CN108956140B - Oil transfer pump bearing fault diagnosis method - Google Patents

Abstract

The invention discloses a fault diagnosis method for a bearing of an oil transfer pump, which is characterized by comprising the following steps of: (1) carrying out data acquisition on a bearing vibration signal by adopting an acceleration sensor, and carrying out EMD (empirical mode decomposition) and IMF (intrinsic mode function) envelope spectrum analysis to obtain sample data characteristic physical quantity; (2) processing the sample data characteristic physical quantity based on the bearing fault characteristic matrix to obtain a BPA value as a BPA evidence; (3) and improving the D-S evidence reasoning rule to fuse the BPA evidence to obtain a bearing fault diagnosis result. The oil delivery pump bearing fault diagnosis method provided by the invention can be used for processing partial data failure or abnormal conditions caused by sensor damage in a plurality of bearing acceleration sensors.

Description

Oil transfer pump bearing fault diagnosis method

Technical Field

The invention relates to the field of bearing fault diagnosis, and provides a method for diagnosing a bearing fault of an oil delivery pump for ensuring the transportation safety and economic benefit of an oil pipeline.

Background

Petroleum is an important resource for guaranteeing the high-speed and stable development of national economy, pipeline transportation is used as an efficient, economic and safe transportation mode, and the petroleum transportation mode becomes the preferred transportation mode of petroleum transportation by virtue of incomparable advantages of other traditional transportation. The oil transfer pump becomes the most critical movable equipment in pipeline transportation, and the stable operation of the oil transfer pump directly influences the safety of a petroleum pipeline transportation system.

At present, the diagnosis of the bearing fault of the oil delivery pump is simply judged based on a single sensor under most conditions, for example, the fault characteristics are extracted by collecting single state signals of equipment, such as vibration, temperature and flow state, so that the fault is identified. Under natural or human interference conditions, there is a significant risk of using a single sensor, and failure of a single sensor may result in misjudgment of the equipment failure state. The multiple sensors can reduce the misjudgment risk, and the obtained information is inevitably uncertain under the interference condition, so that the conflict information exists. Methods such as mathematical averaging or simple logic processing are often used to identify erroneous fault diagnosis results because of the inability to process such conflicting information.

Disclosure of Invention

The invention aims to provide a novel method for diagnosing the faults of a bearing of an oil transfer pump, which aims at solving the problems of high risk and uncertainty of identifying the faults of the oil transfer pump by a single sensor.

The technical scheme of the invention is as follows:

a fault diagnosis method for a bearing of an oil transfer pump comprises the following steps:

(1) acquiring data of the bearing vibration signals by adopting an acceleration sensor, and performing EMD (empirical mode decomposition) and IMF (intrinsic mode function) envelope spectrum analysis to obtain sample data characteristic physical quantity;

(2) processing the sample data characteristic physical quantity based on the bearing fault characteristic matrix to obtain a BPA value as a BPA evidence;

(3) and improving the D-S evidence reasoning rule to fuse the BPA evidence to obtain a bearing fault diagnosis result.

Specifically, in the step (1), a plurality of acceleration sensors are adopted to detect bearing vibration so as to obtain a fault vibration signal, frequency spectrum analysis is carried out on the signal, on the basis, the fault vibration signal is further decomposed into the sum of at least three intrinsic mode functions IMF through EMD empirical mode, the first three intrinsic mode functions IMF are taken to carry out envelope spectrum analysis respectively, and sample data characteristic physical quantity is obtained.

