CN116822383A - Equipment life prediction model construction method and device, readable storage medium and equipment - Google Patents
Equipment life prediction model construction method and device, readable storage medium and equipment Download PDFInfo
- Publication number
- CN116822383A CN116822383A CN202311111360.4A CN202311111360A CN116822383A CN 116822383 A CN116822383 A CN 116822383A CN 202311111360 A CN202311111360 A CN 202311111360A CN 116822383 A CN116822383 A CN 116822383A
- Authority
- CN
- China
- Prior art keywords
- signal
- storage device
- feature
- characteristic
- prediction model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000003860 storage Methods 0.000 title claims abstract description 181
- 238000010276 construction Methods 0.000 title claims description 23
- 238000000034 method Methods 0.000 claims abstract description 36
- 238000013145 classification model Methods 0.000 claims abstract description 22
- 238000010801 machine learning Methods 0.000 claims abstract description 19
- 239000011159 matrix material Substances 0.000 claims description 51
- 238000012360 testing method Methods 0.000 claims description 30
- 238000012549 training Methods 0.000 claims description 24
- 230000006870 function Effects 0.000 claims description 21
- 238000004422 calculation algorithm Methods 0.000 claims description 18
- 238000013507 mapping Methods 0.000 claims description 14
- 238000012216 screening Methods 0.000 claims description 13
- 238000004590 computer program Methods 0.000 claims description 12
- 238000000605 extraction Methods 0.000 claims description 11
- 230000009467 reduction Effects 0.000 claims description 9
- 238000011478 gradient descent method Methods 0.000 claims description 5
- 238000012163 sequencing technique Methods 0.000 claims description 5
- 239000013598 vector Substances 0.000 claims description 5
- 238000001914 filtration Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 abstract description 8
- 230000000875 corresponding effect Effects 0.000 description 73
- 238000004364 calculation method Methods 0.000 description 6
- 238000000513 principal component analysis Methods 0.000 description 6
- 238000012795 verification Methods 0.000 description 5
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 230000001419 dependent effect Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000009826 distribution Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000007418 data mining Methods 0.000 description 2
- 238000012886 linear function Methods 0.000 description 2
- 238000012417 linear regression Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 230000015556 catabolic process Effects 0.000 description 1
- 238000007635 classification algorithm Methods 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
Abstract
The invention discloses a method, a device, a readable storage medium and equipment for constructing a life prediction model of equipment, which are characterized in that after a storage equipment sample is extracted and corresponding characteristic information is acquired, signal characteristics in the characteristic information are further extracted, namely, the characteristic information is subjected to secondary processing, so that signal characteristics which are more relevant to the life of the storage equipment can be extracted, then the relevance between the signal characteristics in a signal characteristic set and the life of the storage equipment is calculated, the signal characteristics in the signal characteristic set are screened according to the relevance of the life, finally, a characteristic subset which is highly relevant to the life of the storage equipment is extracted, and a machine learning classification model is trained based on the characteristic subset, so that the life prediction model with high precision is obtained, and the accuracy of equipment fault prediction is improved.
Description
Technical Field
The present invention relates to the field of storage device lifetime prediction technologies, and in particular, to a device lifetime prediction model construction method and apparatus, a readable storage medium, and a device.
Background
With the progress of semiconductor manufacturing processes, memory devices have been advancing into the era of ultra-high density and ultra-large capacity. The problem of serious reliability degradation caused by the reduced size and increased number of stacked layers of the memory device is to be solved. The service life of the storage device is an important parameter for judging the reliability of the storage device. The lifetime represents the number of operations that a storage device can perform a read-write function before failing. Therefore, the service life of the storage device is predicted, the loss of the storage device, such as data, economy and the like, caused by the failure of the storage device can be effectively avoided, and meanwhile, an effective use strategy can be formulated according to the predicted service life, so that the data function of the storage device product can be exerted to the maximum extent.
Currently, existing storage device lifetime prediction techniques typically use machine learning classification training by collecting storage device feature values. But generalization is weak and it is difficult to give a reasonable response to a new data set added on the data set. Moreover, since the distance between the reference point and each sample point needs to be calculated through traversal, when the data amount is large, a large amount of calculation amount is generated, and the prediction speed and the accuracy of the service life are further affected.
Disclosure of Invention
The technical problems to be solved by the invention are as follows: provided are a device life prediction model construction method, a device, a readable storage medium and electronic equipment, and accuracy of device failure prediction is improved.
In order to solve the technical problems, the invention adopts the following technical scheme:
the equipment life prediction model construction method comprises the following steps:
extracting a storage device sample, and acquiring characteristic information corresponding to the storage device sample;
extracting signal characteristics in the characteristic information to obtain a signal characteristic set;
calculating the correlation degree between the signal characteristics in the signal characteristic set and the service life of the storage equipment, and obtaining signal characteristic value ordering;
sorting and screening signal features in the signal feature set according to the signal feature value to obtain a feature subset;
And inputting the feature subset into a machine learning classification model for training to obtain a life prediction model.
In order to solve the technical problems, the invention adopts another technical scheme that:
an apparatus for constructing a device lifetime prediction model, comprising: the extraction module is used for extracting a storage device sample and acquiring characteristic information corresponding to the storage device sample; the extraction module is used for extracting signal characteristics in the characteristic information to obtain a signal characteristic set; the calculating module is used for calculating the correlation degree between the signal characteristics in the signal characteristic set and the service life of the storage equipment and obtaining signal characteristic value ordering; the screening module is used for screening the signal characteristics in the signal characteristic set according to the signal characteristic value sequence to obtain a characteristic subset; and the construction module inputs the feature subset into a machine learning classification model for training to obtain a life prediction model.
In order to solve the technical problems, the invention adopts another technical scheme that:
a computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the apparatus life prediction model construction method as described above.
In order to solve the technical problems, the invention adopts another technical scheme that:
an electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps in a device lifetime prediction model building method as described above when the computer program is executed.
