CN102645315B - Automatic, fast and accurate detection method for air resistance characteristics of large heat exchanger - Google Patents

Abstract

The invention discloses an automatic, fast and accurate detection method for air resistance characteristics of a large heat exchanger. The existing plate-fin heat exchanger air resistance characteristic measuring and controlling system cannot achieve automatic and fast measuring. The method includes: firstly performing data mining according to historical test data, building a general relationship between designed parameters and initial frequency of the heat exchanger through a support vector machine, obtaining control characteristics of heat exchanger channels through automatic identification of control characteristic parameters of different heat exchanger channels, setting pining and instruments diagram (PID) parameters of an automatic controller by adopting optimum performance indexes, regulating converted air flue volume standard values to a set standard value portion by adopting an incremental PID controller and obtaining the air resistance characteristics of the large heat exchanger by detecting the air resistance value at the moment. The automatic, fast and accurate detection method can achieve automatic and fast measuring of the air resistance characteristics of the large heat exchanger, is high in measuring accuracy and effectively improves production efficiency.

Description

A kind of large heat exchanger vapour lock characteristic detection method quick and precisely automatically

Technical field

The invention belongs to control technology field, relate to data modeling and the process control of industrial process control field.Be mainly concerned with a kind ofly for large-scale plate type finned heat exchanger vapour lock feature measurement process, automatically and heat interchanger air quantity is controlled to the setting airflow value after conversion fast, then obtain the vapour lock characteristic value of heat interchanger by detection.

Background technology

Vapour lock characteristic is the important indicator that characterizes large heat exchanger flowing property, is also one of important indicator of large-scale plate type finned heat exchanger product necessary test before dispatching from the factory.Mainly comprise two aspects: one is for certain heat exchanger channel, in standard temperature, normal pressure and normal flow situation, the pressure loss at these heat exchanger channel two ends; Another one is the friction factor of heat exchanger channel in the case.Due to test result difference under the condition such as different temperatures, pressure, lack comparability.Therefore required standard situation is that a normal atmosphere is depressed, standard zero degrees celsius is the status of criterion, and the air quantity now recording is standard air quantity, and the air quantity recording under other pressure and temperatures need to be converted into standard air quantity.Past general orifice flowmeter that adopts in measuring process, by measuring the vapour lock parameter of heat interchanger in certain air quantity situation, then converts and obtains the vapour lock characteristic of heat interchanger under normal conditions.Adopt above method to there is obvious shortcoming, one is that the measurement accuracy of orifice flowmeter is poor, the control of air quantity adopts the mode of manual control valve in addition, cannot accurately be controlled at the air quantity after conversion, the 3rd is that the measurement of employing orifice plate and not energy-conservation, inaccurate, the follow-up artificial calculated amount of manual control are large, because each heat interchanger test duration is long, work efficiency is lower.

Adopt the air quantity of Frequency Converter Control blower fan, adopt the nozzle measuring system can resolution system power saving, the precision that nozzle is measured air quantity be much higher than orifice flowmeter.Owing to cannot ensureing in test process that current working is in standard condition, need the temperature and pressure of heat exchanger to convert according to the equation of gas state, conversion method is: , here with represent standard temperature and pressure (STP), for actual measurement air quantity, with for observed pressure and temperature, it is the standard actual measurement air quantity after converting.For the purpose of accurately, air quantity is controlled at fast to the standard actual measurement air quantity after conversion in heat exchanger channel performance test process, namely require actual measurement air quantity by converting actual measurement standard air quantity to, then control this actual measurement standard air quantity and make it equal established standards air quantity.Consider efficiency and the energy-conservation requirement of test, measurement mechanism adopts converter technique to carry out air quantity adjusting.

