CN114934848B - Fuzzy neural network modeling method for optimizing control of combustion performance of diesel engine - Google Patents
Fuzzy neural network modeling method for optimizing control of combustion performance of diesel engine Download PDFInfo
- Publication number
- CN114934848B CN114934848B CN202210686148.XA CN202210686148A CN114934848B CN 114934848 B CN114934848 B CN 114934848B CN 202210686148 A CN202210686148 A CN 202210686148A CN 114934848 B CN114934848 B CN 114934848B
- Authority
- CN
- China
- Prior art keywords
- diesel engine
- data
- rotating speed
- neural network
- model
- 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
Links
Images
Classifications
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F02—COMBUSTION ENGINES; HOT-GAS OR COMBUSTION-PRODUCT ENGINE PLANTS
- F02D—CONTROLLING COMBUSTION ENGINES
- F02D41/00—Electrical control of supply of combustible mixture or its constituents
- F02D41/0002—Controlling intake air
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F02—COMBUSTION ENGINES; HOT-GAS OR COMBUSTION-PRODUCT ENGINE PLANTS
- F02D—CONTROLLING COMBUSTION ENGINES
- F02D41/00—Electrical control of supply of combustible mixture or its constituents
- F02D41/02—Circuit arrangements for generating control signals
- F02D41/14—Introducing closed-loop corrections
- F02D41/1401—Introducing closed-loop corrections characterised by the control or regulation method
- F02D41/1405—Neural network control
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F02—COMBUSTION ENGINES; HOT-GAS OR COMBUSTION-PRODUCT ENGINE PLANTS
- F02D—CONTROLLING COMBUSTION ENGINES
- F02D41/00—Electrical control of supply of combustible mixture or its constituents
- F02D41/30—Controlling fuel injection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/043—Architecture, e.g. interconnection topology based on fuzzy logic, fuzzy membership or fuzzy inference, e.g. adaptive neuro-fuzzy inference systems [ANFIS]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F02—COMBUSTION ENGINES; HOT-GAS OR COMBUSTION-PRODUCT ENGINE PLANTS
- F02D—CONTROLLING COMBUSTION ENGINES
- F02D41/00—Electrical control of supply of combustible mixture or its constituents
- F02D41/0002—Controlling intake air
- F02D2041/0022—Controlling intake air for diesel engines by throttle control
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F02—COMBUSTION ENGINES; HOT-GAS OR COMBUSTION-PRODUCT ENGINE PLANTS
- F02D—CONTROLLING COMBUSTION ENGINES
- F02D41/00—Electrical control of supply of combustible mixture or its constituents
- F02D41/02—Circuit arrangements for generating control signals
- F02D41/14—Introducing closed-loop corrections
- F02D41/1401—Introducing closed-loop corrections characterised by the control or regulation method
- F02D2041/1433—Introducing closed-loop corrections characterised by the control or regulation method using a model or simulation of the system
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Abstract
A fuzzy neural network modeling method for diesel engine combustion performance optimization control belongs to the field of diesel engine combustion process modeling. The method solves the problem that the data in the bench test stage of the diesel engine cannot sample the full-rotating-speed working condition range of the diesel engine, so that the neural network modeling of the diesel engine under the full-operating working condition is difficult. The main technical scheme of the invention is as follows: selecting input parameters and output parameters of a diesel engine simulation model according to a control target for diesel engine combustion performance; secondly, exciting a diesel engine simulation model by using the input parameter data to obtain output parameter data, and taking the input parameter data and the output parameter data as modeling data; establishing a plurality of submodels corresponding to the fixed-value rotating speed by using the obtained modeling data; and step three, weighting the submodels corresponding to the adjacent rotating speeds according to the effectiveness function to obtain a prediction model in the full rotating speed range. The method can be applied to data-driven modeling and combustion performance optimization control of the diesel engine.
Description
Technical Field
The invention belongs to the field of diesel engine combustion process modeling, and particularly relates to a fuzzy neural network modeling method for diesel engine combustion performance optimization control.
