CN110955951B - Product life prediction method and device based on path classification and estimation - Google Patents
Product life prediction method and device based on path classification and estimation Download PDFInfo
- Publication number
- CN110955951B CN110955951B CN201811124089.7A CN201811124089A CN110955951B CN 110955951 B CN110955951 B CN 110955951B CN 201811124089 A CN201811124089 A CN 201811124089A CN 110955951 B CN110955951 B CN 110955951B
- Authority
- CN
- China
- Prior art keywords
- product
- data
- path
- failure
- life
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 49
- 230000015556 catabolic process Effects 0.000 claims abstract description 131
- 238000006731 degradation reaction Methods 0.000 claims abstract description 131
- 239000013074 reference sample Substances 0.000 claims abstract description 77
- 238000006243 chemical reaction Methods 0.000 claims description 38
- 238000012360 testing method Methods 0.000 claims description 22
- 230000001133 acceleration Effects 0.000 claims description 14
- 239000000523 sample Substances 0.000 claims description 13
- 238000004364 calculation method Methods 0.000 claims description 11
- 238000000611 regression analysis Methods 0.000 claims description 9
- 230000009467 reduction Effects 0.000 claims description 5
- 230000006872 improvement Effects 0.000 description 12
- 230000008569 process Effects 0.000 description 9
- 230000008859 change Effects 0.000 description 6
- 238000004458 analytical method Methods 0.000 description 4
- 238000011161 development Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000013459 approach Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24133—Distances to prototypes
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Abstract
The invention discloses a product life prediction method and a device based on path classification and estimation, wherein the method comprises the following steps: s1, respectively acquiring performance degradation data of a product to be detected and a plurality of failure reference samples, obtaining a target path according to the performance degradation data of the product to be detected, and obtaining a plurality of reference paths according to the performance degradation data of each failure reference sample; s2, calculating the similarity between each reference path and the target path respectively, and obtaining the residual life of the product to be tested by using the obtained similarity between each reference path and the target path and the residual life estimation of each failure reference sample; the device comprises a path acquisition module and a service life prediction module. The method has the advantages of simple implementation method, low required cost, high prediction precision, no need of knowing the failure threshold value of the product and the like, and can realize dynamic prediction of the residual life of the product.
Description
Technical Field
The invention relates to the technical field of performance evaluation of electronic and electromechanical products, in particular to a product life prediction method and device based on path classification and estimation.
Background
If the residual service life of the product can be predicted in the use process of the product, the degradation and failure state of the product can be found in time before the product fails, the occurrence of the product failure is avoided, and the operation reliability of the whole machine is improved.
For life prediction of electronic products, the prediction is generally performed directly based on the variation trend of the degradation data of the products at present, that is, the existing degradation data of the products to be predicted are obtained, the development trend of the subsequent degradation data is predicted by the degradation data, and the time of reaching the failure threshold of the degradation parameter in the development trend is used as the life prediction value. However, in the life prediction method, only the degradation data of the product to be predicted is adopted, the data amount is too small, and the life of the product is predicted based on the change trend simply and is difficult to prepare in practice.
The practitioner proposes to use a variation of a degradation parameter to describe the whole failure physical process of the product, and meanwhile, use a series of samples to compare and classify with failure data of the product to be detected, wherein the life of the similar samples after classification is used as the life of the product to be detected.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: aiming at the technical problems existing in the prior art, the invention provides the product life prediction method and the device based on path classification and estimation, which have the advantages of simple implementation method, low required cost, high prediction precision and no need of knowing the failure threshold value of the product, and can realize the dynamic prediction of the residual life of the product.
In order to solve the technical problems, the technical scheme provided by the invention is as follows:
a product life prediction method based on path classification and estimation comprises the following steps:
s1, path acquisition: respectively acquiring performance degradation data of a product to be detected and a plurality of failure reference samples, obtaining a target path according to the performance degradation data of the product to be detected, and obtaining a plurality of reference paths according to the performance degradation data of each failure reference sample;
s2, predicting service life: and respectively calculating the similarity between each reference path and the target path, and obtaining the residual life of the product to be tested by using the obtained similarity between each reference path and the target path and the residual life estimation of each failure reference sample.
As a further improvement of the process of the invention: the performance degradation data in the step S1 is obtained by firstly acquiring performance degradation data under high stress with a preset size and then converting the performance degradation data into normal stress.
As a further improvement of the method of the present invention, the step S1 of obtaining performance degradation data of the product to be tested and the plurality of failure reference samples includes: performing an accelerated life test on a plurality of test samples of a product to be tested, obtaining failure data and performance degradation data of each test sample under high stress with different sizes, respectively performing data conversion to convert the data under the high stress into normal stress, taking one of the test samples as a target product and the rest as failure reference samples, obtaining the performance degradation data of the product to be tested according to the converted data obtained by the corresponding target product, and obtaining the performance degradation data of the failure reference sample according to the converted data obtained by the corresponding failure reference sample.