It is that test has intermittence that the different lane testings of heat interchanger also have a feature: test after a passage, before new passage is received to measurement mechanism, it is zero that measurement mechanism air channel volume requires.Because the model of each heat interchanger is different with type, in test process, the original frequency of how to confirm blower fan, and how according to the characteristic of system under test (SUT), adopt suitable control method that air quantity fast and stable is controlled in the setting value corresponding with standard condition, heat exchanger measuring accuracy and rapidity have material impact.Due to the nonlinear characteristic that the hysteresis quality of measuring system, different passage part throttle characteristics are widely different and frequency variation signal changes, cause system to be difficult to realize automatic Quick Measurement.Therefore be necessary to study a kind of method of observing and controlling air quantity fast that can realize in the case automatically, make it to meet the industrial requirements of plate type finned heat exchanger vapour lock characteristic test.

Summary of the invention

The object of this invention is to provide one to the different channel characteristics of various heat exchange device, carry out the method for automatic Quick Measurement vapour lock characteristic.In the method, first carry out data mining according to historical test data, adopt support vector machine to set up the general relationship between design of heat exchanger parameter and original frequency, then obtain the control characteristic of heat exchanger channel by the control characteristic parameter automatic Identification of various heat exchange device passage, adopt the adjust pid parameter of self-actuated controller of optimal performance index, then adopt increment type PID controller that the standard value after air channel volume conversion is controlled to established standards air quantity place, obtain its vapour lock characteristic by the vapour lock value detecting now.

Concrete steps of the present invention are:

Step (1) gathers dissimilar plate type finned heat exchanger design parameter and detected parameters, sets up the real-time data base that comprises design of heat exchanger parameter and detected parameters; Described design of heat exchanger parameter comprises tunnel name, design standards air quantity, design vapour lock, the friction factor of heat interchanger, and detected parameters comprises actual vapour lock, environment temperature, air pressure and the blower fan frequency of heat interchanger.On this basis, based on historical test data, adopt the strong support vector machine integrated modelling approach of generalization ability to set up the relational model between design standards air quantity, design vapour lock, environment temperature and actual vapour lock, actual blower fan frequency, predict that with this under various heat exchange device path setting standard air quantity and design vapour lock, blower fan is in order to reach the required frequency values of this established standards air quantity .Concrete modeling method is as follows:

Input parameter and output parameter for modeling sample can be expressed as , wherein represent the group, as the parameter vector of input data, comprises design standards air quantity, design vapour lock and environment temperature, represent the group, as the parameter vector of output, comprises actual vapour lock and actual blower fan frequency, for sample size.

For algorithm of support vector machine, its kernel function is elected radial basis function as:

for radial basis function, for mapping function, represent the the parameter vector of group conduct input data, , for radial basis function nuclear parameter, establish required objective function and be: , for the actual vapour lock of model output and the predicted value of blower fan frequency, for weight coefficient vector, for intercept, in order to calculate with value.Introduce relaxation factor with , and allow error of fitting to be , with value can be passed through in constraint:

, under condition, minimize:

Obtain, wherein for structural risk minimization function, constant for penalty coefficient, with for parametric variable.This minimization problem is a convex quadratic programming problem, introduces Lagrangian function:

Wherein: >=0, >=0, be Lagrange's multiplier.

At saddle point place, Lagrangian function be about minimal point, be also maximal point, above minimization problem is converted into the maximization problems of asking its dual problem.Lagrangian function at saddle point place be about minimal point,:

Can obtain the dual function of Lagrangian function :

now,

According to Ku En-Plutarch (KKT) conditional theorem, have following formula to set up at saddle point:

From above formula, , with can not be non-zero simultaneously, can obtain:

Can obtain from above formula .

According to above algorithm of support vector machine, the step of support vector machine integrated modelling approach is as follows:

A. original training data initialization weights are , for weight update times, when initializes weights , set iterations .

B. call above algorithm of support vector machine to training sample modeling, obtain a model , calculate the square value of average forecasting error: .

C. upgrade original training data weight: .

D. distribute according to the new weights of original training data, sample at former training set, sampling condition is: , for the weight sampling threshold values of setting, produce the training set of a Sub-SVM.

E. repeating step b～d obtains new model with new sub-training set, until inferior iteration completes.

F. by obtain individual sub-supporting vector machine model carries out integrated, and Model Weight is: , the final integrated model obtaining is: , for the support vector machine integrated model obtaining.