Background
In order to predict the diesel engine performance emission, further realize combustion optimization, reduce the times and cost of bench tests, a diesel engine numerical simulation model is developed. The development process can be divided into: the system comprises a zero-dimensional model, a quasi-dimensional model and a multi-dimensional model, wherein the zero-dimensional model regards the inside of a cylinder as a uniform field according to the law of conservation of energy and mass in thermodynamics, but the emission performance index of the diesel engine cannot be accurately predicted according to the result; the quasi-dimensional model divides the combustion chamber into a plurality of regions, and establishes a relation of temperature, concentration and other parameters in the regions, which changes along with time, so that the excessive simplification of the combustion process is avoided, the calculated amount is moderate, and the quasi-dimensional model has certain performance prediction capability and is a model which is applied more at present, such as an oil drop evaporation combustion model; the multidimensional model expresses the in-cylinder combustion process by using a control equation on the basis of the most basic physical quantity and is solved by a numerical method, but the calculation amount is huge.
With the development of intelligent algorithms and the proposal of artificial neural networks, the black box model for learning and predicting complex mechanisms by using a large amount of test data is also widely applied to the field of diesel engine performance prediction. The combustion process of the diesel engine is complex, coupling and other relations exist among a plurality of parameters, and the artificial neural network learns and predicts the complex nonlinear relation among the multi-input and multi-output systems according to a large amount of data, so that the artificial neural network can be well adapted to the combustion process of the diesel engine. In 2014, haditaghavifar et al establish an artificial neural network model for the biodiesel engine by taking parameters such as fuel evaporation capacity, diesel mass percentage and the like as input and taking NOx and Soot as output, and the prediction accuracy is as high as 99%. However, the fitting neural network model has higher requirements on data, the expected data has ergodicity, compactness and compatibility, and the data in the bench test stage of the diesel engine generally cannot sample the full-rotating-speed working condition range of the diesel engine, so that the neural network modeling of the full-operating working condition of the diesel engine is difficult. Therefore, a neural network model modeling mode needs to be improved, a diesel engine combustion process model in a full-rotating-speed working condition range is constructed by using less data, and a good foundation is made for further realizing diesel engine combustion performance optimization.
Disclosure of Invention
The invention aims to solve the problem that the neural network modeling of the full-operation working condition of a diesel engine is difficult due to the fact that data of a bench test stage of the diesel engine cannot sample the full-rotating-speed working condition range of the diesel engine, and provides a fuzzy neural network modeling method for optimizing and controlling the combustion performance of the diesel engine.
The technical scheme adopted by the invention for solving the technical problems is as follows:
a fuzzy neural network modeling method for optimizing and controlling combustion performance of a diesel engine specifically comprises the following steps:
selecting input parameters and output parameters of a diesel engine simulation model according to a control target for diesel engine combustion performance;
determining the data amplitude variation range and action time interval of the input parameters, exciting a diesel engine simulation model (the diesel engine simulation model is a GT-Power simulation model) by using the input parameter data to obtain output parameter data, and taking the input parameter data and the output parameter data as modeling data;
establishing a plurality of submodels corresponding to the fixed-value rotating speed by using the obtained modeling data;
and thirdly, designing an effectiveness function for each sub-model in the second step, and weighting the sub-models corresponding to the adjacent rotating speeds according to the effectiveness functions to obtain a prediction model in the full rotating speed range.
Further, in the first step, according to the target of reducing oil consumption and emission of the diesel engine, selecting the output parameters modeled by the diesel simulation model as the output power, the fuel consumption rate, the NOx generation amount, the air-fuel ratio and the detonation pressure of the diesel engine; and selecting the input parameters of the diesel engine simulation model building as the oil injection time, the opening time of the intake valve, the oil injection pressure and the circulating oil injection quantity according to the relation between the output parameters and the diesel engine fuel injection and air intake adjustable parameters.
The purpose of selecting the input and output parameters is to establish a good model base for optimizing control, further optimize combustion and reduce emission.
Further, the specific process of the second step is as follows:
selecting 4 fixed-value rotating speeds, adjusting the value of the circulating fuel injection quantity up and down at the standard circulating fuel injection quantity corresponding to the current rotating speed propelling working condition and the values of other input parameters under each fixed-value rotating speed, and exciting a diesel simulation model by using input parameter data to obtain output parameter data;
and respectively taking the input parameter data and the output parameter data at each constant rotating speed as modeling data at the corresponding constant rotating speed, and then establishing the sub-models at the corresponding constant rotating speeds by utilizing the modeling data, namely respectively obtaining the sub-models corresponding to the 4 constant rotating speeds.