As a further improvement of the method of the present invention, the step of performing data conversion includes:
s11, obtaining failure samples under each step of stress reduction according to the failure data and the degradation data so as to convert the step of stress reduction data into constant stress data;
s12, under the condition of constraint of distribution parameters, carrying out distribution parameter fitting on equivalent life data of each stress to obtain distribution parameter estimation;
s13, carrying out accelerated model regression analysis on the life characteristic parameter estimation and the stress level of the product by using the obtained distribution parameter estimation to obtain model parameter estimation, and completing data conversion.
As a further improvement of the process of the invention: in the step S1, data fitting is performed on the performance degradation data of the product to be tested to obtain a target path, and data fitting is performed on the performance degradation data of each failure reference sample to obtain a plurality of reference paths.
As a further improvement of the process of the invention: and in the step S2, calculating the similarity between each reference path and the target path by using a kernel function.
As a further improvement of the method of the present invention, the step of calculating the similarity between each of the reference paths and the target path using a kernel function includes:
s21, respectively calculating the distance between each reference path and each target path;
s22, converting the calculated distances between the reference paths and the target paths into similarity by using a preset kernel function, and obtaining the similarity between the reference paths and the target paths.
As a further improvement of the process of the invention: the kernel function is a gaussian radial basis kernel function.
As a further improvement of the process of the invention: and when the residual life of the product to be detected is estimated in the step S2, determining a weight value corresponding to each failure reference sample according to the similarity between each failure reference sample path and the target path, and calculating the residual life of the product to be detected according to the residual life of each failure reference sample and the corresponding weight value.
As a further improvement of the method, the residual life of the product to be tested is calculated by the following formula:
wherein RL is a 0 To the remaining life of the product to be tested, RL i The remaining life of the ith failed reference sample, n is the number of failed reference samples, w i And the weight value corresponding to the ith failure reference sample is determined based on the similarity.
A product life prediction device based on path classification and estimation, comprising:
the path acquisition module is used for respectively acquiring performance degradation data of a product to be detected and a plurality of failure reference samples, obtaining a target path according to the performance degradation data of the product to be detected, and obtaining a plurality of reference paths according to the performance degradation data of each failure reference sample;
and the life prediction module is used for respectively calculating the similarity between each reference path and the target path and obtaining the residual life of the product to be detected by using the obtained similarity between each reference path and the target path and the residual life estimation of each failure reference sample.
As a further improvement of the device of the invention: the path acquisition module comprises a degradation data acquisition submodule, wherein the degradation data acquisition submodule is used for acquiring performance degradation data of a product to be detected and a plurality of failure reference samples, the degradation data acquisition submodule is used for acquiring failure data and degradation data of the product to be detected under high stress with preset size, respectively carrying out data conversion to convert the data under the high stress into normal stress, taking the data obtained by the failure data conversion as the performance degradation data of the failure reference samples, and taking the data obtained by the degradation data conversion as the performance degradation data of the product to be detected, and acquiring the performance degradation data of the product to be detected and the plurality of failure reference samples.
As a further development of the inventive arrangement, the degraded data acquiring submodule comprises a conversion unit for performing data conversion, the conversion unit comprising:
the conversion subunit is used for obtaining failure samples under each step-down stress from the failure data and the degradation data so as to convert the step-down stress data into constant stress data;
the distribution parameter estimation subunit is used for carrying out distribution parameter fitting on the equivalent life data of each stress under the condition of distribution parameter constraint to obtain distribution parameter estimation;
and the acceleration model regression analysis subunit is used for carrying out acceleration model regression analysis on the life characteristic parameter estimation and the stress level of the product by using the obtained distribution parameter estimation to obtain model parameter estimation and complete data conversion.
As a further improvement of the device of the invention: the life prediction module comprises a similarity calculation submodule, which is used for calculating the similarity between each reference path and the target path by using a kernel function, and comprises:
a distance calculation unit for calculating the distance between each reference path and the target path;
and the similarity conversion unit is used for converting the calculated distances between each reference path and the target path into similarity by using a preset kernel function respectively, so as to obtain the similarity between each reference path and the target path.
As a further improvement of the device of the invention: the life prediction module comprises a remaining life estimation sub-module for estimating the remaining life of the product to be detected, wherein the remaining life estimation sub-module determines a weight value corresponding to each failure reference sample according to the similarity between each failure reference sample path and the target path, and the remaining life of each failure reference sample and the corresponding weight value are calculated to obtain the remaining life of the product to be detected.