The invention has the beneficial effects that: after the storage device sample is extracted and the corresponding characteristic information is obtained, signal characteristics in the characteristic information are further extracted, namely, the characteristic information is subjected to secondary processing, so that signal characteristics which are more relevant to the service life of the storage device can be extracted, the relevance of the signal characteristics in the signal characteristic set and the service life of the storage device is calculated, the signal characteristics in the signal characteristic set are screened according to the relevance of the service life, a characteristic subset which is highly relevant to the service life of the storage device is finally extracted, and a machine learning classification model is trained based on the characteristic subset, so that a high-precision service life prediction model is obtained, and the accuracy of equipment fault prediction is improved.
Drawings
FIG. 1 is a flow chart of steps of a method for constructing a life prediction model of an apparatus according to an embodiment of the present invention;
FIG. 2 is a flowchart showing another step of a method for constructing a life prediction model of an apparatus according to an embodiment of the present invention;
FIG. 3 is a deviation chart of feature values corresponding to different K values in an equipment life prediction model construction method according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of a linear function calculation result of lifetime analysis in a method for constructing a lifetime prediction model of an apparatus according to an embodiment of the present invention;
FIG. 5 is a comparative image of actual life and predicted life in a method for constructing a model for predicting equipment life according to an embodiment of the present invention;
FIG. 6 is a schematic structural diagram of an apparatus for constructing a life prediction model of a device according to an embodiment of the present invention;
fig. 7 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
In order to describe the technical contents, the achieved objects and effects of the present invention in detail, the following description will be made with reference to the embodiments in conjunction with the accompanying drawings.
Referring to fig. 1, the method for constructing the equipment life prediction model includes the steps of:
extracting a storage device sample, and acquiring characteristic information corresponding to the storage device sample;
extracting signal characteristics in the characteristic information to obtain a signal characteristic set;
calculating the correlation degree between the signal characteristics in the signal characteristic set and the service life of the storage equipment, and obtaining signal characteristic value ordering;
Sorting and screening signal features in the signal feature set according to the signal feature value to obtain a feature subset;
and inputting the feature subset into a machine learning classification model for training to obtain a life prediction model.
From the above description, the beneficial effects of the invention are as follows: after the storage device sample is extracted and the corresponding characteristic information is obtained, signal characteristics in the characteristic information are further extracted, namely, the characteristic information is subjected to secondary processing, so that signal characteristics which are more relevant to the service life of the storage device can be extracted, the relevance of the signal characteristics in the signal characteristic set and the service life of the storage device is calculated, the signal characteristics in the signal characteristic set are screened according to the relevance of the service life, a characteristic subset which is highly relevant to the service life of the storage device is finally extracted, and a machine learning classification model is trained based on the characteristic subset, so that a high-precision service life prediction model is obtained, and the accuracy of equipment fault prediction is improved.
Further, the characteristic information includes a programming time;
the obtaining the characteristic information corresponding to the storage device sample comprises the following steps:
dividing test nodes, reading the time used by the write operation corresponding to the storage device sample at each test node, and stopping recording the cycle number of the received return data;
The programming time is obtained from the time taken for the write operation and the number of cycles that the data stopped recording.
As can be seen from the above description, by dividing the test node into the time for the storage device sample to perform the write operation and the cycle number for which the return data is received to stop recording, the programming time of the storage device sample at the corresponding node can be accurately obtained, the validity of the feature information is ensured, and the programming time can be used as one of the prediction elements in the subsequent life prediction to predict the life of the storage device.
Further, the characteristic information includes an erasure time;
the obtaining the characteristic information corresponding to the storage device sample comprises the following steps:
dividing test nodes, and reading the time used by the erasing operation corresponding to the storage device sample and the period number for which the erasing operation is continuous at each test node;
and obtaining the erasing time according to the time used by the erasing operation and the period number of the duration of the erasing operation.
As can be seen from the above description, by dividing the test node into the time taken by the storage device sample to perform the erase operation and the number of periods during which the erase operation is continued, the erase time of the storage device sample at the corresponding node can be accurately obtained, the validity of the feature information is ensured, and the erase time can be used as one of the prediction elements in the subsequent life prediction to predict the life of the storage device.
Further, the characteristic information includes an error rate;
the obtaining the characteristic information corresponding to the storage device sample comprises the following steps:
sequentially performing read-write operation on each storage unit in the storage device sample, counting the number of storage units with differences between the written data and the read data, and counting the total number of the storage units;
and obtaining the error rate according to the number of the storage units with the difference and the total number of the storage units.
As is clear from the above description, the read-write error rate of the storage device is taken as one of the elements for predicting the lifetime of the storage device, so that the lifetime of the storage device can be determined based on the read-write error rate; when the error rate is obtained, the difference between the read data and the write data is compared in the operation process of reading the write data by the storage units, and the number of the storage units with the difference and the total number of the storage units are counted, so that the error rate corresponding to the storage device sample is obtained.
Further, the characteristic information includes a margin current;
the obtaining the characteristic information corresponding to the storage device sample comprises the following steps:
judging whether the storage device sample completes the erasing operation or not, if so, applying a preset control voltage to a control gate of the storage device sample;
And reading a current value between the drain electrode and the source electrode of the storage device sample to obtain the residual current.
As is apparent from the above description, the residual current of the memory device after being erased is taken as one of the elements for predicting the life of the memory device, so that the life of the memory device can be determined based on the residual current; when the residual current is obtained, after a preset control voltage is applied to the control gate of the storage device sample, the current value between the drain electrode and the source electrode of the storage device sample is recorded, so that the residual current corresponding to the storage device sample is obtained.
Further, before the preset control voltage is applied to the control gate of the storage device sample, the method further comprises:
setting at least one threshold time;
judging whether the time to be operated corresponding to the storage device sample after the erasing operation is completed reaches the threshold time, if so, executing the step of applying a preset control voltage to the control gate of the storage device sample.
As is apparent from the above description, the margin current of the memory device after a plurality of different wait time values have been erased is taken as one of the elements for predicting the lifetime of the memory device, thereby improving the accuracy of predicting the lifetime of the memory device based on the margin current.
Further, the extracting the signal features in the feature information to obtain a signal feature set includes:
extracting corresponding time domain signal characteristics from the time domain of the characteristic information;
extracting corresponding frequency domain signal characteristics from the frequency domain of the characteristic information;
and obtaining the signal feature set according to the time domain signal features and the frequency domain signal features.