By adjusting the amplitude variation range of the input parameters, the power of the output power of the diesel engine under different rotating speed working conditions can realize the power coverage of adjacent rotating speed sections. And determining the time interval of the action of each input parameter for acquiring the modeling data according to the response speed of the output parameter to the change of the input parameter.
Further, the input parameter excitation signal of the diesel engine simulation model is a step signal.
Further, the duration of the step signal is 2s.
The duration of each input parameter excitation signal is determined according to the actual condition of the simulation model, and the principle is to sufficiently excite the system, reflect the difference of different signals, and consider the calculation time to ensure higher efficiency. And selecting the time interval when the steady-state data trend of the simulation sampling model is the same and the specific numerical value is close to the steady-state data trend, and selecting 2s as the minimum time interval of the signal by the method.
Further, in the second step, after the modeling data under the corresponding constant value rotating speed is obtained, a sub-model corresponding to each constant value rotating speed is established by using a BP neural network, and the sub-model is used for representing the relation of input and output parameters.
Further, weighting the submodels corresponding to the adjacent rotating speeds according to an effectiveness function to obtain a prediction model in a full rotating speed range; the specific process comprises the following steps:
the invention makes the neural network submodels with different constant rotating speed working conditions perform pairwise weighted interpolation in the adjacent rotating speed intervals, namely in the rotating speed intervals of the neural network submodels with the two adjacent constant rotating speed working conditions, only the two submodels play a role, the rotating speeds of the other submodels are not in the coverage range of the validity function, and the weight is 0.
y＝f a ·h 1 +f b ·h 2
Wherein:
wherein f is a For the output, f, obtained by submodel 1 of the input parameter data b Output obtained by sub-model 2 for input parameter data; h is 1 Is f a The weight of (c); h is a total of 2 Is f b The weight of (c); x is the rotating speed value at the current moment; e is the base number of the natural logarithm;
is the validity function of sub-model 1; c. C 1 The central point of the submodel 1 is the fixed value rotating speed value of the submodel 1; sigma 1 Is the standard deviation of submodel 1.
Further, the method further comprises a fourth step, and the specific process of the fourth step is as follows:
determining the standard rotating speed of the diesel engine simulation model under a new working condition, and obtaining the oil injection time and the circulating oil injection quantity under the new working condition through DOE optimization; respectively establishing new working conditions of different propeller curves corresponding to loads by modifying the rotating speed, the oil injection time and the circulating oil injection quantity;
designing an input parameter excitation signal, and acquiring verification data of the prediction model based on the designed input parameter excitation signal;
and comparing the predicted output of the prediction model with the actual output of the diesel engine simulation model to complete the verification of the prediction model.
Further, the input parameter excitation signal is a given speed change torque signal or a speed change torque signal.
The invention has the beneficial effects that:
the invention provides a fuzzy neural network modeling method for diesel engine combustion performance optimization control, a local sub-model adopts an artificial neural network method to establish a diesel engine performance parameter prediction model, expected diesel engine performance parameters can be obtained quickly according to diesel engine input, and compared with a diesel engine GT-Power simulation model with a slow operation speed, the neural network model is more suitable for a scene needing repeated iterative computation during control method design. The method combining the artificial neural network and the T-S fuzzy theory is provided, the effectiveness function is used for carrying out fuzzy weighted interpolation on the neural network submodels under different rotating speed working conditions to obtain the prediction parameters of the intermediate working conditions, the problems that the neural network model is difficult to model and low in prediction precision under the full rotating speed range are solved, the established neural network model is good in generalization effect, and the combustion performance parameters and the emission of the diesel engine can be accurately predicted.
And the prediction result is compared with the simulation result, the average relative error percentage is small, the prediction data and the simulation data have good consistency, and the established diesel engine combustion performance prediction model has good prediction generalization capability.
