CN106401757A - Cylinder shutting down mode implementing method and system of engine, and vehicle - Google Patents
Cylinder shutting down mode implementing method and system of engine, and vehicle Download PDFInfo
- Publication number
- CN106401757A CN106401757A CN201510450509.0A CN201510450509A CN106401757A CN 106401757 A CN106401757 A CN 106401757A CN 201510450509 A CN201510450509 A CN 201510450509A CN 106401757 A CN106401757 A CN 106401757A
- Authority
- CN
- China
- Prior art keywords
- cylinder
- neural network
- engine
- wavelet neural
- layer
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 46
- 238000013528 artificial neural network Methods 0.000 claims abstract description 103
- 238000012549 training Methods 0.000 claims abstract description 76
- 230000008569 process Effects 0.000 claims abstract description 19
- 238000002347 injection Methods 0.000 claims abstract description 15
- 239000007924 injection Substances 0.000 claims abstract description 15
- 238000012360 testing method Methods 0.000 claims abstract description 13
- 230000009849 deactivation Effects 0.000 claims description 36
- 230000006870 function Effects 0.000 claims description 20
- 238000012937 correction Methods 0.000 claims description 12
- 238000012545 processing Methods 0.000 claims description 8
- 238000010606 normalization Methods 0.000 claims description 7
- 230000000644 propagated effect Effects 0.000 claims description 6
- 239000002826 coolant Substances 0.000 claims description 5
- 239000000446 fuel Substances 0.000 abstract description 18
- 230000000694 effects Effects 0.000 abstract description 10
- 238000004458 analytical method Methods 0.000 description 12
- 239000003921 oil Substances 0.000 description 8
- 238000010586 diagram Methods 0.000 description 5
- 210000004027 cell Anatomy 0.000 description 4
- 210000002569 neuron Anatomy 0.000 description 4
- 230000009466 transformation Effects 0.000 description 4
- 230000001133 acceleration Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000005520 cutting process Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000002485 combustion reaction Methods 0.000 description 1
- 239000000110 cooling liquid Substances 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000000295 fuel oil Substances 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005086 pumping Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000010792 warming Methods 0.000 description 1
Landscapes
- Combined Controls Of Internal Combustion Engines (AREA)
Abstract
The invention provides a cylinder shutting down mode implementing method and system for an engine, and a vehicle. The method includes: selecting a plurality of engine rotation speeds, and performing load characteristic test in a plurality of preset cylinder shutting down modes at each engine rotation speed so as to acquire a plurality of groups of training samples; training a wavelet neural network according to the training samples; inputting running parameters of a vehicle into the trained wavelet neural network so as to determine the cylinder shutting down number and a cylinder shutting down sequence when it is determined that the vehicle enters a cylinder shutting down mode, generating a corresponding cylinder shutting down mode signal, and acquiring the air intake flow and the fuel injection amount of working cylinders according to the cylinder shutting down mode signal; and controlling a cylinder shutting down process of the engine according to the cylinder shutting down mode signal, and the air intake flow and the fuel injection amount of the working cylinders. The cylinder shutting down mode implementing method of the engine can ensure a cylinder shutting down effect (i.e., fuel economy, can improve the uniformity of the cylinders and the stability of working, can improve the reliability of the whole vehicle, and can improve the comfort of a driver.
Description
Technical Field
The invention relates to the technical field of automobiles, in particular to a method and a system for realizing a cylinder-breaking mode of an engine and a vehicle.
Background
The engine works in a surface working condition interval, the rotating speed and the load range of the engine are wide, the fuel economy under low load rate is poor, and along with the stricter fuel consumption regulation, the requirement of reducing the fuel consumption of the engine with a plurality of cylinders and large displacement is more urgent. The cylinder-breaking technology can close one or more cylinders when the engine is partially loaded, and in order to ensure that the power of the engine is unchanged, the load rate of the working cylinder needs to be increased, so that the mechanical efficiency of the engine is improved, the pumping loss is reduced, and the fuel economy is improved.
The key of cylinder deactivation control is when to deactivate the cylinder, which cylinders to close, and the cylinder deactivation time and which cylinders to close will directly affect the stability of the engine operation, the comfort of passengers, the uniformity of the wear of each cylinder, the reliability of the whole engine, and also affect the improvement effect of fuel economy.
The common realization mode of the existing cylinder-breaking mode is that the work of certain specific cylinders is stopped under the partial load working condition of an engine, the cylinder-breaking time and the selection of a working cylinder and a stopping cylinder cannot be accurately adjusted according to the change of the working condition, and the economical efficiency of cylinder breaking is further influenced; in addition, the working cylinder and the stopping cylinder are usually fixed, which easily causes the non-uniformity of the work among the cylinders and influences the reliability of the whole machine.
Disclosure of Invention
In view of this, the present invention is directed to a method for implementing a cylinder deactivation mode of an engine, which improves uniformity among cylinders and stability of operation, improves reliability of a finished vehicle, and improves comfort of a driver while ensuring a cylinder deactivation effect (i.e., fuel economy).
In order to achieve the purpose, the technical scheme of the invention is realized as follows:
a method for realizing a cylinder-failure mode of an engine comprises the following steps: selecting a plurality of engine rotating speeds, and performing load characteristic tests in a plurality of preset cylinder-breaking modes at each engine rotating speed to obtain a plurality of groups of training samples, wherein the training samples comprise the engine rotating speeds, output torques and torque fluctuation conditions along with the crankshaft rotating angle; training a preset wavelet neural network according to the training sample until a preset condition is met to obtain a trained wavelet neural network; acquiring running parameters of a vehicle, and judging whether to enter a cylinder-breaking mode according to the running parameters; if so, inputting the running parameters of the vehicle into the trained wavelet neural network to determine the cylinder failure number and the cylinder failure sequence according to the running parameters of the vehicle, generating corresponding cylinder failure mode signals, and obtaining the air intake flow and the oil injection flow of the working cylinder according to the cylinder failure mode signals; and controlling the cylinder cutting process of the engine according to the cylinder cutting mode signal, the air intake flow and the oil injection quantity of the working cylinder.