Compared with the prior art, the invention has the advantages that:
1. according to the method and the device for predicting the service life of the product based on path classification and estimation, the target path corresponding to the performance degradation data of the product to be detected and the multiple reference paths corresponding to the performance degradation data of each failure reference sample are obtained by obtaining the performance degradation data of the product to be detected and the multiple failure reference samples, the residual service life of the product to be detected is finally comprehensively obtained according to the residual service life of each failure reference sample and the similarity between the target path and each reference path, the predicted data quantity is large based on the degradation data of the product to be detected and each failure reference sample, the service life prediction of the product can be realized by fully utilizing the performance degradation data of the product to be detected and the performance degradation data of each failure reference sample and the correlation between the product to be detected and each failure reference sample, and the service life prediction precision of the product is effectively improved.
2. According to the method and the device for predicting the service life of the product based on path classification and estimation, the residual service life can be predicted by the performance degradation data of the product to be detected and the failure reference sample, the failure threshold value of the product is not required to be known in advance, and compared with the traditional prediction mode based on the change trend of the degradation data, the method and the device for predicting the service life of the product can realize dynamic prediction of the residual service life of the product.
3. According to the method and the device for predicting the service life of the product based on path classification and estimation, the kernel function is further used for calculating the similarity between each reference path and the target path, the kernel function is introduced for realizing the kernel path classification and the estimated service life prediction, the similarity between each reference path and the target path is obtained based on the kernel function, the influence degree of a reference sample with high similarity in the prediction can be effectively improved, the influence degree of a reference sample with low similarity is reduced, and therefore the correlation between a failure reference sample and degradation data of the product to be detected can be fully utilized for realizing accurate residual service life prediction.
Drawings
Fig. 1 is a schematic flow chart of an implementation of a product life prediction method based on path classification and estimation according to embodiment 1 of the present invention.
FIG. 2 is a schematic diagram of a specific implementation of product life prediction in example 1 of the present invention.
Fig. 3 is a schematic diagram of the implementation flow of data conversion by using the three-step method in embodiment 1 of the present invention.
FIG. 4 is a graph showing the results of degradation data of the drift coefficient under high stress of a certain type of electromechanical product obtained in example 2 of the present invention.
FIG. 5 is a graph showing the results of degradation data of drift coefficients under normal stress of a certain type of electromechanical product obtained in example 2 of the present invention.
FIG. 6 is a schematic diagram showing the result of predicting the remaining life of the product obtained in example 2 of the present invention.
Detailed Description
The invention is further described below in connection with the drawings and the specific preferred embodiments, but the scope of protection of the invention is not limited thereby.
Example 1:
as shown in fig. 1 and 2, the product life prediction method based on path classification and estimation in this embodiment includes the following steps:
s1, path acquisition: respectively acquiring performance degradation data of a product to be detected and a plurality of failure reference samples, obtaining a target path according to the performance degradation data of the product to be detected, and obtaining a plurality of reference paths according to the performance degradation data of each failure reference sample;
s2, predicting service life: and respectively calculating the similarity between each reference path and the target path, and obtaining the residual life of the product to be tested by using the obtained similarity between each reference path and the target path and the residual life estimation of each failure reference sample.
According to the method, the target path corresponding to the performance degradation data of the product to be detected and the performance degradation data of the plurality of failure reference samples are obtained through obtaining the performance degradation data of the product to be detected and the plurality of reference paths corresponding to the performance degradation data of the failure reference samples, the residual life of the product to be detected is finally obtained comprehensively through the residual life of the failure reference samples and the similarity between the target path and the reference paths, the predicted data amount is large based on the degradation data of the product to be detected and the failure reference samples, the performance degradation data of the product to be detected and the failure reference samples and the correlation between the product to be detected and the failure reference samples can be fully utilized, the life prediction of the product is achieved, and the life prediction precision of the product is effectively improved.
By adopting the method, the residual life prediction can be realized by the performance degradation data of the product to be detected and the failure reference sample, the failure threshold value of the product is not required to be known in advance, and compared with the traditional prediction mode based on the degradation data change trend, the dynamic residual life prediction of the product can be realized.
In this embodiment, the performance degradation data in step S1 is obtained by first obtaining performance degradation data under high stress of a predetermined size and then converting the data to normal stress. The performance degradation data under high stress can be obtained by carrying out accelerated life test and the like on the product, and then converting the performance degradation data under high stress obtained by the test into performance degradation data under normal stress to obtain the performance degradation data required by prediction.