Drawings
FIG. 1 is a flow chart of the method of the present invention;
FIG. 2a is a graph of simulation data results for different durations of the excitation signal during input sampling according to the present invention;
FIG. 2b is a graph of simulation data results for different durations of the excitation signal during input sampling according to the present invention;
FIG. 2c is a third graph of simulation data results for different durations of the excitation signal during input sampling according to the present invention;
FIG. 3 is a graph of the input excitation signals of the GT-Power simulation model of the present invention;
FIG. 4 is a schematic diagram of a GT-Power simulation model and Simulink coupled input sampling model of the present invention;
FIG. 5 is a schematic diagram of an effectiveness function of a neural network model under different load conditions under a full-load condition;
FIG. 6 is a schematic diagram of an intelligent prediction model for combustion performance of a diesel engine according to the present invention;
the model was implemented in Simulink;
FIG. 7 is a graph of simulation model excitation signals for a fixed speed torque converter for verifying the data type according to the present invention;
FIG. 8 is a graph of simulation model excitation signals for verifying the type of data as variable speed and variable torque in accordance with the present invention;
FIG. 9a is a first verification result of the combustion parameter prediction accuracy of the all-condition combustion prediction model of the present invention;
FIG. 9b is a second graph illustrating the result of verifying the combustion parameter prediction accuracy of the full condition combustion prediction model according to the present invention;
FIG. 9c is a third graph illustrating the results of the verification of the combustion parameter prediction accuracy of the all-condition combustion prediction model according to the present invention;
FIG. 9d is a fourth verification result of the combustion parameter prediction accuracy of the all-condition combustion prediction model of the present invention;
note: in the attached drawings, the circulating fuel injection quantity, the rotating speed and other diesel engine performance parameters of the diesel engine are normalized by the following formulas:
wherein, the first and the second end of the pipe are connected with each other,
is a value after parameter normalization; x is a radical of a fluorine atom min Is the minimum value of the parameter; x is the number of max Is the maximum value of the parameter.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail below with reference to the accompanying drawings. The invention adopts the following technical scheme: firstly, according to a control target facing the combustion performance of the diesel engine, determining input and output parameters of a model, determining input signals, changing circulating fuel injection quantity according to different working conditions with rotating speed as a characteristic to realize Power coverage between working conditions of adjacent rotating speeds, exciting a GT-Power simulation model to obtain modeling data, respectively establishing a neural network sub-model corresponding to each rotating speed by using the data, combining the sub-models into a full-rotating-speed range model based on a T-S fuzzy principle, and verifying the established full-working-condition intelligent prediction model of the combustion performance of the diesel engine.
FIG. 1 is a flow chart of a modeling method of a combustion process of a diesel engine based on a T-S fuzzy neural network. The method comprises the following specific steps:
1. and determining input and output parameters of the model according to the control target for controlling the combustion performance of the diesel engine. In order to meet the requirements of optimizing combustion and reducing emission, the invention selects the performance parameters of the diesel engine: diesel engine output power, fuel consumption, NOx formation, and combustion process constraint parameters: the air-fuel ratio and the detonation pressure are used as output parameters of the model; and selecting the oil injection timing, the opening time of the intake valve, the oil injection pressure and the circulating oil injection quantity as input parameters.
2.1, power coverage under different working conditions with the rotation speed as the characteristic: in order to optimally control the model obtained by the method of the invention at any given rotating speed of the diesel engine, the power coverage among submodels with adjacent rotating speeds is required to be satisfied for the established submodels corresponding to a plurality of fixed-value rotating speeds. The output power of the diesel engine is changed by adjusting the circulating fuel injection quantity and other input parameters up and down according to the standard fuel injection quantity corresponding to each rotating speed propelling working condition, so that the power is changed in a certain range, and the power coverage of adjacent rotating speed sections is realized by the power under different rotating speed working conditions. And data acquisition is carried out in the GT-Power simulation model. Given that the rated power of a certain marine diesel engine is 1, the corresponding standard rotating speed and the corresponding standard circulating fuel injection amount are A, a respectively, the powers of 75%, 50% and 25% load working conditions are 0.75, 0.5 and 0.25 respectively, the corresponding rotating speed according to a propeller propulsion characteristic curve is B, C, D respectively, and the corresponding standard circulating fuel injection amount is b, c and d respectively.
The rotating speed of the fixed value rotating speed working condition is standard rotating speed A, B, C, D, the fuel injection quantity is changed to change the output power of the fuel injection quantity about the standard circulating fuel injection quantity of each working condition, and finally the power is changed to be [85%,110% ] under the load working condition with the rotating speed A as the characteristic; under the load working condition with the rotating speed B as the characteristic, the power change is [60%,85% ]; under the load condition characterized by the C rotating speed, the power change is [35%,60% ]; under load conditions characterized by D speed, the power change was [10%,35% ].