Further, the wavelet neural network includes an input layer, a hidden layer and an output layer, where the input layer has n input nodes, the output layer has m output nodes, n and m are positive integers, and the number of nodes in the hidden layer is obtained by the following formula:
wherein k is the number of nodes of the hidden layer, and β is any constant between 1 and 10.
Further, the training of the preset wavelet neural network according to the training sample until a predetermined condition is met to obtain the trained wavelet neural network specifically includes:
normalizing the training samples, wherein the training samples are normalized through a sigmoid function, and the sigmoid function is as follows:
wherein c is a coefficient of the sigmoid function;
inputting a group of training samples after normalization processing into the wavelet neural network for forward propagation;
acquiring an actual torque fluctuation difference value when the full-cylinder mode is switched to the cylinder-off mode, and comparing the actual torque fluctuation difference value with an expected torque fluctuation difference value;
when the error between the actual torque fluctuation difference value and the expected torque fluctuation difference value is larger than a first error threshold value, the error is reversely propagated through the wavelet neural network so as to correct the connection weight value between the input layer and the hidden layer in the wavelet neural network and the connection threshold value of the hidden layer until the error between the actual torque fluctuation difference value and the expected torque fluctuation difference value is smaller than the first error threshold value, and the trained wavelet neural network is obtained.
Further, from the output layer as a start to the input layer, by adjusting the connection weight and connection threshold of the node of each layer, error back propagation is performed, wherein the connection weight between the input layer and the hidden layer in the wavelet neural network is modified by the following formula:
wherein, the η is the learning rate of weight correction, 0<η<1, said wijThe connection weight from the ith input layer to the jth hidden layer;
correcting the connection threshold of the hidden layer in the wavelet neural network through the following formula:
wherein μ is a learning rate of threshold correction, 0<μ<1, said thetajIs the connection threshold of the jth hidden layer cell;
correcting the connection weight between the hidden layer and the output layer in the wavelet neural network through the following formula:
wherein, v isjiAnd connecting the ith hidden layer to the jth output layer.
Further, the operating parameters of the vehicle include: an engine speed signal, an engine torque signal, an accelerator pedal signal, a gear signal, a vehicle speed signal, and a coolant temperature signal.
Compared with the prior art, the method for realizing the cylinder-cut-off mode of the engine has the following advantages:
the method for realizing the cylinder-breaking mode of the engine combines wavelet analysis and artificial neural network control, can accurately control the cylinder-breaking system of the engine in real time, has the self-adaption and self-learning capabilities of the neural network, can overcome the uncertainty and the time-varying property of a control object, trains the network by using less test data as samples, and can enable the neural network to have higher learning and convergence speed by utilizing the wavelet analysis and improve the real-time performance of control. The method improves the uniformity among cylinders and the working stability, improves the reliability of the whole vehicle and improves the comfort of a driver while ensuring the cylinder-breaking effect (namely fuel economy).
The invention also aims to provide a cylinder-failure mode realization system of the engine, which improves the uniformity among cylinders and the working stability, improves the reliability of a whole vehicle and improves the comfort of a driver while ensuring the cylinder-failure effect (namely fuel economy).
In order to achieve the purpose, the technical scheme of the invention is realized as follows:
a cylinder deactivation mode implementing system of an engine, comprising: the training sample acquisition module is used for selecting a plurality of engine rotating speeds and performing load characteristic tests in a plurality of preset cylinder failure modes at each engine rotating speed to obtain a plurality of groups of training samples, wherein the training samples comprise the engine rotating speeds, output torques and torque fluctuation conditions along with the rotating angles of the crankshafts; the training module is used for training a preset wavelet neural network according to the training sample until a preset condition is met to obtain a trained wavelet neural network; the judging module is used for acquiring the running parameters of the vehicle and judging whether to enter a cylinder-failure mode according to the running parameters; and the control module is used for inputting the running parameters of the vehicle into the trained wavelet neural network when the judgment module judges that the vehicle enters the cylinder failure mode, determining the cylinder failure number and the cylinder failure sequence according to the running parameters of the vehicle, generating corresponding cylinder failure mode signals, obtaining the air intake flow and the oil injection flow of a working cylinder according to the cylinder failure mode signals, and controlling the cylinder failure process of the engine according to the cylinder failure mode signals, the air intake flow and the oil injection flow of the working cylinder.
Further, the wavelet neural network includes an input layer, a hidden layer and an output layer, where the input layer has n input nodes, the output layer has m output nodes, n and m are positive integers, and the number of nodes in the hidden layer is obtained by the following formula:
wherein k is the number of nodes of the hidden layer, and β is any constant between 1 and 10.
Further, the training module is configured to:
normalizing the training samples, wherein the training samples are normalized through a sigmoid function, and the sigmoid function is as follows:
wherein c is a coefficient of the sigmoid function;
inputting a group of training samples after normalization processing into the wavelet neural network for forward propagation;
acquiring an actual torque fluctuation difference value when the full-cylinder mode is switched to the cylinder-off mode, and comparing the actual torque fluctuation difference value with an expected torque fluctuation difference value;
when the error between the actual torque fluctuation difference value and the expected torque fluctuation difference value is larger than a first error threshold value, the error is reversely propagated through the wavelet neural network so as to correct the connection weight value between the input layer and the hidden layer in the wavelet neural network and the connection threshold value of the hidden layer until the error between the actual torque fluctuation difference value and the expected torque fluctuation difference value is smaller than the first error threshold value, and the trained wavelet neural network is obtained.