As can be seen from the above description, the time domain signal features and the frequency domain signal features corresponding to the feature information are extracted from the time domain and the frequency domain respectively; therefore, the signal characteristic sets which are highly correlated with the service life of the storage equipment and are mutually uncorrelated can be extracted from the characteristic information, the accuracy of the training set is improved, and further, the model prediction is more accurate.
Further, the calculating the correlation between the signal features in the signal feature set and the service life of the storage device, and obtaining the signal feature value ranking includes:
forming an original matrix with a preset row-column size by the signal features in the signal feature set;
zero-equalizing the original matrix to obtain a zero-equalizing matrix;
calculating a covariance matrix corresponding to the zero-mean matrix;
calculating the eigenvalue of the covariance matrix and the eigenvector corresponding to the eigenvalue;
And sequencing the signal features according to the magnitude of the feature values corresponding to the signal features to obtain the signal feature value sequencing.
From the above description, it can be seen that by forming the signal features in the signal feature set into an original matrix with a preset rank size and zero-equalizing the original matrix, sequentially calculating to obtain a covariance matrix corresponding to the zero-equalized matrix, feature values of the covariance matrix and feature vectors corresponding to the feature values, the signal features are accurately ordered according to the sizes of the feature values corresponding to the signal features.
Further, the filtering the signal features in the signal feature set according to the signal feature value ranking, and obtaining the feature subset includes:
acquiring a preset number of feature vectors according to the magnitude of the feature value corresponding to the signal feature to form a feature matrix according to rows;
performing dimension reduction on the original matrix according to the feature matrix to obtain a dimension reduction matrix;
and taking the dimension reduction matrix as the feature subset.
From the above description, the feature vectors preset before the feature values corresponding to the signal features are obtained form the feature matrix according to the rows, so that the screening and redundancy processing of the signal features are realized, the main content of the data is ensured not to be lost, the data quantity of the signal features is reduced, and the training efficiency and accuracy of the model are improved.
Further, the inputting the feature subset into a machine learning classification model for training, and obtaining a life prediction model comprises:
performing feature classification on the feature subset through a K nearest neighbor or K-dimensional tree algorithm to obtain a mapping relation between a feature value corresponding to the signal feature in the feature subset and the service life of the storage device;
and fitting the mapping relation between the characteristic value corresponding to the signal characteristic and the service life of the storage equipment according to a regression algorithm to obtain a service life prediction model.
From the description, the K nearest neighbor or K-dimensional tree algorithm is adopted to classify the features of the feature subset, so that different requirements under different application scenes are met, and the practicability of the method is improved.
Further, the fitting the mapping relation between the feature value corresponding to the signal feature and the service life of the storage device according to the regression algorithm to obtain a service life prediction model includes:
establishing a life prediction function, and taking a characteristic value corresponding to the signal characteristic as a function independent variable;
calculating a mapping relation between a characteristic value corresponding to the signal characteristic and the service life of the storage equipment by a least square method or a gradient descent method to obtain an independent variable coefficient;
And carrying out regression fitting according to the independent variable and the independent variable coefficient to obtain the life prediction model.
As can be seen from the above description, after the mapping relationship between the characteristic value corresponding to the signal characteristic and the service life of the storage device is obtained, the mapping relationship between the characteristic value corresponding to the signal characteristic and the service life of the storage device is calculated by a least square method or a gradient descent method to obtain the independent variable coefficient, so that the functional relationship between the service life of the storage device and the signal characteristic can be fitted, and the effect of accurately predicting the service life fault of the storage device according to the characteristic corresponding to the storage device to be predicted is achieved.
Further, the characteristic information includes a storage device class;
the establishing a life prediction function includes:
and establishing a life prediction function corresponding to the storage equipment type according to the storage equipment type corresponding to the characteristic information.
As is clear from the above description, the corresponding lifetime prediction function is established based on different storage device types, so that lifetime of the storage device such as eMMC, UFS, BGASSD can be predicted according to different lifetime prediction functions.
Another embodiment of the present invention provides an apparatus for constructing a life prediction model of a device, including: the extraction module is used for extracting a storage device sample and acquiring characteristic information corresponding to the storage device sample; the extraction module is used for extracting signal characteristics in the characteristic information to obtain a signal characteristic set; the calculating module is used for calculating the correlation degree between the signal characteristics in the signal characteristic set and the service life of the storage equipment and obtaining signal characteristic value ordering; the screening module is used for screening the signal characteristics in the signal characteristic set according to the signal characteristic value sequence to obtain a characteristic subset; and the construction module inputs the feature subset into a machine learning classification model for training to obtain a life prediction model.
Another embodiment of the present invention provides a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the device lifetime prediction model construction method as described above.
Another embodiment of the present invention provides an electronic device including a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor implementing the steps in the device lifetime prediction model building method as described above when executing the computer program.
The method, the device, the computer-readable storage medium and the electronic apparatus for constructing the lifetime model of the device according to the present invention are applicable to lifetime prediction of various storage devices, for example, lifetime prediction of a storage device such as eMMC, UFS, BGASSD, and will be described in detail below by way of specific embodiments:
example 1
Referring to fig. 1 and 2, a method for constructing a life prediction model of an apparatus includes the steps of:
s1, extracting a storage device sample, and acquiring characteristic information corresponding to the storage device sample, wherein the characteristic information is specifically: if the service lives of the memory devices of the eMMC, the UFS and the BGA SSD are predicted, respectively extracting memory device samples of the eMMC, the UFS and the BGA SSD as training sets, specifically: and extracting samples of eMMC, UFS and BGA SSD manufactured in different batches at different times under the same manufacturing process level, such as TLCNAND under a certain manufacturing process. The extraction mode is completely random, the number of samples is 2% of the total amount of the sampled batch equipment, for example, 2000 products are extracted respectively as machine learning input samples; and the sample number is not fixed, the configuration can be adjusted according to the machine learning precision requirement, and the proportion of the training set to the verification set is changed. In the life test process of the storage device, the storage device test platform executes erasing operation on the storage blocks of the storage device until the storage device reaches the life limit, and the test system counts the total programming/erasing operation cycle number of the storage device as the life of the storage device.