2.2, determining the type of the simulation model excitation signal and the signal change time interval. The step signal is selected as an excitation signal for model identification, and meanwhile, in order to sufficiently excite the system, the input signal needs to last as long as possible, but in practical simulation, too long excitation leads to too long calculation time, and the obtained data does not greatly help the result accuracy of model identification, so that the excitation signal needs to be selected with proper duration. The step signal durations were chosen to be 1s, 2s respectively and to reach steady state, and the normalized GT-Power simulation results are shown in fig. 2a to 2 c.
As can be seen from fig. 2a to 2c, when the signal duration is 2s, the change trend of the steady-state data value is very close to the change trend of the steady-state data value, and in this case, the excitation effect of different signals on the simulation model can be reflected, and the sampling time can be reduced, so that the duration of the step signal is selected to be 2s.
For example, the excitation signal of the GT-Power model is shown in FIG. 3, and the Simulink model coupled with the GT-Power simulation model is shown in FIG. 4.
3.1, establishing a neural network sub-model under the working condition of constant rotating speed. Dividing 330 data of each constant rotating speed working condition obtained by sampling through a GT-Power simulation model into training data and verification data, randomly extracting 10% as the verification data, and taking the rest as the training data.
All training data are randomly grouped (70%, 15% and 15%) according to a certain proportion before each training and are sequentially used for training, verifying and testing during model training, and Mean Square Error (MSE) and goodness of fit (R) are used 2 And judging the training effect of the neural network. And after sampling is finished, taking 10% of verification data extracted as a group of data completely different from the training data, fitting by using a trained neural network model, comparing the obtained data with the original data for error analysis, and judging the generalization effect of the model by using the average relative error percentage MAPE.
Wherein: y is i For the actual value of the data, f (x) i ) In order to predict the value of the model,
is the average of the actual values of the data.
During training, an MATLAB neural network toolbox is used, a Levenberg-Marquardt algorithm is selected as a training algorithm, and the hidden layer structures of the neural networks under different load working conditions and error analysis indexes thereof are shown in a table 1:
TABLE 1 neural network model for different working conditions
And substituting randomly selected 10% of verification data different from the training data into the established single-working-condition neural network model, and evaluating the model through the average relative error percentage MAPE, wherein the smaller the MAPE is, the stronger the prediction generalization capability of the neural network model is. The average relative error percentage MAPE of the validation data output parameters and the fitted output parameters under each condition is shown in table 2:
TABLE 2 analysis of errors in different load conditions
3.2, after establishing a neural network submodel with a fixed rotating speed, combining the submodel by a Takagi-Sugeno fuzzy neural network principle to enable the local submodel to be combined into a model with a full rotating speed range. The T-S fuzzy neural network is a Takagi-Sugeno fuzzy model and belongs to a hierarchical model. The basic principle is a divide and conquer strategy, where the system space is subdivided into a number of smaller subspaces and identified using a single sub-model. Complex problems are subdivided into simpler problems which are then identified using simpler models.
The output of the fuzzy neural network may be expressed as:
wherein y represents the model prediction output over the full speed range, being the weighted sum of all local sub-models f (·); function of validity
The weight coefficient used for determining weighted submodel uses Gaussian function as validity function
c ij Is the center point, u n For different values of speed of rotation within the interval, σ ij Is standard variance, is based on>
Principle of design 1)
2) Determining based on the speed range of a sub-model>
Medium parameter.
The method of the invention makes the neural network submodels with different constant rotating speed working conditions perform pairwise weighted interpolation in the adjacent rotating speed intervals, namely, only the two submodels play a role in the rotating speed intervals of the neural network submodels with the two adjacent constant rotating speed working conditions, the rotating speeds of the other submodels are not in the coverage range of the validity function, and the weight is 0. Therefore, formula
Can be simplified into:
y＝f a ·h 1 +f b ·h 2
wherein:
wherein f is a Equal to the output obtained by fitting the input data through the neural network submodel 1; f. of b Equal to the output obtained by fitting the input data through the neural network submodel 2; h is a total of 1 Is f a The weight of (c); h is 2 Is f b The weight of (c); c. C 1 The central point of the neural network submodel 1 is the constant value rotating speed value of the neural network submodel 1; sigma 1 Is the standard deviation of the neural network submodel 1.