Further, the training module performs error back propagation by adjusting a connection weight and a connection threshold of a node of each layer from the output layer as a start to the input layer, wherein the connection weight between the input layer and the hidden layer in the wavelet neural network is modified by a formula:
wherein, the η is the learning rate of weight correction, 0<η<1, said wijThe connection weight from the ith input layer to the jth hidden layer;
correcting the connection threshold of the hidden layer in the wavelet neural network through the following formula:
wherein μ is a learning rate of threshold correction, 0<μ<1, said thetajIs the connection threshold of the jth hidden layer cell;
correcting the connection weight between the hidden layer and the output layer in the wavelet neural network through the following formula:
wherein, v isjiAnd connecting the ith hidden layer to the jth output layer.
Compared with the prior art, the cylinder-cut-off mode implementation system of the engine and the cylinder-cut-off mode implementation method of the engine have the same advantages, and are not described again.
Still another object of the present invention is to provide a vehicle, which improves the uniformity among cylinders and the stability of operation, improves the reliability of the entire vehicle, and improves the comfort of the driver while ensuring the cylinder-cutoff effect (i.e., fuel economy).
In order to achieve the purpose, the technical scheme of the invention is realized as follows:
a vehicle is provided with the cylinder deactivation mode realization system of the engine as described in the above embodiment.
Compared with the prior art, the vehicle and the engine cylinder-failure mode realization system have the same advantages, and the detailed description is omitted.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate an embodiment of the invention and, together with the description, serve to explain the invention and not to limit the invention. In the drawings:
FIG. 1 is a flow chart of a method for implementing a cylinder deactivation mode of an engine according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a neural network in a method for implementing a cylinder deactivation mode of an engine according to an embodiment of the present invention;
FIG. 3 is a schematic diagram illustrating a neural network self-learning process in a method for implementing a cylinder deactivation mode of an engine according to an embodiment of the present invention;
FIG. 4 is a schematic diagram illustrating a control of a cylinder deactivation mode in a method for implementing a cylinder deactivation mode of an engine according to an embodiment of the present invention;
fig. 5 is a block diagram showing a cylinder deactivation mode implementing system of an engine according to an embodiment of the present invention.
Description of reference numerals:
500-a cylinder-failure mode implementation system of an engine, 510-a training sample acquisition module, 520-a training module, 530-a judgment module and 540-a control module.
Detailed Description
It should be noted that the embodiments and features of the embodiments may be combined with each other without conflict.
The present invention will be described in detail below with reference to the embodiments with reference to the attached drawings.
FIG. 1 is a flow chart of a method of implementing a cylinder deactivation mode of an engine according to one embodiment of the present invention. As shown in fig. 1, a method for implementing a cylinder deactivation mode of an engine according to an embodiment of the present invention includes the steps of:
step S101: selecting a plurality of engine rotating speeds, and performing load characteristic tests in a plurality of preset cylinder-breaking modes at each engine rotating speed to obtain a plurality of groups of training samples, wherein the training samples comprise the engine rotating speeds, output torques and torque fluctuation conditions along with the crankshaft rotating angle.
Specifically, a large number of unknown operating conditions may be mapped by collecting a small amount of test data. For example: selecting p different engine rotating speeds, carrying out load characteristic tests on each rotating speed in q preset cylinder-breaking modes, and collecting p × q samples as learning samples. And recording corresponding engine speed, output torque and torque fluctuation conditions along with the crankshaft angle for the p × q samples, and storing the data into an engine control unit as a training sample of the wavelet neural network. It should be noted that, the more the actual measurement data is, the more the sample library is full, the wider the coverage condition is, and the better the training effect of the wavelet neural network in the later stage is.
Step S102: and training the preset wavelet neural network according to the training sample until a preset condition is met to obtain the trained wavelet neural network.
As shown in fig. 2, in one embodiment of the present invention, the wavelet neural network comprises an input layer, a hidden layer and an output layer, wherein the input layer has n input nodes, the output layer has m output nodes, n and m are positive integers, and the number of nodes of the hidden layer is obtained by the following formula:
where k is the number of nodes in the hidden layer, and β is any constant between 1 and 10.
Specifically, the input parameters of the input layer are corresponding sensor signals after wavelet transformation, and the sensor signals comprise engine speed, torque, gearbox gear, vehicle speed and the like. The hidden layer is an internal information processing layer that adjusts the corresponding coefficients by forward propagation of information and back propagation of errors to achieve the goal that the actual output matches the desired output. The output parameters of the output layer are the cylinder-breaking number and the cylinder-breaking sequence, and the control target is to improve the fuel economy and improve the uniformity of power output before and after switching between the full-cylinder mode and the cylinder-breaking mode. And establishing a functional relation between input parameters and output parameters by using the training samples as learning samples of the wavelet neural network.
Specifically, the input pattern vector is,
Xa＝(x1,x2,...,xn),a＝1,2,...,n；
the corresponding output pattern vector is that of,
Yb＝(y1,y2,...,ym),b＝1,2,...,m；
the neural network structure is a non-linear mapping of the n inputs and m outputs as shown in fig. 2.
In addition, according to the principle of neuron model, there are:
wherein s isjIs the input of the jth hidden layer, wijIs the firstConnection weights, θ, of the i input layers to the jth hidden layerjIs the connection threshold for the jth hidden layer. Obtaining the input of each unit of the output layer according to the thought of mode forward propagation,
wherein ljIs the input of the jth output layer, vjiIs the connection weight, γ, of the ith hidden layer to the jth output layerjIs the connection threshold for the jth output layer. The neural network (i.e. wavelet neural network) realizes the control process by adjusting the connection weight and the connection threshold in the above two formulas.
After the structure of the wavelet neural network according to the embodiment of the present invention is described, the training process of the wavelet neural network is described in detail below.