Specifically, in the step (2), based on the gray correlation degree, calculating the gray correlation degree of the collected sample data characteristic physical quantity and the bearing fault characteristic matrix to obtain the BPA evidence, wherein the process is as follows:

a: the rolling bearing faults are specifically divided into inner ring faults, outer ring faults and rolling body faults, and are sequentially recorded as fault domains { A }₁,A₂,A₃Taking the respective failure characteristic frequency as the characteristic physical quantity, soBarrier { A₁,A₂,A₃The characteristic reference sequence of (x)_t1,x_t2,x_t3) T is 1,2, 3; characteristic reference sequence (x)_t1,x_t2,x_t3) The element values in t 1,2 and 3 are obtained by theoretical calculation in advance or processing and counting a large number of historical faults and corresponding sample data; forming a bearing reference fault characteristic matrix:

b: recording the characteristic physical quantity of sample data obtained after EMD processing and envelope spectrum analysis of bearing vibration signals acquired by the ith acceleration sensor as P⁽ⁱ⁾＝(y⁽ⁱ⁾ ₁,y⁽ⁱ⁾ ₂，y⁽ⁱ⁾ ₃) I is 1,2, …, n, n represents the number of sensors; calculating P⁽ⁱ⁾For fault domain { A₁,A₂,A₃Gray scale of { gamma) } degree of gray scale_i1,γ_i2,γ_i3}：

B1: calculating P⁽ⁱ⁾For fault A_tCorrelation coefficient alpha in q dimension_tq：

Wherein t is 1,2,3, q is 1,2,3, and rho is a resolution coefficient and is between 0 and 1;

B2：P⁽ⁱ⁾to A_tHas a gray scale relation of gamma_it

C: normalized gray degree of association { gamma_i1,γ_i2,γ_i3Get the ith evidence information to failure A_tEvidence of BPA of m_i(A_t)：

Specifically, in the step (3), BPA evidences are fused based on the inter-evidence conflict factor k, different fusion formulas are adopted for different k values, and the calculation of the inter-evidence conflict factor k specifically includes:

recording the focal yuan as A_tT is 1,2, 3; BPA evidences acquired by data acquired by the ith and the j acceleration sensors are respectively expressed as m_i＝(m_i(A₁),m_i(A₂),m_i(A₃) And m) and_j＝(m_j(A₁),m_j(A₂),m_j(A₃) I, j denotes the number of the acceleration sensors, i, j ═ 1,2, …, n, n denotes the number of bearing acceleration sensors, the collision factor k of n BPA evidence sets is calculated:

combining evidence groups pairwise, and calculating the conflict factor k_ijTake k_ijAs the collision factor k of the data source, i.e.:

k＝max(k_ij) (5)

specifically, in step (3), when k is less than 0.779, evidence is fused with the formula m (A)_t) Comprises the following steps:

obtaining probability distribution m (A) deduced based on each evidence source, namely acceleration sensor information_t) Selecting m (A)_t) Fault a corresponding to maximum value_tAs a final oil transfer pump bearing fault diagnosis result.

Specifically, in the step (3), when k is greater than or equal to 0.779, the evidence fusion step specifically comprises:

a: calculate for the same focal element A_tT 1,2,3, distance d between two evidence of BPA_ij(A_t)：

Aligning n BPA evidences to a certain focal length A_tThe distance matrix of (A) is regularized to obtain a matrix D^*(A_t)：

Wherein: d_ij ^*＝d_ij(A_t)/d_max，d_max＝max{d_ij(A_t)}；0＜d_ij ^*＜1；

B: support matrix and other confidence calculation

Defining a support matrix Y (A)_t)＝1-D^*(A_t) I.e. Y (A)_t) Middle element y_ij＝1-d_ij ^*,y_ii＝1：

Mixing Y (A)_t) Adding according to columns and normalizing to obtain the credibility vector Sup (A)_t)：

Sup(A_t)＝[Sup₁(A_t),Sup₂(A_t),…Sup_n(A_t)](10)

Sup_i(A_t) Representing for the same focal element A_tOther evidence to evidence E_iThe degree of support of;

separately calculate { A ] for different focal elements₁,A₂,A₃Sup of (A)_t)，

Define his confidence matrix Z for evidence:

line i of Z reflects other evidence for evidence E_iThe degree of support of;

averaging the rows of Z to obtain evidence E_iHis degree of confidence t_i；

Absolute confidence of others T_i＝t_i/max{t_i}；

C: evidence correction

Based on the obtained absolute confidence T_iAnd correcting the trust distribution value of the evidence and recording as m^*：m_i ^*(A_t)＝T_i*m_i(A_t) I is 1,2, … n, and since the absolute confidence level is not always 1, adding the confidence distribution value m of the unknown focal element into the corrected data source_i ^*(X)：

