CN112610344A - Common rail injector fault diagnosis method based on CEEMD and improved level discrete entropy - Google Patents

Abstract

The invention aims to provide a common rail fuel injector fault diagnosis method based on CEEMD and improved level discrete entropy. And then calculating improved level discrete entropy of the reconstructed signal, and inputting the improved level discrete entropy as fault characteristics into the LSSVM multi-classifier, thereby realizing fault diagnosis of the common rail fuel injector. The method is suitable for fault diagnosis of the common rail fuel injector in a field industrial environment with strong noise interference, and has the advantages of high fault diagnosis accuracy and strong anti-interference performance.

Description

Common rail injector fault diagnosis method based on CEEMD and improved level discrete entropy

Technical Field

The invention relates to a diesel engine control method, in particular to a diesel engine oil injection control method.

Background

With the rapid development of science and technology, the energy crisis and environmental problems are more and more severe, and a high-pressure common rail diesel engine is developed and produced for relieving the problems. With the rapid increase of the number of high-pressure common rail diesel engines, the problem of fault diagnosis and maintenance becomes a difficult problem to be ignored. The research of the engine management of the Japan Ship east Association shows that the failure rate of the oil injector accounts for 17.1 percent of the failure rate of the high-pressure common rail diesel engine, the failure of the oil injector causes the combustion deterioration, the power performance, the economic performance and the reliability of the diesel engine to be reduced, and the harmful emissions are increased. Therefore, how to accurately identify the fault of the fuel injector is an urgent problem to be solved.

The working state information of the oil injector can be embodied by the fuel pressure wave of the common rail pipe, but the fuel pressure wave is a nonlinear and non-stable signal, and in order to extract time domain and frequency domain information at the same time, a joint time-frequency analysis method is required to be applied for fault detection. The EMD is a time-frequency analysis method proposed by Huang and the like, is widely applied in the field of fault diagnosis, but the EMD has some defects such as mode aliasing, end effect, stopping conditions and the like. Aiming at the inherent defect of EMD, Yeh et al propose a Complementary overall average Empirical Mode Decomposition (CEEMD) which eliminates the residual noise by adding auxiliary white noise with positive and negative alternation to the original signal. Compared with EMD, CEEMD is more stable, avoids the mode mixing problem, and CEEMD can also reduce residual noise and improve calculation efficiency by adding the positive and negative alternating auxiliary noise pairs.

After filtering processing, how to effectively extract the weak fault characteristics of the oil injector becomes the key for realizing the weak fault of the oil injector. Entropy is one of the most effective tools for measuring time series uncertainty and dynamic characteristics, and due to such advantages, various information entropy methods are applied to fault diagnosis, such as sample entropy, fuzzy entropy, discrete entropy, and the like. The sample entropy overcomes the problem of template self-matching in the approximate entropy, and has the characteristics of strong anti-interference capability, good consistency in a large parameter value range and the like. However, since it defines vector similarity based on unit step function, it cannot accurately define input category. Therefore, in order to overcome the defect of the sample entropy, a concept of the fuzzy entropy is proposed, such as chenweiting, but the fuzzy entropy calculation is relatively complex, and the calculation efficiency is influenced. Azami proposes Discrete Entropy (DE) to alleviate the respective disadvantages of sample Entropy and fuzzy Entropy. For sample entropy and fuzzy entropy, discrete entropy is more efficient to compute and more robust to interference. To ensure the integrity and accuracy of information measurement, Azami proposed Multi-Scale discrete Entropy (MDE) based on DE. However, the multi-scale Discrete Entropy only considers the low-frequency components of the time sequence and ignores the high-frequency part of the time sequence, and based on the above, songi et al provides a Hierarchical Discrete Entropy (HDE) based on Hierarchical analysis and Discrete Entropy. However, as the number k of decomposition layers of the HDE increases, the time sequence is shortened, which results in the decrease of statistical reliability, and the HDE has a certain limitation on the signal length, which reduces the universality of the HDE algorithm.

Disclosure of Invention

The invention aims to provide a common rail injector fault diagnosis method based on CEEMD and improved level discrete entropy, which overcomes the defects of HDE, can provide comprehensive evaluation for irregularity and uncertainty of timing sequences and solves the problem of low common rail injector fault diagnosis precision under complex working conditions and noise environments.