Specifically, training a preset wavelet neural network according to a training sample until a predetermined condition is met to obtain the trained wavelet neural network, specifically comprising:
1. carrying out normalization processing on the training samples, wherein the training samples are normalized through a sigmoid function, and the sigmoid function is as follows:
wherein c is the coefficient of the sigmoid function.
2. And inputting a group of training samples subjected to normalization processing into a wavelet neural network for forward propagation.
3. And acquiring an actual torque fluctuation difference value when the all-cylinder mode is switched to the cylinder-cut mode, and comparing the actual torque fluctuation difference value with an expected torque fluctuation difference value.
4. When the error between the actual torque fluctuation difference value and the expected torque fluctuation difference value is larger than a first error threshold value, the error is reversely propagated through the wavelet neural network so as to correct the connection weight value between the input layer and the hidden layer in the wavelet neural network and the connection threshold value of the hidden layer until the error between the actual torque fluctuation difference value and the expected torque fluctuation difference value is smaller than the first error threshold value, and the trained wavelet neural network is obtained. The first error threshold refers to an allowable error, and may be determined empirically.
In the above description, the error back propagation is performed by adjusting the connection weight and the connection threshold of the node of each layer from the output layer as the start to the input layer, wherein the connection weight between the input layer and the hidden layer in the wavelet neural network can be modified by the following formula:
wherein η is the learning rate of weight correction, 0<η<1，wijThe connection weight from the ith input layer to the jth hidden layer;
correcting the connection threshold of the hidden layer in the wavelet neural network by the following formula:
where μ is the learning rate of the threshold correction, 0<μ<1，θjIs the connection threshold of the jth hidden layer cell;
correcting the connection weight between the hidden layer and the output layer in the wavelet neural network by the following formula:
wherein v isjiAnd connecting the ith hidden layer to the jth output layer.
More specifically, as shown in fig. 3, the training process (i.e., the self-learning process) of the wavelet neural network can be summarized into several stages, such as data preprocessing, sample selection, forward propagation, error calculation, and error back propagation.
Data preprocessing:
from the beginning of network training, input data are normalized to be changed between 0 and 1 through the sigmoid function with the same status for each input component, the adjustment range of the network weight is reduced, and the difficulty in network training is reduced.
Sample selection:
and selecting a group of data from the preprocessed training sample set, inputting the data into the neural network, and calculating an output value of the neural network.
Forward propagation:
the forward propagation process is a process of propagation in the direction of an arrow as shown in fig. 2, and when the learning process starts, the forward propagation process propagates from the input layer to the output layer via the intermediate hidden layers, and each neuron in the output layer obtains an input response of the network.
And (3) error analysis:
the error analysis link aims at the analysis of the output torque fluctuation of the crankshaft before and after mode switching, wherein a torque fluctuation signal can be measured by an electromagnetic rotating speed sensor and a speed measuring fluted disc. The torque fluctuation is too large, resonance is easily formed, the performance and the power performance of the whole machine are affected, and the comfort of passengers in the vehicle is reduced. In this link, the actual torque fluctuation value is an absolute value of a difference between the torque fluctuation value acquired before the mode switching and the torque fluctuation value acquired after the mode switching, and the difference is compared with a set difference of the expected torque fluctuation.
Under the ideal condition, namely when the mode switching of cylinder breaking/full cylinder does not occur, when the engine keeps the same mode to work, the torque fluctuation value is acquired twice, the difference value is zero, namely the performance and the power performance of the whole vehicle are not changed, and the comfort of passengers in the vehicle is not changed.
For node j, the error of the actual output from the desired output is defined as,
wherein,and ojRespectively, the desired output and the actual output of node j.
Set to a specified tolerance, E, if the actual outputs of the m output nodes all satisfy the corresponding m desired outputs of the samplejAnd j is less than or equal to 1, 2. Otherwise, if Ej>J 1,2, m enters the error back propagation process.
And (3) error back propagation:
from the output layer to the input layer, error back propagation is carried out by adjusting the weight and the threshold of each layer of neuron; the artificial neural network stores 'new knowledge' by adjusting the weight of each neuron, and in the error back propagation process, the principle of coefficient adjustment is to modify the weight w according to the negative gradient of the error and to correct the threshold according to the formula.
The network training is repeatedly executed according to the steps, and in the executing process, the weight and the threshold are adjusted according to the difference value between the actual output and the expected output, so that the actual output is closer to the expected output, and the training of the wavelet neural network is completed.
Step S103: and acquiring the running parameters of the vehicle, and judging whether to enter a cylinder-breaking mode according to the running parameters. Wherein the operating parameters of the vehicle include, but are not limited to: an engine speed signal, an engine torque signal, an accelerator pedal signal, a gear signal, a vehicle speed signal, and a coolant temperature signal.
In the on-line control stage, the working conditions of the engine under various complex working conditions need to be considered by the cylinder-breaking control system, and when the idling working condition or the rotating speed is too low, the cylinder-breaking can cause the whole engine to vibrate violently and the whole engine needs to work in a full-cylinder mode; when the engine is in cold start or the temperature of the cooling liquid is too low, the fuel oil is not completely vaporized, and at the moment, the whole cylinder is required to work, so that the stable combustion is ensured; under the condition of quick acceleration, in order to meet the power requirement of the whole machine, the whole-cylinder mode is required to work, and good power performance is ensured; the gear signal, the accelerator pedal signal and the like are directly related to the running condition of the vehicle, and the gear signal, the accelerator pedal signal and the like are also taken as sufficient conditions for entering the cylinder-cut mode.