The characteristic information corresponding to the storage device sample includes, but is not limited to, a storage device chip corresponding to: programming time, reading time, erasing time, current, operating frequency, chip power consumption, threshold voltage distribution, memory block number, memory page number, number of program/erase cycles currently experienced by a memory device chip, number of conditional error pages, number of conditional error blocks, number of error bits and error rate, and margin current after sampling a plurality of different latency values of the memory device after being erased.
The specific implementation manner of acquiring the specific feature information is described below by taking the program time, the erase time, the error rate and the residual current feature as examples:
the obtaining programming time corresponding to the storage device sample comprises the following steps: dividing test nodes, reading the time used by the write operation corresponding to the storage device sample at each test node, and stopping recording the cycle number of the received return data; the programming time is obtained from the time taken for the write operation and the number of cycles that the data stopped recording.
The obtaining the erasing time corresponding to the storage device sample comprises the following steps: dividing test nodes, and reading the time used by the erasing operation corresponding to the storage device sample and the period number for which the erasing operation is continuous at each test node; and obtaining the erasing time according to the time used by the erasing operation and the period number of the duration of the erasing operation.
The obtaining the error rate corresponding to the storage device sample comprises the following steps: sequentially performing read-write operation on each storage unit in the storage device sample, counting the number of storage units with differences between the written data and the read data, and counting the total number of the storage units; and obtaining the error rate according to the number of the storage units with the difference and the total number of the storage units.
Obtaining residual currents corresponding to the storage device samples after the erased plurality of different waiting time values comprises: setting at least one threshold time; judging whether the time to be operated corresponding to the storage device sample after the erasing operation is completed reaches the threshold time, if so, judging whether the storage device sample is completed with the erasing operation, if so, applying a preset control voltage to a control gate of the storage device sample, and reading a current value between a drain electrode and a source electrode of the storage device sample to obtain the residual current.
S2, extracting signal features in the feature information to obtain a signal feature set, wherein the step comprises the following steps:
s21, extracting corresponding time domain signal features from the time domain of the feature information; the time domain signal characteristics comprise the mean value, variance and skewness of the measured amplitude corresponding to the characteristic information.
S22, extracting corresponding frequency domain signal characteristics from the frequency domain of the characteristic information; the frequency domain signal characteristics include power in calculating different frequency ranges from the characteristic information, and other frequency-dependent parameter information.
S23, obtaining the signal characteristic set according to the time domain signal characteristics and the frequency domain signal characteristics, namely finally obtaining a characteristic parameter set possibly related to the service life of storage equipment, and finishing secondary processing of the signal characteristics; if 56 groups of characteristics of the time domain signal characteristics and the frequency domain signal characteristics are obtained through calculation, three groups of 2000 x 56 signal characteristic sets are finally obtained; wherein the number of groups (56 groups) of the signal feature set is not fixed and can be adjusted according to the machine learning accuracy requirement.
S3, calculating the correlation degree between the signal characteristics in the signal characteristic set and the service life of the storage equipment, and obtaining signal characteristic value sequencing; because the data dimension of the signal feature set is quite large, the signal features in the signal feature set need to be selectively removed, and redundant data is processed, specifically:
s31, forming the signal characteristics in the signal characteristic set into an original matrix with a preset row-column size; forming an original matrix X of n rows and m columns by using the signal features in the signal feature set;
S32, carrying out zero-mean on the original matrix to obtain a zero-mean matrix; zero-equalizing each row of the original matrix X, namely subtracting the average value of the row; wherein each row of the original matrix X represents an attribute field;
s33, calculating a covariance matrix corresponding to the zero-mean matrix; i.e. covariance matrix c=1/mxx ζ, where C is p-x p-dimensional symmetric matrix;
s34, calculating the eigenvalue of the covariance matrix and the eigenvector corresponding to the eigenvalue; if the eigenvalues lambda 1, lambda 2 … lambda p and the corresponding eigenvectors are obtained;
and S35, sorting the signal features according to the magnitude of the feature values corresponding to the signal features to obtain the signal feature value sorting.
S4, sorting and screening the signal features in the signal feature set according to the signal feature value to obtain a feature subset, wherein the method comprises the following steps:
s41, acquiring a preset number of feature vectors according to the magnitude of the feature value corresponding to the signal feature to form a feature matrix according to the rows; the magnitudes of the eigenvalues lambda 1, lambda 2 … lambda p are arranged into a matrix from top to bottom in rows;
s42, performing dimension reduction on the original matrix according to the feature matrix to obtain a dimension reduction matrix, and specifically: selecting k rows of which the magnitudes are arranged into a matrix from top to bottom according to the sizes of the eigenvalues lambda 1 and lambda 2 … lambda P to form a matrix P, wherein Y=PX is the data after the dimension is reduced to k dimensions; if the dimension of the matrix is greatly reduced (i.e. the k value is too small), the main content of the data is easily lost; referring to fig. 3, for the distribution of the eigenvalues obtained by one principal component analysis, it can be seen that several eigenvalues at the front end of the curve are larger but decay rate is fast, while those at the rear end are smaller but decay is slow; selecting the inflection point (i.e., k=3) as the k value can secure an optimal effect according to the nature of the image inflection point.
S43, taking the dimension reduction matrix as the feature subset; taking the extracted feature subset as a training set of a machine learning classification model; dividing the training set into a verification set and a test set;
s5, inputting the feature subset into a machine learning classification model for training to obtain a life prediction model; specific: and constructing a classification model and a regression model according to the test set, namely randomly dividing 2000 characteristic samples in each group of signal characteristic sets into two groups, wherein the two groups comprise 1500 signal characteristic samples in a training group and 500 signal characteristic samples in a test group. When the program runs, each signal characteristic in the test group is selected and put into a coordinate system; wherein, the signal characteristics are reduced to only three dimensions after the secondary processing. And constructing a verification classification module and a verification regression model through the verification set, and verifying the classification model and the regression model. Simultaneously, before training, each signal characteristic sample in the training set is placed in a coordinate system by taking each value as an axis; the samples in the test set are then placed one by one in the coordinate system, several closest training samples are found, and the test point is classified by calculating the largest category to which the closest training samples belong. The regression model is to fit a regression algorithm to calculate each coefficient of the dependent variable (the service life corresponding to the storage device sample) and the independent variable (the characteristic information corresponding to the storage device sample) to generate the relationship between the input and the output of the model, so as to achieve the purpose of predicting the service life of the device according to the physical quantity data obtained intermittently/from the factory.