Step (2) is according to the design standards air quantity of real exchanger passage, design vapour lock and ambient temperature conditions, and the model prediction heat interchanger passage that utilizes step (1) to set up reaches the initial value of established standards air quantity leeward unit frequency , there is not strong mechanism relevance owing to being subject to test condition and input and output, blower fan is at original frequency although the air quantity in situation and actual set standard air quantity require to be more or less the same, but cannot meet the demands.Need to again revise and by automatic control, actual air volume be controlled to established standards air quantity place.

The frequency initial value that step (3) obtains according to step (2) , blower fan frequency is adjusted to from 0 80%, be designated as , when frequency reaches 80% and stable operation 10-20s after, will become blower fan frequency adjustment to 100% , and stable 10-20s, actual measurement standard air quantity is at this time designated as , blower fan frequency reaches 80% and actual measurement standard air quantity when stable operation 10-20s be designated as .And in frequency from 80% to 100% and in stable 10-20s process, record actual measurement standard air quantity, actual measurement vapour lock and the blower fan frequency of heat exchanger channel with cycle 0.5-1s.Then, by frequency from 80% to 100% and actual measurement standard air quantity, blower fan frequency that in stable 10-20s process, record obtains deduct respectively with , obtain air quantity changing value, be designated as , and frequency change value, be designated as , wherein with represent respectively the of record individual air quantity changes and frequency change value.

Step (4) flows and controls model feature according to pipeline gas, sets up the transport function between frequency and air quantity.Air quantity control system transport function can be considered as one order inertia and add delay link, therefore the transport function between frequency and standard air quantity can be made as , wherein represent respectively open loop enlargement factor, time constant and time delay, for plural number.Obtain according to step (3) , will give respectively positive initial value, pass through transport function calculate in sample point controlled quentity controlled variable output under effect , then with for target, adopt least-squares algorithm to simulate three parameters in transport function , obtain the concrete transport function of this heat interchanger passage.

Step (5) has been set up in step (4) after the transport function of heat exchanger channel control characteristic, air quantity is adjusted to preset standard air quantity, the parameter of the PID that adjusts in order to ensure fast and stable.Taking IATE integral performance as optimum index, with parameter in PID controller , , for variable, with PID+ as open-loop transfer function, taking closed loop transfer function, as equation of constraint, with , , be on the occasion of being variable bound, adopt nonlinear optimization solution technique to be optimized and solve, obtain optimum PID controller parameter , , value, wherein , , represent respectively ratio, integration and differentiation parameter.

Step (6) obtains original frequency according to step (2) and the frequency that obtains of step (3) , calculate linear relationship parameter approximate between air quantity and frequency, and the modified value of definite original frequency , blower fan frequency adjustment is arrived , and stable second.According to the principle of similitude, frequency exist approximation relation to be with air quantity . with be the linear relationship parameter that need to obtain.The frequency that will obtain in step (3) and under air quantity, bring above relational expression into and obtain parameter with .Obtaining with after, according to relation, order for established standards air quantity, can obtain in established standards air quantity leeward unit frequency forecast value revision value .

Step (7) exists blower fan frequency setting frequency point, equifrequent reaches setting value stable after second.Then adopt increment type PID controller that the actual measurement air quantity of heat exchanger channel is controlled to established standards air quantity place, obtain the vapour lock characteristic of heat exchanger channel under established standards air quantity.The output form of PID is:

；

here three of increment type PID controller parameters , , for obtain obtaining those three parameters in step (5). represent the sampling period, represent the error between setting value and the value of feedback of corresponding step number, , , represent respectively current frequency values, frequency change value and next step frequency values.By increment type PID control, actual measurement airflow value can be controlled to established standards airflow value automatically, control accuracy is in 0.5%.Be now the vapour lock characteristic under established standards air quantity by the vapour lock characteristic measuring, by the comparison between actual measurement vapour lock and design vapour lock, can obtain heat interchanger vapour lock characteristic performance index.