As shown in fig. 4, firstly, according to signals of the sensor (including an engine speed signal, an engine torque signal, an accelerator pedal signal, a gear signal, a vehicle speed signal, a coolant temperature signal, etc.), the actual operating conditions of the engine and the vehicle are determined, in order to avoid influencing the normal operation of the engine, the engine does not enter a cylinder deactivation mode under the conditions of warming up, idling, acceleration/deceleration, too low/too high rotational speed, etc., according to the collected sensor signals, a control interval (including a rotational speed interval, a torque interval, a vehicle speed interval, a coolant temperature interval, etc.) corresponding to the operating conditions is firstly set, and whether the engine operates in the operating condition of the interval is determined: if the machine runs under the working condition in the interval, the normal mode of the original machine is kept; if the engine does not operate in the working condition of the interval, the basic requirement of the cylinder-breaking mode is considered to be met, and the next link of the control system is entered. The specific parameters of the cylinder deactivation mode are controlled by a control system.
Step S104: if so, inputting the running parameters of the vehicle into the trained wavelet neural network, determining the cylinder failure number and the cylinder failure sequence according to the running parameters of the vehicle, generating corresponding cylinder failure mode signals, and obtaining the air intake flow and the oil injection flow of the working cylinder according to the cylinder failure mode signals.
Step S105: and controlling the cylinder breaking process of the engine according to the cylinder breaking mode signal, the air inlet flow and the oil injection quantity of the working cylinder.
Specifically, when the cylinder-breaking mode is met, input parameters are subjected to wavelet transformation through a wavelet signal analysis link, and the wavelet transformation has the characteristic of high operation speed and is beneficial to real-time processing of an engine control unit. And (4) the parameters after the wavelet transformation enter a neural network control system, and output parameters are obtained through calculation, wherein the output parameters comprise 2 variables, namely the cylinder failure number and the cylinder failure sequence. The cylinder-breaking number refers to the number of the working cylinders which are stopped, the cylinder-breaking sequence refers to the working time sequence of each cylinder (oil injector) after the cylinder-breaking number is determined, and if the cylinder-breaking number is 2, the specific cylinder is determined which two cylinders are stopped. More specifically, a neural network control system is used for selecting a proper cylinder deactivation number according to the working condition of the engine, and a proper cylinder deactivation sequence is selected according to the principle that the torque fluctuation is minimum before and after mode switching. And then according to the output cylinder-breaking mode signal, obtaining the air intake flow of the working cylinder in the cylinder-breaking mode, correspondingly controlling the air distribution phase, and correspondingly controlling the oil injection quantity to complete the control of the whole cylinder-breaking process.
According to the method for realizing the cylinder-failure mode of the engine, disclosed by the embodiment of the invention, wavelet analysis and artificial neural network control are combined, the cylinder-failure system of the engine can be accurately controlled in real time, the neural network has self-adaption and self-learning capabilities, the uncertainty and the time-varying property of a control object can be overcome, less test data is used as a sample to train the network, the neural network can be enabled to have higher learning and convergence speeds by utilizing the wavelet analysis, and the real-time performance of control is improved. The method improves the uniformity among cylinders and the working stability, improves the reliability of the whole vehicle and improves the comfort of a driver while ensuring the cylinder-breaking effect (namely fuel economy).
Fig. 5 is a block diagram of a cylinder deactivation mode implementing system of an engine according to an embodiment of the present invention. As shown in fig. 5, a cylinder deactivation mode implementing system 500 of an engine according to an embodiment of the present invention includes: a training sample acquisition module 510, a training module 520, a determination module 530, and a control module 540.
The training sample obtaining module 510 is configured to select multiple engine speeds, and perform a load characteristic test in multiple preset cylinder deactivation modes at each engine speed to obtain multiple sets of training samples, where the training samples include engine speeds, output torques, and torque fluctuation conditions associated with a crankshaft angle. The training module 520 is configured to train a preset wavelet neural network according to the training sample until a predetermined condition is met, so as to obtain a trained wavelet neural network. The determining module 530 is configured to obtain an operation parameter of the vehicle, and determine whether to enter a cylinder deactivation mode according to the operation parameter. The control module 540 is configured to, when the judgment module judges that the vehicle enters the cylinder deactivation mode, input the operating parameters of the vehicle into the trained wavelet neural network, determine the number and the sequence of cylinder deactivation according to the operating parameters of the vehicle, generate a corresponding cylinder deactivation mode signal, obtain the intake air flow and the fuel injection quantity of the working cylinder according to the cylinder deactivation mode signal, and control the cylinder deactivation process of the engine according to the cylinder deactivation mode signal, the intake air flow and the fuel injection quantity of the working cylinder.
In one embodiment of the present invention, the wavelet neural network includes an input layer, a hidden layer and an output layer, wherein the input layer has n input nodes, the output layer has m output nodes, n and m are positive integers, the number of nodes of the hidden layer is obtained by the following formula:
wherein k is the number of nodes of the hidden layer, and β is any constant between 1 and 10.
In one embodiment of the invention, training module 520 is configured to:
normalizing the training samples, wherein the training samples are normalized through a sigmoid function, and the sigmoid function is as follows:
wherein c is a coefficient of the sigmoid function;
inputting a group of training samples after normalization processing into the wavelet neural network for forward propagation;
acquiring an actual torque fluctuation difference value when the full-cylinder mode is switched to the cylinder-off mode, and comparing the actual torque fluctuation difference value with an expected torque fluctuation difference value;
when the error between the actual torque fluctuation difference value and the expected torque fluctuation difference value is larger than a first error threshold value, the error is reversely propagated through the wavelet neural network so as to correct the connection weight value between the input layer and the hidden layer in the wavelet neural network and the connection threshold value of the hidden layer until the error between the actual torque fluctuation difference value and the expected torque fluctuation difference value is smaller than the first error threshold value, and the trained wavelet neural network is obtained. The first error threshold refers to an allowable error, and may be determined empirically.