Namely, the correction sequence form is as follows: original evidence BPA is m_i＝(m_i(A₁),m_i(A₂),m_i(A₃) Corrected BPA of m)_i ^*，

m_i ^*＝(m_i ^*(A₁),m_i ^*(A₂),m_i ^*(A₃),m_i ^*(X))

＝[T_im_i(A₁),T_im_i(A₂),T_im_i(A₃),1-T_im_i(A₁)-T_im_i(A₂)-T_im_i(A₃)]

D: blending formula m (A)_t)：

Wherein the content of the first and second substances,

p(A_t) Is A_t、A_aAssign A in conflict_tThe method comprises the following steps:

p^*(A_t) Is A_a、A_bAssign A in conflict_tThe method comprises the following steps:

during calculation, for n evidences obtained by n acceleration sensors, the 1 st evidence m is taken firstly₁ ^*2 nd evidence m₂ ^*Performing rational synthesis according to the formula (12), and then combining the synthetic result with the 3 rd evidence m₃ ^*Synthesizing and repeatedly executing until the last evidence m_n ^*Synthesizing to obtain probability distribution m (A) deduced based on information of each evidence source, namely each acceleration sensor_t) Selecting m (A)_t) Fault A corresponding to medium probability maximum_tAs a final oil transfer pump bearing fault diagnosis result.

The invention has the beneficial effects that:

1. the oil delivery pump bearing fault diagnosis method provided by the invention can be used for processing partial data failure or abnormal conditions caused by sensor damage in a plurality of bearing acceleration sensors.

2. The method is suitable for the condition of consistent information of the multi-bearing acceleration sensor and the condition of information conflict of the multi-bearing acceleration sensor.

3. The invention provides a new information reasoning and synthesizing formula, which is more reasonable in distribution of conflict information, fuses the obtained information to the maximum extent and enhances the reliability of the recognition result.

4. The invention provides a new information reasoning and synthesizing formula, has faster convergence speed and better fusion effect, and can identify the correct target under the condition of less evidence groups.

5. The misjudgment rate of the diagnosis of the bearing fault of the oil transfer pump is reduced.

Drawings

FIG. 1 is a flow chart of a method for diagnosing a bearing fault of an oil transfer pump according to the present invention

FIG. 2 is a time domain waveform and a frequency spectrum of an original vibration signal of a fault signal 1 to be detected in a data processing step of the sensor 1 in the embodiment

FIG. 3 is the three IMF time domain waveforms after EMD decomposition of the fault signal to be detected 1 in the data processing step of the sensor 1 in the embodiment

FIG. 4 shows three IMF spectra of data processing steps taken by the sensor 1 in an embodiment

FIG. 5 shows three IMF envelope spectra of data processing steps taken by the sensor 1 in an embodiment

Detailed Description

The present invention will be further described with reference to the following examples.

A fault diagnosis method for a bearing of an oil transfer pump adopts a plurality of bearing acceleration sensors to acquire bearing vibration signals and diagnoses faults according to the processing of the vibration signals. The novel oil transfer pump fault diagnosis method mainly comprises the following steps: the method comprises three parts, namely a data acquisition processing step, a Basic Probability Assignment (BPA) step and an evidence reasoning synthesis fault identification step.

Take a 1.5KW motor, for example, with the bearing being tested supporting the motor shaft. The method comprises the steps of arranging a single-point fault on a bearing by using an electric spark machining technology, wherein the fault diameter is 0.1778mm, collecting vibration signals by using an acceleration sensor, collecting the vibration signals by using a DAT recorder with 16 channels, and processing the vibration signals in an MATLAB environment at the later stage. The sampling frequency is 12000Hz, and the bearing rotating speed is 1772 r/min.