Beneficial effect of the present invention: this method not only replaces in the past manual inspection heat interchanger vapour lock characteristic method completely, has automatic detection, automatically calculates and the high feature of measuring accuracy.The robotization of this method and rapidity are very good, can adapt to different loads and air quantity requirement, in improving measuring accuracy, reducing labor workload, can greatly accelerate heat interchanger test lot number.

Embodiment

A kind of large heat exchanger vapour lock characteristic detection method quick and precisely automatically, specifically implement to adopt following steps:

Step (1) gathers dissimilar plate type finned heat exchanger design parameter and detected parameters, sets up the real-time data base that comprises design of heat exchanger parameter and detected parameters; Described design of heat exchanger parameter comprises tunnel name, design standards air quantity, design vapour lock, the friction factor of heat interchanger, and detected parameters comprises actual vapour lock, environment temperature, air pressure and the blower fan frequency of heat interchanger.On this basis, based on historical test data, adopt the strong support vector machine integrated modelling approach of generalization ability to set up the relational model between design standards air quantity, design vapour lock, environment temperature and actual vapour lock, actual blower fan frequency, predict that with this under various heat exchange device path setting standard air quantity and design vapour lock, blower fan is in order to reach the required frequency values of this established standards air quantity .Concrete modeling method is as follows:

Input parameter and output parameter for modeling sample can be expressed as , wherein represent the group, as the parameter vector of input data, comprises design standards air quantity, design vapour lock and environment temperature, represent the group, as the parameter vector of output, comprises actual vapour lock and actual blower fan frequency, for sample size.

For algorithm of support vector machine, its kernel function is elected radial basis function as:

for radial basis function, for mapping function, represent the the parameter vector of group conduct input data, , for radial basis function nuclear parameter, establish required objective function and be: , for the actual vapour lock of model output and the predicted value of blower fan frequency, for weight coefficient vector, for intercept, in order to calculate with value.Introduce relaxation factor with , and allow error of fitting to be , with value can be passed through in constraint:

, under condition, minimize:

Obtain, wherein for structural risk minimization function, constant for penalty coefficient, with for parametric variable.This minimization problem is a convex quadratic programming problem, introduces Lagrangian function:

Wherein: >=0, >=0, be Lagrange's multiplier.

At saddle point place, Lagrangian function be about minimal point, be also maximal point, above minimization problem is converted into the maximization problems of asking its dual problem.Lagrangian function at saddle point place be about minimal point,:

Can obtain the dual function of Lagrangian function :

now,

According to Ku En-Plutarch (KKT) conditional theorem, have following formula to set up at saddle point:

From above formula, , with can not be non-zero simultaneously, can obtain:

Can obtain from above formula .

According to above algorithm of support vector machine, the step of support vector machine integrated modelling approach is as follows:

A. original training data initialization weights are , for weight update times, when initializes weights , set iterations .

B. call above algorithm of support vector machine to training sample modeling, obtain a model , calculate the square value of average forecasting error: .

C. upgrade original training data weight: .

D. distribute according to the new weights of original training data, sample at former training set, sampling condition is: , for the weight sampling threshold values of setting, produce the training set of a Sub-SVM.

E. repeating step b～d obtains new model with new sub-training set, until inferior iteration completes.

F. by obtain individual sub-supporting vector machine model carries out integrated, and Model Weight is: , the final integrated model obtaining is: , for the support vector machine integrated model obtaining.

Step (2) is according to the design standards air quantity of real exchanger passage, design vapour lock and ambient temperature conditions, and the model prediction heat interchanger passage that utilizes step (1) to set up reaches the initial value of established standards air quantity leeward unit frequency , there is not strong mechanism relevance owing to being subject to test condition and input and output, blower fan is at original frequency although require to be more or less the same with actual set standard air quantity, but cannot meet the demands.Need to again revise and by automatic control, actual air volume be controlled to established standards air quantity place.