According to the data characteristics among different constant value rotating speed working conditions, when h is 1 =1, the output y is equal to the fitted output of the neural network submodel 1, when h 1 When =0, the output y is equal to the fitted output of the neural network submodel 2. I.e. within a weighted interval, the left end h of the interval 1 =1, right end point h 1 ＝0。
So, according to the gaussian function, when θ =0,
if x =3 δ:
f(x)＝e -4.5 ≈0
therefore, when the half-width length of the interval is equal to three times the standard deviation, the left end point h of the interval can be realized 1 =1, right end point h 1 And =0. The validity function, i.e., the fuzzy weighting function, between different constant speed conditions is shown in FIG. 5.
The obtained intelligent prediction model of the combustion performance of the diesel engine based on the T-S fuzzy full-rotating-speed working condition range is shown in figure 6, and the neural network sub-models of different constant-value rotating-speed working conditions are subjected to fuzzy weighting through an effectiveness function to finally obtain a fitting output result of the intelligent prediction model.
4. Designing and sampling a new simulation model working condition, and taking data obtained by sampling as verification data of the diesel engine combustion process model in the full rotating speed range and substituting the verification data into the model. And determining the rotating speed of a new working condition according to a propeller characteristic formula of the marine diesel engine, and optimizing the oil injection time and the circulating oil injection quantity through DOE (data object analysis) to obtain optimal parameters.
Wherein: n is a radical of P -the power absorbed by the propeller; k N -a power coefficient; n is o -the rotational speed of the propeller; m-index, m =3.
In GT-Power, the rotation speed, the oil injection time, the circulating oil injection quantity and the boundary conditions (including the air inlet pressure, the exhaust pressure and the exhaust temperature) are modified, new working conditions of loads corresponding to curves of three propellers of 85 percent, 65 percent and 35 percent are established, and excitation signals are designed into two types: one is for a given torque signal to change speed, the excitation is still a step signal with a minimum time interval of 2s, as shown in FIG. 7; the excitation signal is shown in figure 8, different inputs are sampled in a simulation model to obtain simulation output data, the same input is substituted into an intelligent prediction model of the full-rotating-speed working condition to obtain fitting data, and the fitting data and the simulation data are compared to calculate the average relative error of the fitting data and the simulation data. The comparison of the simulation verification data of 85% working conditions and the fitting data of the intelligent prediction model is shown in fig. 9a to 9d, and the specific verification effect MAPE obtained by comparing the simulation verification data of different new working conditions with the fitting data is shown in table 3:
TABLE 3 data verification effect of different working conditions of full-working-condition model
It can be seen from fig. 9a to 9d that the fitting data error of the verification working condition basically meets the precision requirement, the high-power Kuang Nige error is very small, the fitting data prediction is accurate, and the average relative error becomes large due to combustion deterioration and small performance parameters of the low working condition, but the data error is in a reasonable range. The full-working-condition intelligent prediction model can better predict the combustion performance parameters of unknown working conditions and has better prediction generalization capability.
Through the verification of completely unknown working condition data different from training working conditions, the intelligent prediction model for the full-speed working condition combustion of the diesel engine, namely the fuzzy neural network model for the combustion optimization control of the diesel engine, has good prediction generalization capability.
The above-described calculation examples of the present invention are merely to explain the calculation model and the calculation flow of the present invention in detail, and are not intended to limit the embodiments of the present invention. It will be apparent to those skilled in the art that other variations and modifications of the present invention can be made based on the above description, and it is not intended to be exhaustive of all embodiments, and obvious variations and modifications can be made without departing from the scope of the invention.
Claims (7)
1. A fuzzy neural network modeling method for optimizing and controlling combustion performance of a diesel engine is characterized by comprising the following steps:
selecting input parameters and output parameters of a diesel engine simulation model according to a diesel engine combustion performance control target;
determining the data amplitude variation range and action time interval of the input parameters, exciting a diesel engine simulation model by using the input parameter data to obtain output parameter data, and taking the input parameter data and the output parameter data as modeling data;
establishing a plurality of submodels corresponding to the fixed-value rotating speed by using the obtained modeling data;
selecting 4 fixed value rotating speeds, adjusting the value of the circulating fuel injection quantity up and down at the standard circulating fuel injection quantity corresponding to the current rotating speed propelling working condition and the values of other input parameters under each fixed value rotating speed, and exciting a diesel engine simulation model by using input parameter data to obtain output parameter data;
respectively taking the input parameter data and the output parameter data at each constant value rotating speed as modeling data at corresponding constant value rotating speeds, and establishing sub-models at corresponding constant value rotating speeds by utilizing the modeling data, namely respectively obtaining 4 sub-models corresponding to the constant value rotating speeds;
thirdly, designing an effectiveness function for each sub-model in the second step, and weighting the sub-models corresponding to the adjacent rotating speeds according to the effectiveness functions to obtain a prediction model in the full rotating speed range;
the method comprises the following steps of enabling the neural network submodels under different constant-value rotating speed working conditions to carry out pairwise weighted interpolation in adjacent rotating speed intervals, wherein the specific process is as follows:
y＝f a ·h 1 +f b ·h 2
wherein:
wherein, f a For the output of the input parameter data obtained via submodel 1, f b Output obtained by sub-model 2 for input parameter data; h is 1 Is f a The weight of (c); h is 2 Is f b The weight of (c); x is the rotating speed value at the current moment; e is the base number of the natural logarithm;
is the validity function of sub-model 1; c. C 1 The central point of the submodel 1 is the fixed value rotating speed value of the submodel 1; sigma 1 For the standard deviation of the submodel 1, the validity function->
Used to determine the weight coefficients of the weighted submodels.