The size of the bearing is as follows: (Unit: inch)

Diameter of inner ring	Outer diameter	Thickness of	Diameter of rolling element	Pitch diameter
					0.9843	2.0472	0.5906	0.3126	1.537

Failure frequency (multiple of rotating speed Hz)

Inner ring	Outer ring	Holding rack	Rolling body
				5.4152	3.5848	0.39828	4.7135

1. Collected data processing step

And calculating the fault frequency according to the parameters to obtain the fault frequency of the inner ring of 159.93Hz, the fault frequency of the outer ring of 105.87Hz and the fault frequency of the rolling body of 139.21 Hz. And carrying out spectrum analysis on the acquired vibration signal by utilizing matlab, further decomposing the fault vibration signal into the sum of at least three intrinsic mode functions IMF by EMD empirical mode on the basis, and respectively carrying out envelope spectrum analysis on the first three intrinsic mode functions IMF to obtain the fault characteristic quantity of the fault signal to be detected.

The data collected by the sensor 1 and the processing steps are shown in fig. 2-5.

Finally, obtaining a fault signal to be detected: p₁＝(159.7 159.7 159.7)

2. Basic probability assignment step

Recording inner ring fault, outer ring fault and rolling body fault as fault domains { A₁,A₂,A₃Obtaining fault characteristic quantity (x) of each fault domain based on theoretical calculation in advance (the specific method refers to the above collected data processing steps, other embodiments can also obtain fault characteristic quantity (x) of each fault domain through a large number of historical faults and corresponding sample data processing statistics thereof)_t1,x_t2,x_t3) T is 1,2,3, i.e.:

the characteristic quantities of the inner ring fault sample are as follows: (159.93,159.93,159.93)

The outer ring fault sample characteristic quantity is as follows: (105.87,105.87,105.87)

The characteristic quantities of the inner ring fault sample are as follows: (139.21,139.21,139.21)

Establishing a fault sample space:

then P is calculated according to the equations (2) and (3)₁To A_tHas a gray scale relation of gamma_1tAs shown in Table 1

TABLE 1P₁To A_tDegree of gray scale correlation of

The signal P to be detected can be obtained from the formula (4)₁Evidence of BPA for each failure is

m₁(A₁)＝0.5239，m₁(A₂)＝0.1761，m₁(A₃)＝0.3；

And repeating the steps, and sequentially calculating the BPA value of the data obtained by the 4 sensors to the sample space.

3. Evidence reasoning synthesis fault identification step

The module synthesizes a plurality of groups of evidence BPA according to an improved D-S reasoning rule to realize the final identification of the fault. The following description will be made in terms of the case where there is a relatively high agreement or a relatively high conflict between the evidences of the plurality of sensors.

(1) If different evidence information obtained by data collected by the 4 sensors is consistent, setting the BPA values corresponding to different sensors as follows (k is 0.5813):

m₁(A₁)＝0.5239，m₁(A₂)＝0.1761，m₁(A₃)＝0.3；

m₂(A₁)＝0.6511，m₂(A₂)＝0.071，m₂(A₃)＝0.2779；

m₃(A₁)＝0.708，m₃(A₂)＝0.07，m₃(A₃)＝0.222；

m₄(A₁)＝0.5905，m₄(A₂)＝0.1087，m₄(A₃)＝0.3008；

the combined reasoning results for 4 evidences are shown in Table 2

TABLE 2 reasoning identification results on the consensus evidence

As shown in Table 1, the method of the invention has obvious convergence, is effective for low-conflict evidence combination, and has a bearing fault diagnosis result A₁I.e. inner ring failure.