The frequency initial value that step (3) obtains according to step (2) , blower fan frequency is adjusted to from 0 80%, be designated as , when frequency reaches 80% and stable operation 10-20s after, frequency converter frequency is adjusted to 100% , and stable 10-20s, standard actual measurement air quantity is at this time designated as , frequency converter frequency reaches 80% and standard actual measurement air quantity when stable operation 10-20s be designated as .And in frequency from 80% to 100% and in stable 10-20s process, record standard actual measurement air quantity, actual measurement vapour lock and the blower fan frequency after heat exchanger channel converts with cycle 0.5-1s.Then, by frequency from 80% to 100% and the standard that in stable 10-20s process, record obtains is surveyed air quantity, blower fan frequency deducts respectively with , obtain air quantity changing value, be designated as , and frequency change value, be designated as , wherein with represent respectively the of record individual air quantity changes and frequency change value.

Step (4) flows and controls model feature according to pipeline gas, sets up the transport function between frequency and air quantity.Air quantity control system transport function can be considered as one order inertia and add delay link, therefore the transport function between frequency and standard air quantity can be made as , wherein represent respectively open loop enlargement factor, time constant and time delay, for plural number.Obtain according to step (3) , will give respectively positive initial value, pass through transport function calculate in sample point controlled quentity controlled variable output under effect , then with for target, adopt least-squares algorithm to simulate three parameters in transport function , obtain the concrete transport function of this heat interchanger passage.

Step (5) has been set up in step (4) after the transport function of heat exchanger channel control characteristic, air quantity is adjusted to preset standard air quantity, the parameter of the PID that adjusts in order to ensure fast and stable.Taking IATE integral performance as optimum index, with parameter in PID controller , , for variable, with PID+ as open-loop transfer function, taking closed loop transfer function, as equation of constraint, with , , be on the occasion of being variable bound, adopt nonlinear optimization solution technique to be optimized and solve, obtain optimum PID controller parameter , , value, wherein , , represent respectively ratio, integration and differentiation parameter.

Step (6) obtains original frequency according to step (2) and the frequency that obtains of step (3) , calculate linear relationship parameter approximate between air quantity and frequency, and the modified value of definite original frequency , blower fan frequency adjustment is arrived , and stable second.According to the principle of similitude, frequency exist approximation relation to be with air quantity . with be the linear relationship parameter that need to obtain.The frequency that will obtain in step (3) and under air quantity, bring above relational expression into and obtain parameter with .Obtaining with after, according to relation, order for established standards air quantity, can obtain in established standards air quantity leeward unit frequency forecast value revision value .

Step (7) exists blower fan frequency setting frequency point, equifrequent reaches setting value stable after second.Then adopt increment type PID controller that the actual measurement air quantity of heat exchanger channel is controlled to established standards air quantity place, obtain the vapour lock characteristic of heat exchanger channel under established standards air quantity.The output form of PID is:

; here three of increment type PID controller parameters , , for obtain obtaining those three parameters in step (5). represent the sampling period, represent the error between setting value and the value of feedback of corresponding step number, , , represent respectively current frequency values, frequency change value and next step frequency values.By increment type PID control, actual measurement airflow value can be controlled to established standards airflow value automatically, control accuracy is in 0.5%.Be now the vapour lock characteristic under established standards air quantity by the vapour lock characteristic measuring, by the comparison between actual measurement vapour lock and design vapour lock, can obtain heat interchanger vapour lock characteristic performance index.

Claims (1)

1. a large heat exchanger vapour lock characteristic detection method quick and precisely automatically, is characterized in that the step of the method comprises:

Step (1) gathers dissimilar design of heat exchanger parameter and detected parameters, sets up the real-time data base that comprises design of heat exchanger parameter and detected parameters;

Described design of heat exchanger parameter comprises tunnel name, design standards air quantity, design vapour lock and the friction factor of heat interchanger; Detected parameters comprises actual vapour lock, environment temperature, air pressure and the blower fan frequency of heat interchanger;