In this embodiment, the obtaining performance degradation data of the product to be tested and the plurality of failure reference samples in step S1 specifically includes: performing an accelerated life test on a plurality of test samples of a product to be tested, obtaining failure data and performance degradation data of each test sample under high stress with different sizes, respectively performing data conversion to convert the data under the high stress into normal stress, taking one of the test samples as a target product and the rest as failure reference samples, obtaining the performance degradation data of the product to be tested according to the converted data obtained by the corresponding target product, and obtaining the performance degradation data of the failure reference sample according to the converted data obtained by the corresponding failure reference sample. The data conversion is performed by performing time scale change to change the performance degradation data to normal stress. By the method, a plurality of test samples of the same type of product are tested, so that the performance degradation data of the product to be tested and the performance degradation data of the failure reference sample can be obtained simultaneously, and the life prediction of the product to be tested can be realized based on the performance degradation data of the product to be tested and the failure reference sample.
According to the embodiment, firstly, a step-down stress acceleration life test is implemented, failure data and performance degradation data of a product under high stress are obtained, an Arrhenius acceleration model is built by adopting the failure data and a three-step analysis method, and the performance degradation data under the high stress is converted into normal stress through a data conversion formula.
In this embodiment, a three-step method is specifically adopted to convert performance degradation parameters under high stress into normal stress, as shown in fig. 3, and the steps of performing data conversion by using the three-step method specifically include:
s11, obtaining a complete failure sample under each step-down stress from failure data and degradation data so as to convert the step-down stress data into constant stress data;
s12, under the condition of constraint of distribution parameters, carrying out distribution parameter fitting on equivalent life data of each stress to obtain distribution parameter estimation;
s13, carrying out accelerated model regression analysis on the life characteristic parameter estimation and the stress level of the product by using the obtained distribution parameter estimation to obtain model parameter estimation, and completing data conversion.
In the step S11, a data conversion method and an inverse distance estimation method are specifically used to obtain a complete failure sample under each step of stress reduction, so as to realize conversion from step stress test data to constant stress test data.
After the performance degradation data of the product to be detected and the performance degradation data of each failure reference sample are obtained through the steps, further performing data fitting to obtain a target path and a plurality of reference paths, namely performing data fitting on the performance degradation data of the product to be detected to obtain the target path, and performing data fitting on the performance degradation data of each failure reference sample to obtain a plurality of reference paths so as to realize product life prediction based on path classification and estimation.
In step S2 of this embodiment, the similarity between each reference path and the target path is calculated by using a kernel function, that is, the kernel function is introduced on the basis of obtaining the target path and implementing the path classification and estimation of the plurality of reference paths, so as to implement the life prediction of the kernel path classification and estimation. The method and the device have the advantages that when the distance between the two curves is larger, the similarity value is correspondingly smaller, the similarity value is larger for the smaller distance, when the similarity of the paths between the to-be-tested product and the failure reference sample is smaller, the data accuracy of the failure reference sample is lower, the influence degree of the failure reference sample data in prediction should be reduced, conversely, when the similarity of the paths between the to-be-tested product and the failure reference sample is higher, the data accuracy of the failure reference sample is higher, the influence degree of the failure reference sample data in prediction should be improved, the above characteristics of the kernel function are utilized, the similarity between each reference path and the target path is obtained based on the kernel function, and then the classification and estimation of the kernel path can be realized by the similarity state between the failure reference sample and the to-be-tested product, and the influence degree of the reference sample with high similarity in prediction can be improved by the characteristics of the kernel function, and the influence degree of the reference sample with low similarity is reduced, so that the accurate residual life prediction can be realized by fully utilizing the correlation between the failure reference sample and the degradation data of the to-be-tested product.
In this embodiment, the step of calculating the similarity between each reference path and the target path using the kernel function includes:
s21, respectively calculating the distance between each reference path and the target path;
s22, converting the calculated distances between each reference path and the target path into similarity by using a preset kernel function, and obtaining the similarity between each reference path and the target path.
In this embodiment, when calculating the similarity between each reference path and the target path, the euclidean distance between the target path and the sample path is specifically calculated first, so as to measure the proximity degree of the target path and the sample path, where the distance between the target path and the i-th sample path is:
the calculated distance is reconverted into a similarity by means of a kernel function, the gaussian radial basis kernel function is specifically adopted in the embodiment, the kernel function is used, a smaller similarity value exists for a larger distance, a larger similarity value exists for a smaller distance, and the similarity value is obtained based on the kernel function specifically according to the formula (2):
in the above formula, sigma refers to the bandwidth of the kernel function, the value is usually 0 < sigma less than or equal to 1, sigma=0.1 represents the condition of small bandwidth, and the weight generated when the distance is close to 0 is large; σ=1.0, a larger weight will result over a wider range of distances.