The purpose of the invention is realized as follows:

the invention relates to a common rail oil sprayer fault diagnosis method based on CEEMD and improved level discrete entropy, which is characterized by comprising the following steps:

(1) collecting pressure fluctuation signals of the high-pressure oil pipe through a pressure sensor arranged on the high-pressure oil pipe, and dividing the collected pressure signals into training signals and testing signals;

(2) filtering the pressure signal by using a CEEMD algorithm, eliminating redundant components through correlation coefficients, and reconstructing the pressure signal;

(3) calculating an improved level discrete entropy of the reconstructed pressure signal, and taking IHDE as a fault feature of the fuel pressure signal;

(4) inputting a multi-classifier of a least square support vector machine by taking IHDE of all training samples as a feature vector for training;

(5) and carrying out fault diagnosis and pattern recognition on the IHDE of the test sample by adopting the trained least square support vector machine multi-classifier, and outputting a diagnosis result.

The present invention may further comprise:

1. the CEEMD algorithm comprises the following processes:

(1) when EEMD decomposition is performed on the signal X (t), let the total average number be N, and white noise w is added to the ith time of X (t)_iAfter (t) obtaining

Let X (t) subtract w_i(t) obtaining

(2) To pair

And

EEMD decomposition is respectively carried out to obtain a group of intrinsic mode components IMF respectively_i ⁺And IMF_i ^-Then, then

The above formula is averaged to obtain

Then, a correlation coefficient of each IMF component with the original signal is calculated.

2. The calculation process of IHDE is described as follows:

(1) given a time sequence { x (i), i ═ 1, 2., N }, an averaging operator Q for the time sequence is defined₀Sum difference operator Q₁：

Wherein Q is₀(x) And Q₁(x) Respectively representing the low-frequency and high-frequency components of signal x (i);

(2) of k levels

Is expressed as:

j is 0 or 1;

(3) construct a vector { gamma₁,γ₂,...,γ_kThe integer e can be expressed as

Wherein, { gamma., (gamma.)_m M 1,2, k ∈ {0,1} denotes an average or difference operator of the mth layer;

(4) based on vector { gamma₁,γ₂,...,γ_kThe hierarchical components of the time series X (i) are defined as follows.

(5) Calculating the discrete entropy of each layer component, wherein the improved layer discrete entropy is defined as follows:

IHDE＝DE(X_k,e,m,c,d)。

the invention has the advantages that: the invention effectively utilizes CEEMD to filter the fuel pressure signal, has low noise interference and obtains an effective fault signal; and then, the fault information of the fuel pressure signal is comprehensively and accurately reflected through the improved hierarchical discrete entropy, the fault diagnosis method is suitable for completing the fault diagnosis of the common rail fuel injector in a strong noise environment, and the fault diagnosis method has the advantages of high fault diagnosis accuracy and strong anti-interference performance.

Drawings

FIG. 1 is a flow chart of the present invention;

FIG. 2 is a time domain waveform of a high pressure fuel line fuel pressure signal for three injector states;

FIG. 3 shows the decomposition results of the CEEMD of the pressure waves in three states of the fuel injector;

FIG. 4 illustrates the failure recognition rate of the CEEMD-IHDE method;

fig. 5 shows correlation coefficients of each IMF component with the original signal.

Detailed Description

The invention will now be described in more detail by way of example with reference to the accompanying drawings in which:

with reference to fig. 1-5, the invention relates to a common rail injector fault diagnosis method based on CEEMD and improved hierarchical discrete entropy, which comprises the following steps:

and S1, collecting pressure fluctuation signals of the high-pressure oil pipe through a pressure sensor arranged on the high-pressure oil pipe, wherein the time domain waveform of the high-pressure oil pipe is shown in figure 2, and the collected pressure signals are divided into training signals and testing signals.

S2, performing adaptive filtering processing on the pressure signal by using a CEEMD algorithm, wherein the IMF component is shown in fig. 3. And eliminating redundant components through the correlation coefficient to obtain effective fault components, and reconstructing a fuel pressure fault signal. The process of the CEEMD algorithm is briefly described as follows:

(1) when EEMD decomposition is performed on the signal X (t), let the total average number be N, and white noise w is added to the ith time of X (t)_iAfter (t) obtaining

Let X (t) subtract w_i(t) obtaining

(2) To pair

And

EEMD decomposition is respectively carried out to obtain a group of intrinsic mode components IMF respectively_i ⁺And IMF_i ^-Then, then

The above formula is averaged to obtain

Then, the correlation coefficient of each IMF component with the original signal is calculated as shown in fig. 5. As can be seen in FIG. 5, the three components IMF1, IMF4, and IMF5 have relatively large correlation coefficients, and therefore are selected to reconstruct the common rail fuel pressure signal.