(2) If there is a large conflict between different pieces of evidence information acquired by the data acquired by the 4 sensors, setting BPA values corresponding to the different sensors as follows (k is 0.9134):

m₁(A₁)＝0.5239，m₁(A₂)＝0.1761，m₁(A₃)＝0.3；

m₂(A₁)＝0，m₂(A₂)＝0.8907，m₂(A₃)＝0.1093；

m₃(A₁)＝0.708，m₃(A₂)＝0.07，m₃(A₃)＝0.222；

m₄(A₁)＝0.5905，m₄(A₂)＝0.1087，m₄(A₃)＝0.3008；

the combined reasoning results for 4 evidences are shown in Table 3

TABLE 3 reasoning and identification results for conflict evidences

The results show that the method of the present invention can correctly identify the target when the 3 rd proof is received. The analysis reason can find that: the evidence 2 is greatly deviated from the actual situation due to the possible factors of unreliable sensors and bad environment. The inference rule considers the mutual relation among a plurality of BPA evidences, not only considers the global effectiveness represented by the credibility of each evidence, but also reasonably distributes local conflicts, reduces the influence of 'bad values' on a fusion result and a decision to the greatest extent, improves the convergence speed, reduces the decision risk, and can converge into a correct target under the condition of less evidences.

Therefore, the method is effective in identifying the operating state of the system when the evidences of the sensors are consistent or large conflicts exist.

The parts not involved in the present invention are the same as or can be implemented using the prior art.

Claims (2)

1. A fault diagnosis method for a bearing of an oil transfer pump is characterized by comprising the following steps:

(1) acquiring data of a bearing vibration signal by adopting an acceleration sensor, performing EMD empirical mode decomposition and intrinsic mode function IMF (intrinsic mode function) envelope spectrum analysis to obtain sample data characteristic physical quantity, specifically, detecting the vibration of the bearing by adopting a plurality of acceleration sensors to obtain a fault vibration signal, performing frequency spectrum analysis on the signal, further performing EMD empirical mode decomposition on the fault vibration signal into the sum of at least three intrinsic mode functions IMF on the basis, and performing envelope spectrum analysis on the first three intrinsic mode functions IMF respectively to obtain sample data characteristic physical quantity;

(2) processing the sample data characteristic physical quantity based on the bearing fault characteristic matrix to obtain a BPA value as a BPA evidence; based on the grey correlation degree, calculating the grey correlation degree of the collected sample data characteristic physical quantity and the bearing fault characteristic matrix to obtain the BPA evidence, wherein the process comprises the following steps:

a: the rolling bearing faults are specifically divided into inner ring faults, outer ring faults and rolling body faults, and are sequentially recorded as fault domains { A }₁,A₂,A₃And taking the respective fault characteristic frequency as a characteristic physical quantity to obtain a fault domain { A }₁,A₂,A₃The characteristic reference sequence of (x)_t1,x_t2,x_t3) T is 1,2, 3; characteristic reference sequence (x)_t1,x_t2,x_t3) The element values in t 1,2 and 3 are obtained by theoretical calculation in advance or processing and counting a large number of historical faults and corresponding sample data; forming a bearing reference fault characteristic matrix:

b: the characteristic physical quantity of sample data obtained after EMD processing and envelope spectrum analysis of bearing vibration signals acquired by the ith acceleration sensor is recorded as P⁽ⁱ⁾＝(y⁽ⁱ⁾ ₁,y⁽ⁱ⁾ ₂，y⁽ⁱ⁾ ₃) I is 1,2, …, n, n represents the number of acceleration sensors; calculating P⁽ⁱ⁾For fault domain { A₁,A₂,A₃Gray correlation degree of { gamma }_i1,γ_i2,γ_i3}：

B1: calculating P⁽ⁱ⁾For fault A_tCorrelation coefficient alpha in q dimension_tq：

Wherein t is 1,2,3, q is 1,2,3, and rho is a resolution coefficient and is between 0 and 1;

B2：P⁽ⁱ⁾to A_tGray scale of gamma_it

C: normalized Gray correlation [ gamma ]_i1,γ_i2,γ_i3Get the ith evidence information to failure A_tEvidence of BPA of m_i(A_t)：