Obtaining a similarity vector by nuclear regression:
w＝[w 1 ,w 2 ,...,w n ] (3)
in the step S2 of this embodiment, when the remaining life of the product to be tested is estimated, determining the weight value corresponding to each failure reference sample according to the similarity between each failure reference sample path and the target path, and calculating the remaining life of the product to be tested from the remaining life of each failure reference sample and the corresponding weight value, that is, the similarity value w obtained based on the kernel function i The weight of each failure reference sample data is determined so as to improve the influence degree of the failure reference sample data with high similarity and reduce the influence degree of the failure reference sample data with low similarity, and finally, the residual life of each failure reference sample and the correlation can be synthesized to obtain an accurate product residual life value.
In this embodiment, the remaining life of the product to be measured, that is, the similarity between the target path and each sample path is calculated according to formulas (4) and (5), and is used as a weighted average of the remaining life of the samples to implement life prediction.
I.e. the residual life of the product to be tested is calculated by the formulas (4) and (5), wherein RL 0 To the remaining life of the product to be tested, RL i The remaining life of the ith failed reference sample, n is the number of failed reference samples, w i And the weight value corresponding to the ith failure reference sample.
Firstly, implementing a step-down stress acceleration life test to obtain failure data and performance degradation data of a product under high stress; establishing an Arrhenius acceleration model by adopting failure data and a three-step analysis method, and converting performance degradation data under high stress into normal stress by a data conversion formula; comparing and classifying the target degradation path and a series of sample degradation paths, and calculating Euclidean distance between the target path and each sample path; the Euclidean distance between the target path and the sample path is converted into the similarity by introducing the kernel function, the residual life of the target product is finally calculated through weighting by utilizing the similarity and the residual life of the sample, and the data of each failure reference sample and the correlation prediction can be fully utilized to obtain an accurate residual life value of the product.
The product life prediction device based on path classification and estimation in this embodiment includes:
the path acquisition module is used for respectively acquiring performance degradation data of a product to be detected and a plurality of failure reference samples, obtaining a target path according to the performance degradation data of the product to be detected, and obtaining a plurality of reference paths according to the performance degradation data of each failure reference sample;
and the life prediction module is used for respectively calculating the similarity between each reference path and the target path, and obtaining the residual life of the product to be detected by using the obtained similarity between each reference path and the target path and the residual life estimation of each failure reference sample.
In this embodiment, the path acquisition module includes a degradation data acquisition sub-module, configured to acquire performance degradation data of a product to be tested and a plurality of failure reference samples, where the degradation data acquisition sub-module acquires failure data and degradation data of the product to be tested under a high stress of a preset size, and performs data conversion respectively to convert the data under the high stress to normal stress, where the data obtained by converting the failure data is used as the performance degradation data of the failure reference samples, and the data obtained by converting the degradation data is used as the performance degradation data of the product to be tested, and obtain the performance degradation data of the product to be tested and the plurality of failure reference samples.
In this embodiment, the degraded data acquiring submodule includes a converting unit for converting data, and the converting unit includes:
the conversion subunit is used for obtaining failure samples under each step-down stress from the failure data and the degradation data so as to convert the step-down stress data into constant stress data;
the distribution parameter estimation subunit is used for carrying out distribution parameter fitting on the equivalent life data of each stress under the condition of distribution parameter constraint to obtain distribution parameter estimation;
and the acceleration model regression analysis subunit is used for carrying out acceleration model regression analysis on the life characteristic parameter estimation and the stress level of the product by using the obtained distribution parameter estimation to obtain model parameter estimation and complete data conversion.
In this embodiment, the lifetime prediction module includes a similarity calculation submodule, configured to calculate a similarity between each reference path and the target path using a kernel function, where the similarity calculation submodule includes:
a distance calculation unit for calculating the distance between each reference path and the target path;
and the similarity conversion unit is used for converting the calculated distances between each reference path and the target path into similarity by using a preset kernel function respectively, so as to obtain the similarity between each reference path and the target path.
In this embodiment, the life prediction module includes a remaining life estimation sub-module for estimating a remaining life of a product to be measured, where the remaining life estimation sub-module determines a weight value corresponding to each failure reference sample according to a similarity between each failure reference sample path and a target path, and calculates the remaining life of the product to be measured from the remaining life of each failure reference sample and the corresponding weight value.
The product life prediction device based on the path classification and the estimation in this embodiment corresponds to the product life prediction method based on the path classification and the estimation one by one, and will not be described in detail here.
Example 2:
in this example, the present invention will be further described with reference to the case where the residual life is predicted for a certain type of electromechanical product by the method of example 1.
Step one: and taking the temperature as acceleration stress, and performing a step-down stress acceleration life test on a certain electromechanical product produced in 2009 so as to obtain statistical data and performance degradation data of the electromechanical product under high stress.