And S3, calculating improved level discrete entropy of the reconstructed pressure signal, and taking IHDE as a fault feature of the fuel pressure signal. The calculation process of IHDE can be described as follows:

(1) given a time series { x (i) ═ 1,2, …, N }, the average operator Q of the time series is defined₀Sum difference operator Q₁。

Wherein Q is₀(x) And Q₁(x) Respectively representing the low frequency component and the high frequency component of the signal x (i).

(2) Of k levels

The matrix form of (j ═ 0 or 1) can be expressed as:

(3) construct a vector { gamma₁,γ₂,…,γ_kH, the integer e can be expressed as

Wherein, { gamma., (gamma.)_mAnd m is 1,2, …, k } ∈ {0,1} represents an average value or difference operator of the mth layer.

(4) Based on vector { gamma₁,γ₂,…,γ_kThe hierarchical components of the time series X (i) are defined as follows.

(5) Calculating the discrete entropy of each layer component, and defining the improved layer discrete entropy as follows

IHDE＝DE(X_k,e,m,c,d)

And S4, inputting the least squares support vector machine multi-classifier by taking the IHDE of all training samples as the feature vector for training.

S5, fault diagnosis and pattern recognition are carried out on the IHDE of the test sample by adopting the trained least square support vector machine multi-classifier, and a diagnosis result is output, wherein the classification result is shown in figure 4.

Claims (3)

1. A common rail fuel injector fault diagnosis method based on CEEMD and improved level discrete entropy is characterized in that:

(1) collecting pressure fluctuation signals of the high-pressure oil pipe through a pressure sensor arranged on the high-pressure oil pipe, and dividing the collected pressure signals into training signals and testing signals;

(2) filtering the pressure signal by using a CEEMD algorithm, eliminating redundant components through correlation coefficients, and reconstructing the pressure signal;

(3) calculating an improved level discrete entropy of the reconstructed pressure signal, and taking IHDE as a fault feature of the fuel pressure signal;

(4) inputting a multi-classifier of a least square support vector machine by taking IHDE of all training samples as a feature vector for training;

(5) and carrying out fault diagnosis and pattern recognition on the IHDE of the test sample by adopting the trained least square support vector machine multi-classifier, and outputting a diagnosis result.

2. The common rail injector fault diagnosis method based on CEEMD and improved hierarchical discrete entropy as claimed in claim 1, characterized in that: the CEEMD algorithm comprises the following processes:

(1) when EEMD decomposition is performed on the signal X (t), let the total average number be N, and white noise w is added to the ith time of X (t)_iAfter (t) obtaining

Let X (t) subtract w_i(t) obtaining

(2) To pair

And

EEMD decomposition is respectively carried out to obtain a group of intrinsic mode components IMF respectively_i ⁺And IMF_i ^-Then, then

The above formula is averaged to obtain

Then, a correlation coefficient of each IMF component with the original signal is calculated.

3. The common rail injector fault diagnosis method based on CEEMD and improved hierarchical discrete entropy as claimed in claim 1, characterized in that: the calculation process of IHDE is described as follows:

(1) given a time sequence { x (i), i ═ 1, 2., N }, an averaging operator Q for the time sequence is defined₀Sum difference operator Q₁：

Wherein Q is₀(x) And Q₁(x) Respectively representing the low-frequency and high-frequency components of signal x (i);

(2) of k levels

Is expressed as:

j is 0 or 1;

(3) construct a vector { gamma₁,γ₂,...,γ_kThe integer e can be expressed as

Wherein, { gamma., (gamma.)_mM 1,2, …, k ∈ {0,1} represents an average or difference operator of the mth layer;

(4) based on vector { gamma₁,γ₂,...,γ_kThe hierarchical components of the time series X (i) are defined as follows.

(5) Calculating the discrete entropy of each layer component, wherein the improved layer discrete entropy is defined as follows:

IHDE＝DE(X_k,e,m,c,d)。

Images