Example two
The difference between the present embodiment and the first embodiment is that the generation mode of the model is specifically defined;
s5, inputting the feature subset into a machine learning classification model for training, and obtaining a life prediction model comprises the following steps:
s51, classifying the characteristics of the characteristic subsets through a K nearest neighbor or K-dimensional tree algorithm to obtain a mapping relation between characteristic values corresponding to the signal characteristics in the characteristic subsets and the service life of the storage equipment;
the method is described by taking the construction of a K-dimensional tree algorithm model (e.g., a sample has n=3 features), and is specifically:
s511, calculating variances of the n features respectively, selecting one feature with the largest variance, and assuming that the variance is a;
s512, for the feature a, calculating a median value (mean), and taking the median value as a sample dividing point, namely a root node of the tree;
s513, dividing other data in the sample set according to the characteristic a, dividing all samples smaller than the median value into a left subtree, and dividing all samples larger than the median value into a right subtree;
s514, recursively performing the steps on the left subtree and the right subtree respectively until the number of samples of the left subtree and the right subtree is not more than the number specified by leaf_size (subtree branch).
And when the test group performs the test, backtracking and searching are performed according to the characteristic quantity, and the distance between the test group and each node is calculated until the distance between the test group and the current node is smaller than the distance between the test group and the previous node, namely, the characteristic quantity is judged to belong to the characteristic of the node. Compared with a K Nearest Neighbor (KNN) classification algorithm, the K-dimensional tree algorithm needs to calculate Euclidean distances with all nodes to perform a majority voting method, and the number of times of Euclidean distance calculation performed by the Kd-tree is far smaller than that of the KNN algorithm.
And S52, fitting the mapping relation between the characteristic value corresponding to the signal characteristic and the service life of the storage device according to a regression algorithm to obtain a service life prediction model. Wherein, a life prediction function corresponding to the storage device type needs to be established according to the storage device type corresponding to the feature information, namely life state information is extracted from life related features, a regression model needs to be established for life prediction features of the storage device of three types eMMC, UFS, BGASSD respectively, one regression model can generate a relation between input and output of the model by setting each value in the input as a variable and then calculating each coefficient of the variable, so that after the model is established and the input value is set, the model can multiply each variable with the coefficient calculated by the variable, thereby obtaining the output, and the specific:
S521, establishing a life prediction function, and taking a characteristic value corresponding to the signal characteristic as a function independent variable; namely, taking a characteristic value corresponding to the signal characteristic as a function independent variable to obtain an input/output function;
s522, calculating a mapping relation between a characteristic value corresponding to the signal characteristic and the service life of the storage equipment by a least square method or a gradient descent method to obtain an independent variable coefficient; the least square method takes all inputs of a training sample as an input matrix, takes all outputs as an output matrix, and calculates coefficients of each input, namely independent variable coefficients through the matrix. The gradient descent method further establishes a cost function of the difference between the actual value and the predicted value, and obtains the most suitable coefficient by minimizing the cost function. In this embodiment, the total number of program/erase operation cycles is extracted, that is, the life of the storage device is taken as the output of the model, and then the corresponding feature of the storage device is taken as the input of the model; wherein each feature is set to a variable, and the model is therefore referred to as a multiple linear regression model.
S523, performing regression fitting according to the independent variable and the independent variable coefficient to obtain the life prediction model; a set of coefficients is calculated by using a multi-linear regression model, and then the calculated coefficients are multiplied by the characteristic values to obtain predicted life conditions.
Example III
In order to verify the prediction accuracy of the life prediction model, a specific experimental case is provided for illustration in the example; in this embodiment, a feature extraction method is used: principal Component Analysis (PCA) and two classification methods: k nearest neighbor and K-dimensional tree; and respectively establishing a classification model for comparison of the memory devices of the eMMC type, the UFS type and the BGA SSD type. And establishing a corresponding model aiming at the binary problem of the life prediction of the storage equipment. Initially, the K-nearest neighbor and K-dimensional tree classifier uses the original dataset (i.e., 56 features) for classifier construction; then filtering all 56 features by using a PCA algorithm, and testing a classifier by using the filtered optimal feature set; the classification test results are shown in tables 1 and 2.
TABLE 1 accuracy of classification model
TABLE 2 runtime of classification models
Wherein tables 1 and 2 compare the run time and classification accuracy of the classification model on the dataset and the screened dataset, respectively. As can be seen from tables 1 and 2: 1. the accuracy of the classification model constructed using PCA is significantly improved over the original dataset, i.e., the classifier applying the feature selection dataset has overall higher accuracy than the classifier directly applying the original data. 2. The Kd-tree model based on PCA can achieve an optimal correct classification rate of up to 98% (BGASSD type) in binary classification problems. And can all carry out the classification of comparison accuracy to eMMC, UFS, three kinds of storage device of BGA SSD. Meanwhile, the Kd-tree greatly improves the problems of large calculation amount and long operation time caused by KNN traversal, namely the PCA-Kdtree model is a more effective model.
The model is further analyzed based on the regression model:
referring to fig. 4, as a result of the linear function calculation for the lifetime analysis, it is obvious that almost all samples are calculated along the function with small error.
Referring to fig. 5, a comparison image of actual life and predicted life is shown; the abscissa is the number of samples, and the ordinate is the index corresponding to different features, such as the mean, variance and bias of the measured amplitudes corresponding to BGASSD sample feature information (programming time, reading time, erasing time, current, operating frequency, chip power consumption, threshold voltage distribution, memory block number, memory page number, number of programming/erasing cycles currently experienced by the memory device chip, number of conditional error pages, number of conditional error blocks, number of error bits and error rate, residual current of the sampling memory device after being erased with a plurality of different waiting time values, etc.), where the mean of the amplitudes corresponding to the first ten features is selected. It can be seen that the model can well predict the service life of the storage device and has high precision. Goodness of fit R2 (rζ2= Σ (y prediction-) 2/= Σ (yactually->) 2) is 0.989, wherein the closer R2 is to 1, the higher the fitting accuracy, i.e. the model has higher accuracy.