On this basis, based on historical test data, adopt the strong support vector machine integrated modelling approach of generalization ability to set up the relation between design standards air quantity, design vapour lock, environment temperature and actual vapour lock, actual blower fan frequency, predict that with this under various heat exchange device path setting standard air quantity and design vapour lock, blower fan is in order to reach the required frequency values f of this established standards air quantity ₀; Concrete modeling method is as follows:

Input parameter and output parameter for modeling sample are expressed as wherein x _irepresent the parameter vector of i group as input data, comprise design standards air quantity, design vapour lock and environment temperature, y _irepresent the parameter vector of i group as output, comprise actual vapour lock and actual blower fan frequency, N is sample size;

For algorithm of support vector machine, its kernel function is elected radial basis function as:

$(x_i,x_j)=\mathrm{\φ}\left(x_i\right)\·\mathrm{\φ}\left(x_j\right)=\exp \left|\left(\frac{{\left|\right|x_i-x_j\left|\right|}^2}{{2\mathrm{\σ}}^2}\right)\right|$

K (x _i, x _j) be radial basis function, φ (x) is mapping function, x _jrepresent the parameter vector of j group as input data, j=1 ..., N, σ is radial basis function nuclear parameter, establishes required objective function to be: f (x _i)=w φ (x _i)+b, f (x _i) be the actual vapour lock of model output and the predicted value of blower fan frequency, w is weight coefficient vector, b is intercept, in order to calculate w and b value; Introduce relaxation factor ξ _i>=0 He and to allow error of fitting be ε, w and b value are by constraint:

$\left\{\begin{array}{c}{y}_i-w\·\mathrm{\φ}\left(x_i\right)-b\≤\mathrm{\ϵ}+{\mathrm{\ξ}}_i\\ w\·\mathrm{\φ}\left(x_i\right)+b-y_i\≤\mathrm{\ϵ}+{\mathrm{\ξ}}_i^{*}\\ {\mathrm{\ξ}}_i\≥0\\ {\mathrm{\ξ}}_i^{*}\≥0\end{array}\right.i=1,\·\·\·,N,$ Under condition, minimize:

$\min R(w,\mathrm{\ξ},{\mathrm{\ξ}}^{*})=\frac{1}{2}w\·w+c\underset{i=1}{\overset{k}{\mathrm{\Σ}}}\mathrm{\ξ}+{\mathrm{\ξ}}^{*}$

Obtain wherein R (w, ξ, ξ ^*) be structural risk minimization function, constant c>0 is penalty coefficient, ξ and ξ ^*for parametric variable; This minimization problem is a convex quadratic programming problem, introduces Lagrangian function:

$\begin{array}{c}L(w,b,\mathrm{\ξ},{\mathrm{\ξ}}^{*},a,a^{*},\mathrm{\γ},{\mathrm{\γ}}^{*})=\frac{1}{2}w\·w+c\underset{i=1}{\overset{N}{\mathrm{\Σ}}}(\mathrm{\ξ}+{\mathrm{\ξ}}^{*})-\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{a}_i[y_i-({\mathrm{\ξ}}_i+\mathrm{\ϵ}+f\left(x_i\right))]\\ -\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{a}_i^{*}[{\mathrm{\ξ}}_i^{*}+\mathrm{\ϵ}+f\left(x_i\right)-y_i]-\underset{i=1}{\overset{N}{\mathrm{\Σ}}}({\mathrm{\γ}}_i{\mathrm{\ξ}}_i+{\mathrm{\γ}}_i^{*}{\mathrm{\ξ}}_i^{*})\end{array}$

Wherein: for Lagrange's multiplier;

At saddle point place, LagrangianL be about minimal point, be also maximal point, above minimization problem is converted into the maximization problems of asking its dual problem; LagrangianL at saddle point place be about minimal point,:

$\left\{\begin{array}{c}\frac{\∂}{\∂w}L=0\→w=\underset{i=1}{\overset{N}{\mathrm{\Σ}}}({\mathrm{\α}}_i-{\mathrm{\α}}_i^{*})\mathrm{\φ}\left(x_i\right)\\ \frac{\∂}{\∂b}L=0\→\underset{i=1}{\overset{N}{\mathrm{\Σ}}}({\mathrm{\α}}_i-{\mathrm{\α}}_i^{*})=0\\ \frac{\∂}{\∂{\mathrm{\ξ}}_i}L=0\→c-{\mathrm{\α}}_i-{\mathrm{\γ}}_i=0\\ \frac{\∂}{\∂{\mathrm{\ξ}}_i^{*}}L=0\→c-{\mathrm{\α}}_i^{*}-{\mathrm{\γ}}_i^{*}=0\end{array}\right.$

Can obtain the dual function of Lagrangian function

now,

$\begin{array}{c}w=\underset{i=1}{\overset{N}{\mathrm{\Σ}}}({\mathrm{\α}}_i-{\mathrm{\α}}_i^{*})\mathrm{\φ}\left(x_i\right)\\ f\left(x\right)=\underset{i=1}{\overset{N}{\mathrm{\Σ}}}({\mathrm{\α}}_i-{\mathrm{\α}}_i^{*})K(x,x_i)+b\end{array}$

According to Kuhn-Tucker condition theorem, have following formula to set up at saddle point:

$\left\{\begin{array}{c}{\mathrm{\α}}_i[\mathrm{\ϵ}+{\mathrm{\ξ}}_i-y_i+f\left(x_i\right)]=0\\ {\mathrm{\α}}_i^{*}[\mathrm{\ϵ}+{\mathrm{\ξ}}_i+y_i-f\left(x_i\right)]=0\end{array}i=1,\·\·\·,N\right.$

From above formula, α _iwith can not be non-zero simultaneously, can obtain:

$\left\{\begin{array}{c}{\mathrm{\ξ}}_i{\mathrm{\γ}}_i=0\\ {\mathrm{\ξ}}_i^{*}{\mathrm{\γ}}_i^{*}=0\end{array}\right.i=1,\·\·\·,N$

Can obtain b from above formula;

According to above algorithm of support vector machine, the step of support vector machine integrated modelling approach is as follows:

A. original training data initialization weights are jj is weight update times, and jj=1 when initializes weights sets iterations k;

B. call above algorithm of support vector machine to training sample modeling, obtain a model M _jj, calculate M _jjthe square value of average forecasting error:

C. upgrade original training data weight:

D. distribute according to the new weights of original training data, sample at former training set, sampling condition is: β is the weight sampling threshold values of setting, and produces the training set of a Sub-SVM;

E. repeating step b～d obtains new model M _jj+1with new sub-training set, until k iteration completes;

F. the k of acquisition sub-supporting vector machine model is carried out integratedly, Model Weight is: the final integrated model obtaining is: m _finalfor the support vector machine integrated model obtaining;

Step (2) is according to the design standards air quantity of real exchanger passage, design vapour lock and ambient temperature conditions, and the model prediction heat interchanger passage that utilizes step (1) to set up reaches the frequency initial value f of blower fan under established standards air quantity ₀, there is not strong mechanism relevance owing to being subject to test condition and input and output, blower fan is at frequency initial value f ₀although require to be more or less the same with actual set standard air quantity, but cannot meet the demands; Need to again revise and by automatic control, actual air volume be controlled to established standards air quantity place;

The frequency initial value f that step (3) obtains according to step (2) ₀, blower fan frequency is adjusted to f from 0 ₀80%, be designated as f ₁, when frequency reaches f ₀80% and stable operation 10-20s after, frequency converter frequency is adjusted to 100%f ₀, and stable 10-20s, standard actual measurement air quantity is at this time designated as Q _w, frequency converter frequency reaches f ₀80% and standard actual measurement air quantity when stable operation 10-20s be designated as Q _w1; And in frequency from 80%f ₀to 100%f ₀and in stable 10-20s process, record standard actual measurement air quantity, actual measurement vapour lock and the blower fan frequency after heat exchanger channel converts with cycle 0.5-1s; Then, by frequency from 80%f ₀to 100%f ₀and stable 10-20 _sactual measurement air quantity, blower fan frequency that in process, record obtains deduct respectively Q _w1and f ₁; Obtain air quantity changing value, be designated as and frequency change value, be designated as u=[u ₁, u ₂..., u _cc], wherein and u _cccc the air quantity that represents respectively record changes and frequency change value;