2. The modeling method of the fuzzy neural network for optimizing and controlling the combustion performance of the diesel engine as claimed in claim 1, wherein in the first step, the output parameters of the diesel engine simulation model are selected to be the diesel engine output power, the fuel consumption rate, the NOx generation amount, the air-fuel ratio and the detonation pressure; the input parameters of the diesel engine simulation model are selected as oil injection time, opening time of an inlet valve, oil injection pressure and circulating oil injection quantity.
3. The modeling method of the fuzzy neural network for optimizing and controlling the combustion performance of the diesel engine as claimed in claim 1, wherein the excitation signal of the input parameter of the diesel engine simulation model is a step signal.
4. The modeling method of the fuzzy neural network for optimizing and controlling the combustion performance of the diesel engine as claimed in claim 3, wherein the duration of the step signal is 2s.
5. The modeling method of the fuzzy neural network for optimizing and controlling the combustion performance of the diesel engine as claimed in claim 4, wherein in the second step, after the modeling data under the corresponding constant rotating speed are obtained, the BP neural network is used for establishing the submodel corresponding to each constant rotating speed, and the submodel is used for representing the relation of the input and output parameters.
6. The modeling method of the fuzzy neural network for optimizing and controlling the combustion performance of the diesel engine as claimed in claim 1, wherein the method further comprises a fourth step, and the specific process of the fourth step is as follows:
determining the standard rotating speed of the diesel engine simulation model under a new working condition, and obtaining the oil injection time and the circulating oil injection quantity under the new working condition through DOE optimization; respectively establishing new working conditions of different propeller curves corresponding to loads by modifying the rotating speed, the oil injection time and the circulating oil injection quantity;
designing an input parameter excitation signal, and acquiring verification data of the prediction model based on the designed input parameter excitation signal;
and comparing the predicted output of the prediction model with the actual output of the diesel engine simulation model to finish the verification of the prediction model.
7. The modeling method of the fuzzy neural network for optimizing and controlling the combustion performance of the diesel engine as claimed in claim 6, wherein the input parameter excitation signal is a given speed change torque signal or a speed change torque signal.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210686148.XA CN114934848B (en) | 2022-06-16 | 2022-06-16 | Fuzzy neural network modeling method for optimizing control of combustion performance of diesel engine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210686148.XA CN114934848B (en) | 2022-06-16 | 2022-06-16 | Fuzzy neural network modeling method for optimizing control of combustion performance of diesel engine |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114934848A CN114934848A (en) | 2022-08-23 |
| CN114934848B true CN114934848B (en) | 2023-04-07 |
Family
ID=82869322
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210686148.XA Active CN114934848B (en) | 2022-06-16 | 2022-06-16 | Fuzzy neural network modeling method for optimizing control of combustion performance of diesel engine |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114934848B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117236086B (en) * | 2023-11-16 | 2024-01-19 | 北京理工大学 | Prediction method for highest reinforcement degree of diesel engines with different cylinder diameters |
Family Cites Families (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6941289B2 (en) * | 2001-04-06 | 2005-09-06 | Sas Institute Inc. | Hybrid neural network generation system and method |
| DE102010028266A1 (en) * | 2010-04-27 | 2011-10-27 | Robert Bosch Gmbh | Control device and method for calculating an output for a controller |
| CN107269408B (en) * | 2017-05-15 | 2022-08-05 | 吉林大学 | Diesel engine optimized combustion controller and simulation model control method |