Example IV
Referring to fig. 6, an apparatus for constructing a device lifetime prediction model includes:
the extraction module is used for extracting a storage device sample and acquiring characteristic information corresponding to the storage device sample;
the extraction module is used for extracting signal characteristics in the characteristic information to obtain a signal characteristic set;
the calculating module is used for calculating the correlation degree between the signal characteristics in the signal characteristic set and the service life of the storage equipment and obtaining signal characteristic value ordering;
the screening module is used for screening the signal characteristics in the signal characteristic set according to the signal characteristic value sequence to obtain a characteristic subset;
and the construction module inputs the feature subset into a machine learning classification model for training to obtain a life prediction model.
Example five
A computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of a device lifetime prediction model construction method as described in the first and second embodiments.
Example six
Referring to fig. 7, an electronic device includes a memory, a processor and a computer program stored in the memory and executable on the processor, where the processor implements the steps of a device lifetime prediction model construction method as described in the first and second embodiments when executing the computer program.
In summary, the method and the device for constructing the equipment life prediction model, the readable storage medium and the electronic equipment provided by the application are used for predicting the life of the storage equipment by constructing the data mining classification and regression model based on the K-dimensional tree algorithm. The extracted features of eMMC, UFS and BGA SSD type storage devices are subjected to secondary processing, the value of the features is estimated and sequenced by using a feature extraction algorithm, a feature subset of a feature height set is finally extracted and input into a K-dimensional tree algorithm for data mining, the mapping relation between the service life of the storage devices and the extracted feature values is obtained, each coefficient of a dependent variable and independent variable is calculated through fitting of a regression algorithm to generate the relation between the input and the output of a model, and the purpose of predicting the service life of the device according to physical quantity data obtained by intermittent/product delivery is achieved.
In the foregoing embodiments of the present application, it should be understood that the disclosed method, apparatus, computer readable storage medium and electronic device may be implemented in other manners. For example, the apparatus embodiments described above are merely illustrative, and for example, the division of the modules is merely a logical function division, and there may be additional divisions when actually implemented, for example, multiple components or modules may be combined or integrated into another apparatus, or some features may be omitted or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with respect to each other may be an indirect coupling or communication connection via some interfaces, devices or components or modules, which may be in electrical, mechanical, or other forms.
The components illustrated as separate components may or may not be physically separate, and components shown as components may or may not be physical modules, i.e., may be located in one place, or may be distributed over multiple network modules. Some or all of the components may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional module in each embodiment of the present invention may be integrated into one processing module, or each component may exist alone physically, or two or more modules may be integrated into one module. The integrated modules may be implemented in hardware or in software functional modules.
The integrated modules, if implemented in the form of software functional modules and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied essentially or in part or all of the technical solution or in part in the form of a software product stored in a storage medium, including instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
It should be noted that, for the sake of simplicity of description, the foregoing method embodiments are all expressed as a series of combinations of actions, but it should be understood by those skilled in the art that the present invention is not limited by the order of actions described, as some steps may be performed in other order or simultaneously in accordance with the present invention. Further, those skilled in the art will appreciate that the embodiments described in the specification are all preferred embodiments, and that the acts and modules referred to are not necessarily all required for the present invention.
In the foregoing embodiments, the descriptions of the embodiments are emphasized, and for parts of one embodiment that are not described in detail, reference may be made to the related descriptions of other embodiments.
The foregoing description is only illustrative of the present invention and is not intended to limit the scope of the invention, and all equivalent changes made by the specification and drawings of the present invention, or direct or indirect application in the relevant art, are included in the scope of the present invention.
Claims (15)
1. The equipment life prediction model construction method is characterized by comprising the following steps:
extracting a storage device sample, and acquiring characteristic information corresponding to the storage device sample;
Extracting signal characteristics in the characteristic information to obtain a signal characteristic set;
calculating the correlation degree between the signal characteristics in the signal characteristic set and the service life of the storage equipment, and obtaining signal characteristic value ordering;
sorting and screening signal features in the signal feature set according to the signal feature value to obtain a feature subset;
and inputting the feature subset into a machine learning classification model for training to obtain a life prediction model.
2. The apparatus life prediction model construction method according to claim 1, wherein the characteristic information includes a programming time;
the obtaining the characteristic information corresponding to the storage device sample comprises the following steps:
dividing test nodes, reading the time used by the write operation corresponding to the storage device sample at each test node, and stopping recording the cycle number of the received return data;
the programming time is obtained from the time taken for the write operation and the number of cycles that the data stopped recording.
3. The apparatus life prediction model construction method according to claim 1, wherein the characteristic information includes an erasure time;
the obtaining the characteristic information corresponding to the storage device sample comprises the following steps:
Dividing test nodes, and reading the time used by the erasing operation corresponding to the storage device sample and the period number for which the erasing operation is continuous at each test node;
and obtaining the erasing time according to the time used by the erasing operation and the period number of the duration of the erasing operation.
4. The apparatus life prediction model construction method according to claim 1, wherein the characteristic information includes an error rate;
the obtaining the characteristic information corresponding to the storage device sample comprises the following steps:
sequentially performing read-write operation on each storage unit in the storage device sample, counting the number of storage units with differences between the written data and the read data, and counting the total number of the storage units;
and obtaining the error rate according to the number of the storage units with the difference and the total number of the storage units.
5. The equipment life prediction model construction method according to claim 1, wherein the characteristic information includes a residual current;
the obtaining the characteristic information corresponding to the storage device sample comprises the following steps:
judging whether the storage device sample completes the erasing operation or not, if so, applying a preset control voltage to a control gate of the storage device sample;
And reading a current value between the drain electrode and the source electrode of the storage device sample to obtain the residual current.