In this example, the total number of test samples n=5, the applied temperature stress is 80 ℃, 66 ℃, 53 ℃, 40 ℃, the daily working temperature is 23 ℃, the number of failures under each stress is 2, 1, and the failure time is shown in table 1; selecting acceleration component a of electromechanical product along Y axis and instrument Y axis y Related drift Rate coefficient D (Y) y As an index of performance degradation of the product, a drift rate coefficient D (Y) before data conversion is obtained y The time profile is shown in fig. 4.
Table 1: time to failure of certain electromechanical product
| Sequence number | Stress level (. Degree. C.) | Time to failure (Tian) |
| 1 | 80 | 20,40 |
| 2 | 66 | 20 |
| 3 | 53 | 40 |
| 4 | 40 | 90 |
Step two: and obtaining performance degradation parameters of the electromechanical product under normal stress by adopting a three-step analysis method.
a) The estimates of the distribution parameters under each stress obtained by the data conversion analysis process are shown in table 2.
Table 2: product distribution parameter estimation under each stress
| Sequence number | Stress level (. Degree. C.) | Characteristic life (Tian) | Shape parameters |
| 1 | 80 | 63.69 | 1.65 |
| 2 | 66 | 77.10 | 1.76 |
| 3 | 53 | 143.50 | 1.84 |
| 4 | 40 | 147.94 | 1.83 |
The acceleration model is as follows:
lnη＝-2.9501+2517.7/S (10)
when s=s 0 At =23℃, characteristic lifetime under normal stress,
η 0 ＝257.632 (11)
c) The test time corresponding to the degradation data under high stress shown in fig. 4 is converted to normal stress, and the result of data fitting is shown in fig. 5, and comparing fig. 4 and 5 shows that the data conversion is that the time scale change is performed, and the drift coefficient value at a certain moment under high stress is converted to the drift coefficient value at a corresponding moment under normal stress, and the drift coefficient value is unchanged in the process.
Step three: and taking products 1, 2, 3 and 5 as reference samples, and carrying out residual life prediction on the product 4.
The Euclidean distances between the products 1, 2, 3, 5 and 4 are calculated by using the formula (1) in the embodiment 1, the Euclidean distances are converted into the similarity by using the formula (2) and the formula (3) in the embodiment 1, and finally the residual life prediction of the product 4 is performed by using the formula (4) and the formula (5) in the embodiment 1, and the result is shown in fig. 6.
As can be seen by combining fig. 5 and fig. 6, when the corresponding drift coefficient values of the respective products are continuously input, the remaining life of the product No. 4 can be predicted, and initially, when the time is 0-70 days, the remaining life of the product No. 1, 2, 3 and 5 has an effect on the product No. 4, so that the predicted remaining life line is between the remaining life straight lines of the product No. 3 and the remaining life straight line of the product No. 5 and is close to the position of the product straight line No. 3; after 71 days of storage time, the remaining life line of product No. 4 gradually approaches product No. 5 as product nos. 1, 2, 3 gradually fail, and the predicted remaining life curve will slowly approach the actual remaining life line as storage time increases. The product prediction method can fully utilize the data of each failure reference sample and the similarity between the failure reference sample and the product to be detected to improve the prediction precision.
The foregoing is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. While the invention has been described with reference to preferred embodiments, it is not intended to be limiting. Therefore, any simple modification, equivalent variation and modification of the above embodiments according to the technical substance of the present invention shall fall within the scope of the technical solution of the present invention.
Claims (11)
1. A product life prediction method based on path classification and estimation, comprising the steps of:
s1, path acquisition: respectively acquiring performance degradation data of a product to be detected and a plurality of failure reference samples, obtaining a target path according to the performance degradation data of the product to be detected, and obtaining a plurality of reference paths according to the performance degradation data of each failure reference sample, wherein the performance degradation data is obtained by firstly acquiring the performance degradation data under high stress of a preset size and then converting the performance degradation data into normal stress;
s2, predicting service life: respectively calculating the similarity between each reference path and the target path, and obtaining the residual life of the product to be tested by using the obtained similarity between each reference path and the target path and the residual life estimation of each failure reference sample;
in the step S2, a kernel function is used to calculate the similarity between each reference path and the target path;
and when the residual life of the product to be detected is estimated in the step S2, determining a weight value corresponding to each failure reference sample according to the similarity between each failure reference sample path and the target path, and calculating the residual life of the product to be detected according to the residual life of each failure reference sample and the corresponding weight value.