(3) And improving the D-S evidence reasoning rule to fuse the BPA evidence to obtain a bearing fault diagnosis result: fusing BPA evidences based on the conflict factor k between the evidences, wherein different fusion formulas are adopted for different k values, and the calculation step of the conflict factor k between the evidences specifically comprises the following steps:

recording the focal yuan as A_tT is 1,2, 3; BPA evidences acquired by data acquired by the ith and the j acceleration sensors are respectively expressed as m_i＝(m_i(A₁),m_i(A₂),m_i(A₃) And m) and_j＝(m_j(A₁),m_j(A₂),m_j(A₃) I and j denote the numbers of the acceleration sensors, i, j is 1,2, …, n, n denotes the number of the bearing acceleration sensors, and the inter-evidence conflict factors k of n BPA evidence groups are calculated:

combining evidence groups pairwise, and calculating conflict factor k between evidences_ijTake k_ijAs the inter-evidence conflict factor k of the data source, namely:

k＝max(k_ij) (5)

when k is more than or equal to 0.779, evidence fusion steps are specifically as follows:

a: calculate for the same focal element A_tT 1,2,3, distance d between two evidence of BPA_ij(A_t)：

Aligning n BPA evidences to a certain focal length A_tThe distance matrix of (A) is regularized to obtain a matrix D^*(A_t)：

Wherein: d_ij ^*＝d_ij(A_t)/d_max，d_max＝max{d_ij(A_t)}；0<d_ij ^*<1；

B: support matrix and other confidence calculation

Defining a support matrix Y (A)_t)＝1-D^*(A_t) I.e. Y (A)_t) Middle element y_ij＝1-d_ij ^*,y_ii＝1：

Mixing Y (A)_t) Adding according to columns and normalizing to obtain the credibility vector Sup (A)_t)：

Sup(A_t)＝[Sup₁(A_t),Sup₂(A_t),…Sup_n(A_t)](10)

Sup_i(A_t) Representing for the same focal element A_tOther evidence to evidence E_iThe degree of support of;

separately calculate { A ] for different focal elements₁,A₂,A₃Sup of (A)_t)，

Define his confidence matrix Z for evidence:

line i of Z reflects other evidence for evidence E_iThe degree of support of;

averaging the rows of Z to obtain evidence E_iHis degree of confidence t_i；

Absolute confidence of others T_i＝t_i/max{t_i}；

C: evidence correction

Based on the obtained absolute confidence T_iAnd correcting the trust distribution value of the evidence and recording as m^*：m_i ^*(A_t)＝T_i*m_i(A_t) I is 1,2, … n, and since the absolute confidence level is not always 1, adding the confidence distribution value m of the unknown focal element into the corrected data source_i ^*(X)：

Namely, the correction sequence form is as follows: original evidence BPA is m_i＝(m_i(A₁),m_i(A₂),m_i(A₃) Corrected BPA of m)_i ^*，

m_i ^*＝(m_i ^*(A₁),m_i ^*(A₂),m_i ^*(A₃),m_i ^*(X))

＝[T_im_i(A₁),T_im_i(A₂),T_im_i(A₃),1-T_im_i(A₁)-T_im_i(A₂)-T_im_i(A₃)]

D: blending formula m (A)_t)：

Wherein the content of the first and second substances,

p(A_t) Is A_t、A_aAssign A in conflict_tThe method comprises the following steps:

p^*(A_t) Is A_a、A_bAssign A in conflict_tThe method comprises the following steps:

during calculation, for n evidences obtained by n acceleration sensors, the 1 st evidence m is taken firstly₁ ^*2 nd evidence m₂ ^*Performing rational synthesis according to the formula (12), and then combining the synthetic result with the 3 rd evidence m₃ ^*Synthesizing and repeatedly executing until the last evidence m_n ^*Synthesizing to obtain probability distribution m (A) deduced based on information of each evidence source, namely each acceleration sensor_t) Selecting m (A)_t) Fault A corresponding to medium probability maximum_tAs a final oil transfer pump bearing fault diagnosis result.

2. The method for diagnosing the failure of the bearing of the oil transfer pump as claimed in claim 1, wherein in the step (3), when k is equal to<At 0.779, evidence fuses the formula m (A)_t) Comprises the following steps:

obtaining probability distribution m (A) deduced based on each evidence source, namely acceleration sensor information_t) Selecting m (A)_t) Fault a corresponding to maximum value_tAs a final oil transfer pump bearing fault diagnosis result.

Images