6. The method for constructing a life prediction model of a device according to claim 5, wherein before applying a predetermined control voltage to the control gate of the storage device sample, further comprises:
setting at least one threshold time;
judging whether the time to be operated corresponding to the storage device sample after the erasing operation is completed reaches the threshold time, if so, executing the step of applying a preset control voltage to the control gate of the storage device sample.
7. The method for constructing a life prediction model of a device according to claim 1, wherein extracting signal features in the feature information to obtain a signal feature set includes:
extracting corresponding time domain signal characteristics from the time domain of the characteristic information;
extracting corresponding frequency domain signal characteristics from the frequency domain of the characteristic information;
and obtaining the signal feature set according to the time domain signal features and the frequency domain signal features.
8. The method of claim 1, wherein the calculating the correlation between the signal features in the signal feature set and the lifetime of the storage device, and obtaining the signal feature value ranking comprises:
Forming an original matrix with a preset row-column size by the signal features in the signal feature set;
zero-equalizing the original matrix to obtain a zero-equalizing matrix;
calculating a covariance matrix corresponding to the zero-mean matrix;
calculating the eigenvalue of the covariance matrix and the eigenvector corresponding to the eigenvalue;
and sequencing the signal features according to the magnitude of the feature values corresponding to the signal features to obtain the signal feature value sequencing.
9. The method of claim 8, wherein said filtering signal features in said signal feature set according to said signal feature value ranking to obtain feature subsets comprises:
acquiring a preset number of feature vectors according to the magnitude of the feature value corresponding to the signal feature to form a feature matrix according to rows;
performing dimension reduction on the original matrix according to the feature matrix to obtain a dimension reduction matrix;
and taking the dimension reduction matrix as the feature subset.
10. The method of claim 1, wherein inputting the subset of features into a machine learning classification model for training to obtain a life prediction model comprises:
Performing feature classification on the feature subset through a K nearest neighbor or K-dimensional tree algorithm to obtain a mapping relation between a feature value corresponding to the signal feature in the feature subset and the service life of the storage device;
and fitting the mapping relation between the characteristic value corresponding to the signal characteristic and the service life of the storage equipment according to a regression algorithm to obtain a service life prediction model.
11. The method for constructing a life prediction model of a device according to claim 10, wherein the fitting the mapping relationship between the feature value corresponding to the signal feature and the life of the storage device according to the regression algorithm to obtain the life prediction model includes:
establishing a life prediction function, and taking a characteristic value corresponding to the signal characteristic as a function independent variable;
calculating a mapping relation between a characteristic value corresponding to the signal characteristic and the service life of the storage equipment by a least square method or a gradient descent method to obtain an independent variable coefficient;
and carrying out regression fitting according to the independent variable and the independent variable coefficient to obtain the life prediction model.
12. The device life prediction model construction method according to claim 11, wherein the characteristic information includes a storage device class;
The establishing a life prediction function includes:
and establishing a life prediction function corresponding to the storage equipment type according to the storage equipment type corresponding to the characteristic information.
13. A device life prediction model construction apparatus, comprising:
the extraction module is used for extracting a storage device sample and acquiring characteristic information corresponding to the storage device sample;
the extraction module is used for extracting signal characteristics in the characteristic information to obtain a signal characteristic set;
the calculating module is used for calculating the correlation degree between the signal characteristics in the signal characteristic set and the service life of the storage equipment and obtaining signal characteristic value ordering;
the screening module is used for screening the signal characteristics in the signal characteristic set according to the signal characteristic value sequence to obtain a characteristic subset;
and the construction module inputs the feature subset into a machine learning classification model for training to obtain a life prediction model.
14. A computer readable storage medium having stored thereon a computer program, which when executed by a processor performs the steps of the method for constructing an equipment life prediction model according to any one of claims 1-12.
15. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the steps of the device lifetime prediction model building method of any one of claims 1-12 when the computer program is executed by the processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311111360.4A CN116822383A (en) | 2023-08-31 | 2023-08-31 | Equipment life prediction model construction method and device, readable storage medium and equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311111360.4A CN116822383A (en) | 2023-08-31 | 2023-08-31 | Equipment life prediction model construction method and device, readable storage medium and equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116822383A true CN116822383A (en) | 2023-09-29 |
Family
ID=88141402
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311111360.4A Pending CN116822383A (en) | 2023-08-31 | 2023-08-31 | Equipment life prediction model construction method and device, readable storage medium and equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116822383A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117033912A (en) * | 2023-10-07 | 2023-11-10 | 成都态坦测试科技有限公司 | Equipment fault prediction method and device, readable storage medium and electronic equipment |
| CN117250942A (en) * | 2023-11-15 | 2023-12-19 | 成都态坦测试科技有限公司 | Fault prediction method, device, equipment and storage medium for determining model |
| CN117251830A (en) * | 2023-11-14 | 2023-12-19 | 成都态坦测试科技有限公司 | Flash memory life prediction method and device, readable storage medium and electronic equipment |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130007343A1 (en) * | 2011-06-28 | 2013-01-03 | Seagate Technology Llc | Parameter Tracking for Memory Devices |