2. The method for predicting the life of a product based on path classification and estimation according to claim 1, wherein the step S1 of obtaining performance degradation data of a product to be measured and a plurality of failure reference samples comprises: performing an accelerated life test on a plurality of test samples of a product to be tested, obtaining failure data and performance degradation data of each test sample under high stress with different sizes, respectively performing data conversion to convert the data under the high stress into normal stress, taking one of the test samples as a target product and the rest as failure reference samples, obtaining the performance degradation data of the product to be tested according to the converted data obtained by the corresponding target product, and obtaining the performance degradation data of the failure reference sample according to the converted data obtained by the corresponding failure reference sample.
3. The method of path-based classification and estimation of product life prediction as claimed in claim 2, wherein said step of performing data conversion comprises:
s11, obtaining failure samples under each step of stress reduction according to the failure data and the degradation data so as to convert the step of stress reduction data into constant stress data;
s12, under the condition of constraint of distribution parameters, carrying out distribution parameter fitting on equivalent life data of each stress to obtain distribution parameter estimation;
s13, carrying out accelerated model regression analysis on the life characteristic parameter estimation and the stress level of the product by using the obtained distribution parameter estimation to obtain model parameter estimation, and completing data conversion.
4. The method for predicting the life of a product based on path classification and estimation according to any one of claims 1 to 3, wherein in the step S1, a target path is obtained by performing data fitting on the performance degradation data of the product to be detected, and a plurality of reference paths are obtained by performing data fitting on the performance degradation data of each failure reference sample.
5. The method of claim 1, wherein the step of calculating the similarity between each of the reference paths and the target path using a kernel function comprises:
s21, respectively calculating the distance between each reference path and each target path;
s22, converting the calculated distances between the reference paths and the target paths into similarity by using a preset kernel function, and obtaining the similarity between the reference paths and the target paths.
6. The path classification and estimation based product life prediction method of claim 1, wherein the kernel function is a gaussian radial basis kernel function.
7. The method for predicting the life of a product based on path classification and estimation according to claim 1, wherein the remaining life of the product to be measured is specifically calculated by the following formula:
wherein RL is a 0 To the remaining life of the product to be tested, RL i The remaining life of the ith failed reference sample, n is the number of failed reference samples, w i And the weight value corresponding to the ith failure reference sample is determined based on the similarity.
8. A product life prediction apparatus based on path classification and estimation, comprising:
the path acquisition module is used for respectively acquiring performance degradation data of a product to be detected and a plurality of failure reference samples, obtaining a target path according to the performance degradation data of the product to be detected, and obtaining a plurality of reference paths according to the performance degradation data of each failure reference sample, wherein the performance degradation data is obtained by firstly acquiring the performance degradation data under high stress of a preset size and then converting the performance degradation data into normal stress;
the life prediction module is used for calculating the similarity between each reference path and the target path respectively, and obtaining the residual life of the product to be detected by using the obtained similarity between each reference path and the target path and the residual life estimation of each failure reference sample;
the life prediction module comprises a similarity calculation submodule, a calculation submodule and a calculation submodule, wherein the similarity calculation submodule is used for calculating the similarity between each reference path and each target path by using a kernel function;
the life prediction module comprises a remaining life estimation sub-module for estimating the remaining life of the product to be detected, wherein the remaining life estimation sub-module determines a weight value corresponding to each failure reference sample according to the similarity between each failure reference sample path and the target path, and the remaining life of each failure reference sample and the corresponding weight value are calculated to obtain the remaining life of the product to be detected.
9. The device for predicting the life of a product based on path classification and estimation according to claim 8, wherein the path acquisition module includes a degradation data acquisition sub-module for acquiring performance degradation data of a product to be tested and a plurality of failure reference samples, the degradation data acquisition sub-module acquires failure data and degradation data of the product to be tested under high stress of a preset size, and performs data conversion respectively to convert the data under the high stress to normal stress, the data obtained by converting the failure data is used as the performance degradation data of the failure reference samples, and the data obtained by converting the degradation data is used as the performance degradation data of the product to be tested, and the performance degradation data of the product to be tested and the plurality of failure reference samples are acquired.
10. The path-classification and estimation-based product life prediction apparatus according to claim 9, wherein the degradation data acquisition submodule includes a conversion unit for performing data conversion, the conversion unit including:
the conversion subunit is used for obtaining failure samples under each step-down stress from the failure data and the degradation data so as to convert the step-down stress data into constant stress data;
the distribution parameter estimation subunit is used for carrying out distribution parameter fitting on the equivalent life data of each stress under the condition of distribution parameter constraint to obtain distribution parameter estimation;
and the acceleration model regression analysis subunit is used for carrying out acceleration model regression analysis on the life characteristic parameter estimation and the stress level of the product by using the obtained distribution parameter estimation to obtain model parameter estimation and complete data conversion.