The training module 520 performs error back propagation by adjusting the connection weight and the connection threshold of the node of each layer from the output layer as the start to the input layer, wherein the connection weight between the input layer and the hidden layer in the wavelet neural network is modified by the following formula:
wherein, the η is the learning rate of weight correction, 0<η<1，wijThe connection weight from the ith input layer to the jth hidden layer;
correcting the connection threshold of the hidden layer in the wavelet neural network through the following formula:
wherein μ is a learning rate of threshold correction, 0<μ<1, said thetajIs the connection threshold of the jth hidden layer cell;
correcting the connection weight between the hidden layer and the output layer in the wavelet neural network through the following formula:
wherein, v isjiAnd connecting the ith hidden layer to the jth output layer.
According to the engine cylinder-failure mode implementation system provided by the embodiment of the invention, wavelet analysis and artificial neural network control are combined, the cylinder-failure system of the engine can be accurately controlled in real time, the neural network has self-adaption and self-learning capabilities, the uncertainty and time-varying property of a control object can be overcome, less test data is used as a sample to train the network, the neural network can have higher learning and convergence speed by utilizing the wavelet analysis, and the real-time performance of control is improved. The system improves the uniformity and working stability among the cylinders, improves the reliability of the whole vehicle and improves the comfort of a driver while ensuring the cylinder-breaking effect (namely fuel economy).
It should be noted that, a specific implementation manner of the engine cylinder deactivation mode implementation system according to the embodiment of the present invention is similar to a specific implementation manner of the engine cylinder deactivation mode implementation method according to the embodiment of the present invention, and please refer to the description of the method part specifically, and details are not repeated in order to reduce redundancy.
Further, the embodiment of the invention discloses a vehicle which comprises the cylinder-cut-off mode realization system of the engine. The vehicle combines wavelet analysis and artificial neural network control, can accurately control a cylinder-breaking system of an engine in real time, the neural network has self-adaption and self-learning capabilities, uncertainty and time-varying property of a control object can be overcome, less test data are used as samples to train the network, the neural network can have higher learning and convergence speed by utilizing the wavelet analysis, and the real-time performance of control is improved. When the cylinder breaking effect (namely fuel economy) is guaranteed, uniformity among cylinders and working stability are improved, reliability of the whole vehicle is improved, and comfort of a driver is improved.
In addition, other configurations and functions of the vehicle according to the embodiment of the present invention are known to those skilled in the art, and are not described in detail in order to reduce redundancy.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.
Claims (10)
1. A method for realizing a cylinder-cut-off mode of an engine is characterized by comprising the following steps:
selecting a plurality of engine rotating speeds, and performing load characteristic tests in a plurality of preset cylinder-breaking modes at each engine rotating speed to obtain a plurality of groups of training samples, wherein the training samples comprise the engine rotating speeds, output torques and torque fluctuation conditions along with the crankshaft rotating angle;
training a preset wavelet neural network according to the training sample until a preset condition is met to obtain a trained wavelet neural network;
acquiring running parameters of a vehicle, and judging whether to enter a cylinder-breaking mode according to the running parameters;
if so, inputting the running parameters of the vehicle into the trained wavelet neural network to determine the cylinder failure number and the cylinder failure sequence according to the running parameters of the vehicle, generating corresponding cylinder failure mode signals, and obtaining the air intake flow and the oil injection flow of the working cylinder according to the cylinder failure mode signals; and
and controlling the cylinder breaking process of the engine according to the cylinder breaking mode signal, the air intake flow and the oil injection quantity of the working cylinder.
2. The method of claim 1, wherein the wavelet neural network comprises an input layer, a hidden layer and an output layer, wherein the input layer has n input nodes, the output layer has m output nodes, n and m are positive integers, and the number of nodes in the hidden layer is obtained by the following formula:
wherein k is the number of nodes of the hidden layer, and β is any constant between 1 and 10.
3. The method for implementing the cylinder deactivation mode of the engine according to claim 2, wherein the training of the wavelet neural network which is preset according to the training samples is performed until a predetermined condition is met, so as to obtain the trained wavelet neural network, specifically comprising:
normalizing the training samples, wherein the training samples are normalized through a sigmoid function, and the sigmoid function is as follows:
wherein c is a coefficient of the sigmoid function;
inputting a group of training samples after normalization processing into the wavelet neural network for forward propagation;
acquiring an actual torque fluctuation difference value when the full-cylinder mode is switched to the cylinder-off mode, and comparing the actual torque fluctuation difference value with an expected torque fluctuation difference value;
when the error between the actual torque fluctuation difference value and the expected torque fluctuation difference value is larger than a first error threshold value, the error is reversely propagated through the wavelet neural network so as to correct the connection weight value between the input layer and the hidden layer in the wavelet neural network and the connection threshold value of the hidden layer until the error between the actual torque fluctuation difference value and the expected torque fluctuation difference value is smaller than the first error threshold value, and the trained wavelet neural network is obtained.
4. The method according to claim 3, wherein error back propagation is performed by adjusting the connection weights and connection thresholds of the nodes of each layer from the output layer as a start to the input layer, wherein the connection weights between the input layer and the hidden layer in the wavelet neural network are modified by the following formula:
wherein, the η is the learning rate of weight correction, 0<η<1, said wijThe connection weight from the ith input layer to the jth hidden layer;
correcting the connection threshold of the hidden layer in the wavelet neural network through the following formula:
wherein μ is a learning rate of threshold correction, 0<μ<1, said thetajIs the connection threshold of the jth hidden layer cell;
correcting the connection weight between the hidden layer and the output layer in the wavelet neural network through the following formula:
wherein, v isjiAnd connecting the ith hidden layer to the jth output layer.
5. The method of implementing a cylinder deactivation mode of an engine according to claim 1, wherein said vehicle operating parameters include: an engine speed signal, an engine torque signal, an accelerator pedal signal, a gear signal, a vehicle speed signal, and a coolant temperature signal.
