel like this turns the consensus forecast precision of N as neutral net, finally choose consensus forecast precision maximum time node in hidden layer as optimal node number.
8. as claimed in claim 7 based on the tyre vulcanizer energy consumption method for detecting abnormality of neutral net, it is characterized in that: in the formula (8) of described step 8, get a=1.176, b=2/3, thus ensure φ (0) ≈ 0.5, they be approximately linear between [-1,1], and the extreme value of second dervative is at ± 2 places.
9. as claimed in claim 8 based on the tyre vulcanizer energy consumption method for detecting abnormality of neutral net, it is characterized in that: train BP neutral net under standard condition after, be applied in the energy consumption abnormality detection in daily tyre vulcanizing: if actual efficiency and BP neural network prediction efficiency error are within the scope of reasonable interval:
| e ^ - e | ≤ ϵ , - - - ( 13 ) , Then think without exception;
Otherwise then there is abnormal generation.
10., as claimed in claim 1 based on the tyre vulcanizer energy consumption method for detecting abnormality of neutral net, it is characterized in that: sampling period △ T is 1 second.
CN201310261390.3A 2013-06-26 2013-06-26 Neural-network-based energy consumption anomaly detection method of tire vulcanizing machine Active CN103302777B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201310261390.3A CN103302777B (en) 2013-06-26 2013-06-26 Neural-network-based energy consumption anomaly detection method of tire vulcanizing machine

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201310261390.3A CN103302777B (en) 2013-06-26 2013-06-26 Neural-network-based energy consumption anomaly detection method of tire vulcanizing machine

Publications (2)

Publication Number Publication Date
CN103302777A CN103302777A (en) 2013-09-18
CN103302777B true CN103302777B (en) 2015-04-01

Family

ID=49128707

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201310261390.3A Active CN103302777B (en) 2013-06-26 2013-06-26 Neural-network-based energy consumption anomaly detection method of tire vulcanizing machine

Country Status (1)

Country Link
CN (1) CN103302777B (en)

Families Citing this family (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103544544B (en) * 2013-10-29 2016-08-17 广东工业大学 A kind of energy resource consumption Forecasting Methodology and device
CN103729444B (en) * 2013-12-30 2017-07-18 清华大学 The abnormal deviation data examination method of potential relation between a kind of data based on monitoring of equipment
CN103745279B (en) * 2014-01-24 2016-04-27 广东工业大学 A kind of energy consumption method for monitoring abnormality and device
CN103886372B (en) * 2014-04-12 2017-01-11 青岛科技大学 Tire body temperature soft measurement method based on wavelet neural network
CN104376369B (en) * 2014-09-17 2017-10-10 广东工业大学 A kind of tyre vulcanization workshop energy optimization dispatching method based on genetic algorithm
CN104375091A (en) * 2014-11-18 2015-02-25 柳州市金旭节能科技有限公司 Monitoring method for electric vehicle power storage battery
CN104636538B (en) * 2014-12-30 2018-06-19 广东工业大学 The analysis of ball mill particle trajectories and energy consumption modeling method based on DEM
CN107562696B (en) * 2017-09-12 2020-06-05 齐鲁工业大学 On-line detection and control method for tire product quality
CN107992983A (en) * 2017-12-30 2018-05-04 中国移动通信集团江苏有限公司 Energy consumption Forecasting Methodology, system, equipment and storage medium for logistics fair platform
CN109034497A (en) * 2018-08-31 2018-12-18 广东工业大学 Prediction technique, system, medium and the equipment of polycrystalline reduction process energy consumption value
CN109358185A (en) * 2018-10-10 2019-02-19 青岛科技大学 Rubber compounding performance prediction model and prediction technique based on extreme learning machine