| US10634081B2 (en) * | 2018-02-05 | 2020-04-28 | Toyota Jidosha Kabushiki Kaisha | Control device of internal combustion engine |
| JP7231144B2 (en) * | 2019-07-17 | 2023-03-01 | 株式会社トランストロン | Engine control device and neural network program it has |
| JP2021032116A (en) * | 2019-08-22 | 2021-03-01 | トヨタ自動車株式会社 | Vehicular control device, vehicular learning system, and vehicular learning device |
| JP7206407B2 (en) * | 2019-09-26 | 2023-01-17 | 日立Astemo株式会社 | Control device for internal combustion engine |
| CN113804446A (en) * | 2020-06-11 | 2021-12-17 | 卓品智能科技无锡有限公司 | Diesel engine performance prediction method based on convolutional neural network |
| CN112241609A (en) * | 2020-10-15 | 2021-01-19 | 吉林大学 | Real-time estimating system for NOx emission of diesel engine |
| CN112597658B (en) * | 2020-12-28 | 2022-02-18 | 哈尔滨工程大学 | Multi-model fault diagnosis method for marine diesel engine based on working condition identification |
| CN113554153A (en) * | 2021-07-23 | 2021-10-26 | 潍柴动力股份有限公司 | Method and device for predicting emission of nitrogen oxides, computer equipment and medium |
| CN113935125A (en) * | 2021-09-09 | 2022-01-14 | 西华大学 | BP neural network prediction model optimization method for diesel engine emission performance |
-
2022
- 2022-06-16 CN CN202210686148.XA patent/CN114934848B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN114934848A (en) | 2022-08-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zareei et al. | Optimization and study of performance parameters in an engine fueled with hydrogen | |
| CN113572206B (en) | Wind power output interval prediction method | |
| CN104102781A (en) | Multi-objective optimization based engine performance optimizing method | |
| CN106066606B (en) | Hydrogen internal combustion engine ignition timing calibration optimization system and its optimization method based on L-M neural network | |
| Nikzadfar et al. | Investigating a new model-based calibration procedure for optimizing the emissions and performance of a turbocharged diesel engine | |
| Rai et al. | Prediction models for performance and emissions of a dual fuel CI engine using ANFIS | |
| CN114934848B (en) | Fuzzy neural network modeling method for optimizing control of combustion performance of diesel engine | |
| CN110837223A (en) | Combustion optimization control method and system for gas turbine | |
| Alcan et al. | Estimating soot emission in diesel engines using gated recurrent unit networks | |
| CN109344492A (en) | A kind of aeroengine thrust estimation method based on K- mean cluster Yu population core extreme learning machine | |
| CN105116730A (en) | Hydrogen fuel engine ignition advance angle optimization system based on particle swarm fuzzy neural network, and optimization method thereof | |
| CN112182954A (en) | Fluid simulation data prediction model based on LSTM | |
| Wang et al. | Combining GA-SVM and NSGA-Ⅲ multi-objective optimization to reduce the emission and fuel consumption of high-pressure common-rail diesel engine | |
| Martínez-Morales et al. | Modeling engine fuel consumption and NOx with RBF neural network and MOPSO algorithm | |
| CN107269408B (en) | Diesel engine optimized combustion controller and simulation model control method | |
| CN108563844A (en) | A kind of supercharged diesel engine Optimum EGR rate determines method | |
| CN112782974B (en) | Control parameter calibration method | |
| CN114861126A (en) | Method and system for predicting stability in combustion adjustment process of gas turbine | |
| CN114370348A (en) | Control parameter setting method for engine rotating speed control system | |
| Liu et al. | Study on prediction model of diesel engine with regulated two-stage turbocharging system based on hybrid genetic algorithm-particle swarm optimization method at different altitudes | |
| CN113807024B (en) | Gas turbine optimal dynamic working point selection method based on proxy model | |
| CN113392574A (en) | Gasoline engine secondary charging model air inflow estimation method based on neural network model | |
| Jian et al. | Application of Improved Grey-Least Squares Combined Forecasting Model in Gasoline Engine Intake Air Mass Flow | |
| Yongbo et al. | Transient air-fuel ratio control of direct injection gasoline engine based on LADRC algorithm | |
| Sui et al. | Modelling and simulation research of vehicle engines based on computational intelligence methods |
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 |