6. A cylinder deactivation mode enabling system of an engine, comprising:
the training sample acquisition module is used for selecting a plurality of engine rotating speeds and performing load characteristic tests in a plurality of preset cylinder failure modes at each engine rotating speed to obtain a plurality of groups of training samples, wherein the training samples comprise the engine rotating speeds, output torques and torque fluctuation conditions along with the rotating angles of the crankshafts;
the training module is used for training a preset wavelet neural network according to the training sample until a preset condition is met to obtain a trained wavelet neural network;
the judging module is used for acquiring the running parameters of the vehicle and judging whether to enter a cylinder-failure mode according to the running parameters;
and the control module is used for inputting the running parameters of the vehicle into the trained wavelet neural network when the judgment module judges that the vehicle enters the cylinder failure mode, determining the cylinder failure number and the cylinder failure sequence according to the running parameters of the vehicle, generating corresponding cylinder failure mode signals, obtaining the air intake flow and the oil injection flow of a working cylinder according to the cylinder failure mode signals, and controlling the cylinder failure process of the engine according to the cylinder failure mode signals, the air intake flow and the oil injection flow of the working cylinder.
7. The cylinder deactivation mode enabling system of an engine according to claim 6, wherein said wavelet neural network comprises an input layer, a hidden layer and an output layer, wherein said input layer has n input nodes, said output layer has m output nodes, said n and m are positive integers, and the number of nodes of said hidden layer is obtained by the following formula:
wherein k is the number of nodes of the hidden layer, and β is any constant between 1 and 10.
8. The cylinder deactivation mode enabling system of an engine of claim 7, wherein said training module is configured to:
normalizing the training samples, wherein the training samples are normalized through a sigmoid function, and the sigmoid function is as follows:
wherein c is a coefficient of the sigmoid function;
inputting a group of training samples after normalization processing into the wavelet neural network for forward propagation;
acquiring an actual torque fluctuation difference value when the full-cylinder mode is switched to the cylinder-off mode, and comparing the actual torque fluctuation difference value with an expected torque fluctuation difference value;
when the error between the actual torque fluctuation difference value and the expected torque fluctuation difference value is larger than a first error threshold value, the error is reversely propagated through the wavelet neural network so as to correct the connection weight value between the input layer and the hidden layer in the wavelet neural network and the connection threshold value of the hidden layer until the error between the actual torque fluctuation difference value and the expected torque fluctuation difference value is smaller than the first error threshold value, and the trained wavelet neural network is obtained.
9. The system of claim 8, wherein the training module performs error back propagation by adjusting connection weights and connection thresholds of nodes of each layer from the output layer as a start to the input layer, wherein the connection weights between the input layer and the hidden layer in the wavelet neural network are modified by the following formula:
wherein, the η is the learning rate of weight correction, 0<η<1, said wijThe connection weight from the ith input layer to the jth hidden layer;
correcting the connection threshold of the hidden layer in the wavelet neural network through the following formula:
wherein μ is a learning rate of threshold correction, 0<μ<1, said thetajIs the connection threshold of the jth hidden layer cell;
correcting the connection weight between the hidden layer and the output layer in the wavelet neural network through the following formula:
wherein, v isjiAnd connecting the ith hidden layer to the jth output layer.
10. A vehicle characterized by being provided with the cylinder deactivation mode realization system of the engine according to any one of claims 6-9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510450509.0A CN106401757B (en) | 2015-07-28 | 2015-07-28 | Disconnected cylinder mode implementation method, system and the vehicle of engine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510450509.0A CN106401757B (en) | 2015-07-28 | 2015-07-28 | Disconnected cylinder mode implementation method, system and the vehicle of engine |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106401757A true CN106401757A (en) | 2017-02-15 |
| CN106401757B CN106401757B (en) | 2019-07-05 |
Family
ID=58008636
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510450509.0A Active CN106401757B (en) | 2015-07-28 | 2015-07-28 | Disconnected cylinder mode implementation method, system and the vehicle of engine |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106401757B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111022207A (en) * | 2019-11-19 | 2020-04-17 | 潍柴动力股份有限公司 | Control method and control system for fuel injection quantity during cylinder-failure mode switching |
| CN111102090A (en) * | 2019-11-19 | 2020-05-05 | 潍柴动力股份有限公司 | Control method and control system for fuel injection quantity in cylinder-cut-off mode |
| CN112696277A (en) * | 2020-12-29 | 2021-04-23 | 潍柴动力股份有限公司 | Engine cylinder deactivation control method and engine |
| CN113107697A (en) * | 2021-05-07 | 2021-07-13 | 上海柴油机股份有限公司 | Torque fluctuation evaluation method of diesel engine under constant rotating speed |
| CN114909225A (en) * | 2022-04-25 | 2022-08-16 | 潍柴动力股份有限公司 | Oil-saving control method, device and system for AMT vehicle |
| CN114962016A (en) * | 2021-08-18 | 2022-08-30 | 长城汽车股份有限公司 | Engine cylinder deactivation control method, engine cylinder deactivation control device, medium and vehicle |