| CN107967928A (en) * | 2017-10-18 | 2018-04-27 | 武汉忆数存储技术有限公司 | A kind of flash chip life-span prediction method based on mathematical model |
| CN108536938A (en) * | 2018-03-29 | 2018-09-14 | 上海交通大学 | A kind of machine tool life prediction system and prediction technique |
| CN109815534A (en) * | 2018-12-17 | 2019-05-28 | 武汉忆数存储技术有限公司 | A kind of service life of flash memory prediction technique and device based on decision Tree algorithms |
| CN109817267A (en) * | 2018-12-17 | 2019-05-28 | 武汉忆数存储技术有限公司 | A kind of service life of flash memory prediction technique based on deep learning, system and computer-readable access medium |
| CN113934982A (en) * | 2021-10-18 | 2022-01-14 | 河北工业大学 | Method for predicting mechanical life of circuit breaker operating mechanism based on vibration-electric signal fusion |
| CN114548152A (en) * | 2022-01-17 | 2022-05-27 | 上海交通大学 | Method for predicting residual life of marine sliding bearing based on transfer learning |
| CN114757094A (en) * | 2022-03-28 | 2022-07-15 | 广东昭信照明科技有限公司 | Intelligent streetlamp service life prediction method, system, device and medium |
| CN116070527A (en) * | 2023-03-07 | 2023-05-05 | 南京航空航天大学 | Milling cutter residual life prediction method based on degradation model |
| CN116484653A (en) * | 2023-06-20 | 2023-07-25 | 广东电网有限责任公司佛山供电局 | Capacitor residual life prediction method and device |
| CN116631487A (en) * | 2023-05-26 | 2023-08-22 | 北京市合芯数字科技有限公司 | Solid state disk life prediction method, device, terminal and medium based on long-short-term memory network model |
-
2023
- 2023-08-31 CN CN202311111360.4A patent/CN116822383A/en active Pending
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130007343A1 (en) * | 2011-06-28 | 2013-01-03 | Seagate Technology Llc | Parameter Tracking for Memory Devices |
| CN107967928A (en) * | 2017-10-18 | 2018-04-27 | 武汉忆数存储技术有限公司 | A kind of flash chip life-span prediction method based on mathematical model |
| CN108536938A (en) * | 2018-03-29 | 2018-09-14 | 上海交通大学 | A kind of machine tool life prediction system and prediction technique |
| CN109815534A (en) * | 2018-12-17 | 2019-05-28 | 武汉忆数存储技术有限公司 | A kind of service life of flash memory prediction technique and device based on decision Tree algorithms |
| CN109817267A (en) * | 2018-12-17 | 2019-05-28 | 武汉忆数存储技术有限公司 | A kind of service life of flash memory prediction technique based on deep learning, system and computer-readable access medium |
| CN113934982A (en) * | 2021-10-18 | 2022-01-14 | 河北工业大学 | Method for predicting mechanical life of circuit breaker operating mechanism based on vibration-electric signal fusion |
| CN114548152A (en) * | 2022-01-17 | 2022-05-27 | 上海交通大学 | Method for predicting residual life of marine sliding bearing based on transfer learning |
| CN114757094A (en) * | 2022-03-28 | 2022-07-15 | 广东昭信照明科技有限公司 | Intelligent streetlamp service life prediction method, system, device and medium |
| CN116070527A (en) * | 2023-03-07 | 2023-05-05 | 南京航空航天大学 | Milling cutter residual life prediction method based on degradation model |
| CN116631487A (en) * | 2023-05-26 | 2023-08-22 | 北京市合芯数字科技有限公司 | Solid state disk life prediction method, device, terminal and medium based on long-short-term memory network model |
| CN116484653A (en) * | 2023-06-20 | 2023-07-25 | 广东电网有限责任公司佛山供电局 | Capacitor residual life prediction method and device |
Non-Patent Citations (2)
| Title |
|---|
| 孙惠斌 等: "《高品质切削过程的智能感知与预测技术》", pages: 5 - 6 * |
| 郭忠义 等: "装备系统剩余使用寿命预测技术研究进展", 南京航空航天大学学报, vol. 54, no. 03, pages 341 - 364 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117033912A (en) * | 2023-10-07 | 2023-11-10 | 成都态坦测试科技有限公司 | Equipment fault prediction method and device, readable storage medium and electronic equipment |
| CN117033912B (en) * | 2023-10-07 | 2024-02-13 | 成都态坦测试科技有限公司 | Equipment fault prediction method and device, readable storage medium and electronic equipment |
| CN117251830A (en) * | 2023-11-14 | 2023-12-19 | 成都态坦测试科技有限公司 | Flash memory life prediction method and device, readable storage medium and electronic equipment |
| CN117250942A (en) * | 2023-11-15 | 2023-12-19 | 成都态坦测试科技有限公司 | Fault prediction method, device, equipment and storage medium for determining model |
| CN117250942B (en) * | 2023-11-15 | 2024-02-27 | 成都态坦测试科技有限公司 | Fault prediction method, device, equipment and storage medium for determining model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN116822383A (en) | Equipment life prediction model construction method and device, readable storage medium and equipment | |
| CN107967928B (en) | Flash memory chip life prediction method based on mathematical model | |
| US20180082215A1 (en) | Information processing apparatus and information processing method | |
| CN109817267B (en) | Deep learning-based flash memory life prediction method and system and computer-readable access medium | |
| US11841839B1 (en) | Preprocessing and imputing method for structural data | |
| WO2023226423A1 (en) | Auxiliary chip design method and apparatus, device and nonvolatile storage medium | |
| CN111125658B (en) | Method, apparatus, server and storage medium for identifying fraudulent user | |
| CN115394358A (en) | Single cell sequencing gene expression data interpolation method and system based on deep learning | |
| WO2017039684A1 (en) | Classifier | |
| CN110716761A (en) | Automatic and self-optimizing determination of execution parameters of software applications on an information processing platform | |
| CN111863135B (en) | False positive structure variation filtering method, storage medium and computing device | |
| CN113822336A (en) | Cloud hard disk fault prediction method, device and system and readable storage medium | |
| CN112287036A (en) | Outlier detection method based on spectral clustering | |
| CN109815534B (en) | Flash memory life prediction method and system based on decision tree algorithm | |
| Moran et al. | Optimal Tag Sets for Automatic Image Annotation. | |
| US11854631B2 (en) | System and method for dynamic compensation for multiple interference sources in non-volatile memory storage devices | |
| CN114021716A (en) | Model training method and system and electronic equipment | |
| CN114139482A (en) | EDA circuit failure analysis method based on depth measurement learning | |
| CN112817525A (en) | Method and device for predicting reliability grade of flash memory chip and storage medium | |
| CN110970080B (en) | Method for training artificial intelligence to estimate sensing voltage of storage device | |
| US20210166362A1 (en) | Wafer map identification method and computer-readable recording medium | |
| CN112464988A (en) | Optimization method and device for K-NN algorithm near point search | |
| Tomassi et al. | Sufficient dimension reduction for censored predictors | |
| US20230307037A1 (en) | Clustering for read thresholds history table compression in nand storage systems | |
| Bari et al. | 23-bit metaknowledge template towards big data knowledge discovery and management |
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 |