Step (4) flows and controls model feature according to pipeline gas, sets up the transport function between frequency and air quantity; Air quantity control system transport function is considered as one order inertia and adds delay link, therefore the transport function between frequency and standard air quantity can be made as wherein K _z, T _t, τ represents respectively open loop enlargement factor, time constant and time delay, s is plural number; The u=[u1 obtaining according to step (3), u ₂..., u _cc], by K _z, T _t, τ gives respectively positive initial value, passes through transport function calculate the output Q=(Q under sample point controlled quentity controlled variable u effect ₁, Q ₂..., Q _cc), then with for target, adopt least-squares algorithm to simulate three parameter K in transport function _z, T _t, τ, obtain the concrete transport function of this heat interchanger passage;

Step (5) has been set up in step (4) after the transport function of heat exchanger channel control characteristic, air quantity is adjusted to preset standard air quantity, the parameter of the PID that adjusts in order to ensure fast and stable; Taking IATE integral performance as optimum index, with parameter K in PID controller _p, T _i, T _dfor variable, with as open-loop transfer function, taking closed loop transfer function, as equation of constraint, with K _p, T _i, T _dbe on the occasion of being variable bound, adopt nonlinear optimization solution technique to be optimized and solve, obtain optimum PID controller parameter K _p, T _i, T _dvalue, wherein K _p, T _i, T _drepresent respectively ratio, integration and differentiation parameter;

Step (6) obtains original frequency f according to step (2) ₀and the frequency f that obtains of step (3) ₁, calculate linear relationship parameter approximate between air quantity and frequency, and the modified value f ' of definite original frequency ₀, blower fan frequency adjustment is arrived to f ' ₀, and stable τ second; According to the principle of similitude, frequency f and air quantity Q _whaving approximation relation is Q _w≈ d ₁f+d ₂; d ₁and d ₂be the linear relationship parameter that need to obtain; The frequency f that will obtain in step (3) ₀and f ₁under air quantity, bring above relational expression into and obtain parameter d ₁and d ₂; Obtaining d ₁and d ₂after, according to f=(Q _w-d ₂)/d ₁relation, make Q _wfor established standards air quantity, can obtain at established standards air quantity leeward unit frequency forecast value revision value f ' ₀;

Step (7) by blower fan frequency setting at f ' ₀frequency point, after equifrequent reaches setting value and stablizes τ second; Then adopt increment type PID controller that the actual measurement air quantity of heat exchanger channel is controlled to established standards air quantity place, obtain the vapour lock characteristic of heat exchanger channel under established standards air quantity; The output form of PID is:

$\mathrm{\Δ}{u}_{\mathrm{kk}}=K_p[e\left(\mathrm{kk}\right)-e(\mathrm{kk}-1)]+\frac{T}{T_i}e\left(\mathrm{kk}\right)+T_d\frac{e\left(\mathrm{kk}\right)-2e(\mathrm{kk}-1)+e(\mathrm{kk}-2)}{T}];$

Uu (kk+1)=uu (kk)+Δ u _kkhere three of increment type PID controller parameter K _p, T _i, T _dfor obtain obtaining those three parameters in step (5); T represents the sampling period, e (kk), e (kk-1), e (kk-2) represent the error between setting value and the value of feedback of corresponding step number, and uu (kk), Δ uu (kk), uu (kk+1) represent respectively current frequency values, frequency change value and next step frequency values; By increment type PID control, actual measurement airflow value can be controlled to established standards airflow value automatically, control accuracy is in 0.5%; Be now the vapour lock characteristic under established standards air quantity by the vapour lock characteristic measuring, by the comparison between actual measurement vapour lock and design vapour lock, can obtain heat interchanger vapour lock characteristic performance index.