| CN116163846A (en) * | 2023-04-20 | 2023-05-26 | 潍柴动力股份有限公司 | Cylinder deactivation control method and device for engine and engine |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5200898A (en) * | 1989-11-15 | 1993-04-06 | Honda Giken Kogyo Kabushiki Kaisha | Method of controlling motor vehicle |
| US5361213A (en) * | 1990-02-09 | 1994-11-01 | Hitachi, Ltd. | Control device for an automobile |
| US20040230368A1 (en) * | 2003-05-14 | 2004-11-18 | Kropinski Michael A. | Method and apparatus to diagnose intake airflow |
| JP2005263100A (en) * | 2004-03-19 | 2005-09-29 | Mitsubishi Fuso Truck & Bus Corp | Vehicle control device |
| US20060047487A1 (en) * | 2004-08-26 | 2006-03-02 | Volponi Allan J | Bootstrap data methodology for sequential hybrid model building |
| CN1981123A (en) * | 2004-06-25 | 2007-06-13 | Fev电机技术有限公司 | Motor vehicle control device provided with a neuronal network |
| CN101067401A (en) * | 2006-05-02 | 2007-11-07 | 通用汽车环球科技运作公司 | Redundant torque security path |
| CN101198783A (en) * | 2005-04-28 | 2008-06-11 | 雷诺股份公司 | Method for controlling a motor vehicle using a network of neurones |
| JP2011149348A (en) * | 2010-01-22 | 2011-08-04 | Daihatsu Motor Co Ltd | Control device for idle stop vehicle |
-
2015
- 2015-07-28 CN CN201510450509.0A patent/CN106401757B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5200898A (en) * | 1989-11-15 | 1993-04-06 | Honda Giken Kogyo Kabushiki Kaisha | Method of controlling motor vehicle |
| US5361213A (en) * | 1990-02-09 | 1994-11-01 | Hitachi, Ltd. | Control device for an automobile |
| US20040230368A1 (en) * | 2003-05-14 | 2004-11-18 | Kropinski Michael A. | Method and apparatus to diagnose intake airflow |
| JP2005263100A (en) * | 2004-03-19 | 2005-09-29 | Mitsubishi Fuso Truck & Bus Corp | Vehicle control device |
| CN1981123A (en) * | 2004-06-25 | 2007-06-13 | Fev电机技术有限公司 | Motor vehicle control device provided with a neuronal network |
| US20060047487A1 (en) * | 2004-08-26 | 2006-03-02 | Volponi Allan J | Bootstrap data methodology for sequential hybrid model building |
| CN101198783A (en) * | 2005-04-28 | 2008-06-11 | 雷诺股份公司 | Method for controlling a motor vehicle using a network of neurones |
| CN101067401A (en) * | 2006-05-02 | 2007-11-07 | 通用汽车环球科技运作公司 | Redundant torque security path |
| JP2011149348A (en) * | 2010-01-22 | 2011-08-04 | Daihatsu Motor Co Ltd | Control device for idle stop vehicle |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111022207A (en) * | 2019-11-19 | 2020-04-17 | 潍柴动力股份有限公司 | Control method and control system for fuel injection quantity during cylinder-failure mode switching |
| CN111102090A (en) * | 2019-11-19 | 2020-05-05 | 潍柴动力股份有限公司 | Control method and control system for fuel injection quantity in cylinder-cut-off mode |
| CN112696277A (en) * | 2020-12-29 | 2021-04-23 | 潍柴动力股份有限公司 | Engine cylinder deactivation control method and engine |
| CN113107697A (en) * | 2021-05-07 | 2021-07-13 | 上海柴油机股份有限公司 | Torque fluctuation evaluation method of diesel engine under constant rotating speed |
| CN114962016A (en) * | 2021-08-18 | 2022-08-30 | 长城汽车股份有限公司 | Engine cylinder deactivation control method, engine cylinder deactivation control device, medium and vehicle |
| CN114909225A (en) * | 2022-04-25 | 2022-08-16 | 潍柴动力股份有限公司 | Oil-saving control method, device and system for AMT vehicle |
| CN116163846A (en) * | 2023-04-20 | 2023-05-26 | 潍柴动力股份有限公司 | Cylinder deactivation control method and device for engine and engine |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106401757B (en) | 2019-07-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106401757B (en) | Disconnected cylinder mode implementation method, system and the vehicle of engine | |
| Klein et al. | Engine in the loop: Closed loop test bench control with real-time simulation | |
| US10991174B2 (en) | Machine learning device of amount of unburned fuel, machine learning method, learned model, electronic control unit, method of production of electronic control unit, and machine learning system | |
| CN113153521A (en) | System and method for predicting vehicle engine torque by using artificial neural network | |
| DE102007047763B4 (en) | System and method for estimating a torque | |
| JP2010523886A (en) | ENGINE CONTROL METHOD AND CONTROL DEVICE | |
| DE102011113169A1 (en) | System for diagnosing fault conditions of a gas flow control system for turbocharged engines | |
| US10825267B2 (en) | Control system of internal combustion engine, electronic control unit, server, and control method of internal combustion engine | |
| Isermann et al. | Design of computer controlled combustion engines | |
| CN104408271A (en) | Model-based gasoline engine calibration method | |
| US20200123990A1 (en) | Control device of internal combustion engine and control method of same and learning model for controlling internal combustion engine and learning method of same | |
| CN102567782A (en) | Neural-network-based automobile engine torque estimation method | |
| CN112983608A (en) | Particle trap carbon load calculation method and system, controller and storage medium | |
| CN113266481A (en) | Vehicle control method, vehicle control device, and server | |
| JP2022530504A (en) | Engine control system | |
| Wei | Design of Web based expert system of electronic control engine fault diagnosis | |
| Shayler et al. | The exploitation of neural networks in automotive engine management systems | |
| Malikopoulos et al. | Optimal engine calibration for individual driving styles | |
| CN105443259A (en) | Intelligent diesel engine cylinder balance adjusting algorithm based on approximate dynamic planning | |
| CN107035560B (en) | Engine control system based on driver type online identification | |
| CN104653317B (en) | Torque control method based on arbitration voting | |
| Cranmer et al. | Grey-box modeling architectures for rotational dynamic control in automotive engines | |
| Müller et al. | Approximation and control of the engine torque using neural networks | |
| Xie et al. | Using sensors data and emissions information to diagnose engine’s faults | |
| CN111255557B (en) | Automobile engine cooling detection system and control method thereof |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | 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 |