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102072922A (en) * 2009-11-25 2011-05-25 东北林业大学 Particle swarm optimization neural network model-based method for detecting moisture content of wood
CN103077267A (en) * 2012-12-28 2013-05-01 电子科技大学 Parameter sound source modeling method based on improved BP (Back Propagation) neural network
CN103136446A (en) * 2013-01-31 2013-06-05 南信大影像技术工程(苏州)有限公司 Back propagation (BP) neural network landslide forecasting method based on optimized input layer

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH0415860A (en) * 1990-05-09 1992-01-21 Kawasaki Steel Corp Method for making learning of neural network efficient
WO2007106942A1 (en) * 2006-03-21 2007-09-27 Commonwealth Scientific And Industrial Research Organisation Analysis of grape quality using neural network

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102072922A (en) * 2009-11-25 2011-05-25 东北林业大学 Particle swarm optimization neural network model-based method for detecting moisture content of wood
CN103077267A (en) * 2012-12-28 2013-05-01 电子科技大学 Parameter sound source modeling method based on improved BP (Back Propagation) neural network
CN103136446A (en) * 2013-01-31 2013-06-05 南信大影像技术工程(苏州)有限公司 Back propagation (BP) neural network landslide forecasting method based on optimized input layer

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
基于BP人工神经网络的橡胶老化预报模型;方庆红 等;《合成材料老化与应用》;20031231(第2期);第27-28页 *

Also Published As

Publication number Publication date
CN103302777A (en) 2013-09-18

Similar Documents

Publication Publication Date Title
Zeng et al. Improved multi-variable grey forecasting model with a dynamic background-value coefficient and its application
Huang et al. A proposed iteration optimization approach integrating backpropagation neural network with genetic algorithm
Cong et al. Applications of ANNs in flow and heat transfer problems in nuclear engineering: a review work
Zhang et al. Machine tool thermal error modeling and prediction by grey neural network
CN103777627B (en) A kind of batch process on-line monitoring method based on a small amount of batch
CN104155423B (en) A kind of landscape water quality method for early warning of changeable weight combination
Wang et al. Analysis of coupled mechanical and thermal behaviors in hot rolling of large rings of titanium alloy using 3D dynamic explicit FEM
CN103926874B (en) The selection optimization method of numerical control machine heat error compensation modeling temperature point combination
CN103472088B (en) Thermal resistance analysis method
CN106352542B (en) A kind of prediction technique of storage-type electric water heater residue bathing time
Shigemori et al. Optimum quality design system for steel products through locally weighted regression model
Ghritlahre et al. Investigation of thermal performance of unidirectional flow porous bed solar air heater using MLP, GRNN, and RBF models of ANN technique
CN103912966B (en) A kind of earth source heat pump refrigeration system optimal control method
Jassar et al. Adaptive neuro-fuzzy based inferential sensor model for estimating the average air temperature in space heating systems
Bartolini et al. Application of artificial neural networks to micro gas turbines
CN103614498B (en) Method for reconstructing three-dimensional temperature field positioned on blast furnace wall and computer monitoring system
CN106056210A (en) PM 2.5 concentration value prediction method based on hybrid neural network
Tian et al. Electric vehicle air conditioning system performance prediction based on artificial neural network
Sun et al. Case study of performance evaluation of ground source heat pump system based on ANN and ANFIS models
CN103823991B (en) Heavy-duty tool thermal error prediction method taking environmental temperature into account
Zhang et al. A transient quasi-3D entire time scale line source model for the fluid and ground temperature prediction of vertical ground heat exchangers (GHEs)
CN105302848A (en) Evaluation value calibration method of equipment intelligent early warning system
CN104700153A (en) PH (potential of hydrogen) value predicting method of BP (back propagation) neutral network based on simulated annealing optimization
CN105400915B (en) A kind of method and system of quantitative assessment top gas flow distribution
CN106249724B (en) A kind of polynary molten steel quality forecast Control Algorithm of blast furnace and system

Legal Events

Date Code Title Description
PB01 Publication
C06 Publication
SE01 Entry into force of request for substantive examination
C10 Entry into substantive examination
GR01 Patent grant
C14 Grant of patent or utility model