11. The path-classification and estimation-based product life prediction apparatus according to claim 8 or 9 or 10, wherein the similarity calculation submodule includes:
a distance calculation unit for calculating the distance between each reference path and the target path;
and the similarity conversion unit is used for converting the calculated distances between each reference path and the target path into similarity by using a preset kernel function respectively, so as to obtain the similarity between each reference path and the target path.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811124089.7A CN110955951B (en) | 2018-09-26 | 2018-09-26 | Product life prediction method and device based on path classification and estimation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811124089.7A CN110955951B (en) | 2018-09-26 | 2018-09-26 | Product life prediction method and device based on path classification and estimation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110955951A CN110955951A (en) | 2020-04-03 |
| CN110955951B true CN110955951B (en) | 2023-12-29 |
Family
ID=69964566
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811124089.7A Active CN110955951B (en) | 2018-09-26 | 2018-09-26 | Product life prediction method and device based on path classification and estimation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110955951B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2012009549A1 (en) * | 2010-07-14 | 2012-01-19 | Baker Hughes Incorporated | System and method for estimating remaining useful life of a downhole tool |
| US8855954B1 (en) * | 2007-11-30 | 2014-10-07 | Intellectual Assets Llc | System and method for prognosticating capacity life and cycle life of a battery asset |
| CN107194478A (en) * | 2017-06-21 | 2017-09-22 | 中国人民解放军国防科学技术大学 | Merge the unit method for predicting residual useful life of lifetime data and Performance Degradation Data |
-
2018
- 2018-09-26 CN CN201811124089.7A patent/CN110955951B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8855954B1 (en) * | 2007-11-30 | 2014-10-07 | Intellectual Assets Llc | System and method for prognosticating capacity life and cycle life of a battery asset |
| WO2012009549A1 (en) * | 2010-07-14 | 2012-01-19 | Baker Hughes Incorporated | System and method for estimating remaining useful life of a downhole tool |
| CN107194478A (en) * | 2017-06-21 | 2017-09-22 | 中国人民解放军国防科学技术大学 | Merge the unit method for predicting residual useful life of lifetime data and Performance Degradation Data |
Non-Patent Citations (2)
| Title |
|---|
| 任博 ; 董兴辉 ; 郑凯 ; .基于相似性的风电机组轴承剩余寿命预测方法.机械设计与研究.2016,(04),第101-104页. * |
| 魏高乐 ; 陈志军 ; .考虑异常数据剔除的产品可靠性评估.空军工程大学学报(自然科学版).2016,(05),第31-36页. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110955951A (en) | 2020-04-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP1416348B1 (en) | Methodology for temporal fault event isolation and identification | |
| CN107885928B (en) | Stepping stress acceleration performance degradation reliability analysis method considering measurement error | |
| US10162014B2 (en) | Battery state determining device | |
| CN109765490B (en) | Power battery fault detection method and system based on high-dimensional data diagnosis | |
| JP5418408B2 (en) | Simulation parameter calibration method, apparatus and program | |
| CN107220500B (en) | Bayesian reliability evaluation method for performance degradation test based on inverse Gaussian process | |
| CN106446317A (en) | Mathematic model-based sealed relay storage life prediction method | |
| CN110110476B (en) | Sample selection and rejection method for estimating reliability based on Weibull distribution and right truncation data | |
| CN104181457A (en) | Method for selecting optimal semiconductor device temperature and humidity combined stress acceleration model | |
| CN108829878B (en) | Method and device for detecting abnormal points of industrial experimental data | |
| CN106779217A (en) | Detection of Air Quality method and air quality detection system | |
| Shintani et al. | Wafer-level variation modeling for multi-site RF IC testing via hierarchical Gaussian process | |
| CN111046331A (en) | Health monitoring method based on signal abnormity | |
| CN114970157A (en) | Method for predicting test life of small sample of electronic product under voltage stress | |
| CN109145331A (en) | Electronic product performance estimating method and device based on hardware white-box testing | |
| CN110955951B (en) | Product life prediction method and device based on path classification and estimation | |
| CN112329272B (en) | Residual life fitting prediction method integrating photovoltaic module degradation and life data | |
| CN103675712A (en) | Power supply service life detection method and system | |
| CN107677290B (en) | Testing method and device for precision evaluation of inertial navigation system | |
| KR20210129453A (en) | Apparatus and method of detecting bad data in power system measurement data | |
| CN113218537A (en) | Training method, device, equipment and storage medium of temperature anomaly detection model | |
| CN111273126B (en) | Power distribution network topology rapid sensing method | |
| CN109214447B (en) | Disk life prediction method and device | |
| CN109308395A (en) | Wafer scale space measurement abnormal parameters recognition methods based on LOF-KNN algorithm | |
| CN109359375A (en) | A kind of data product Storage Life Prediction method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |