Method for detecting repair-necessary motherboards and device using the method

1. A method of implementing repairable board test of a repairable board detection device, the method comprising:

obtaining, by the repairable board detection device, repair-relevant information of a plurality of sample repairable boards;

extracting, by the repairable board detection device, predetermined features from the repair-relevant information of the plurality of the sample repairable boards;

obtaining, by, the repairable board detection device, a features list which comprises the predetermined features;

encoding a value of each feature of the features list and converting feature values of the predetermined features to truth-values based on a predetermined converting rule to establish a features truth-value list, by the repairable board detection device;

applying, by the repairable board detection device, the features truth-value list and detection results of the plurality of the sample repairable boards as training features;

establishing and training, by the repairable board detection device, a board detection model based on the training features; and

receiving, by the repairable board detection device, repair-relevant information of a repairable board and transmitting, by the repairable board detection device, the repair-relevant information of the repairable board to the board detection model to obtain a detection result of the repairable board;

wherein each feature value is trained to obtain a weighted value, and the board detection model calculates the detection result of the repairable board according to the weighted value of each feature value; and

wherein the predetermined features are selected from the group consisting of agent identification (ID), category, severity, timestamp, message, message ID, fully qualified device descriptor (FQDD), argument (ARG), and raw event data.

2. The method of claim 1 , wherein obtaining the repair-relevant information of the plurality of sample repairable boards comprises:

obtaining the repair-relevant information of the plurality of the sample repairable boards from a baseboard management controller (BMC) of each sample repairable board.

3. The method of claim 1 , wherein establishing and training the board detection model based on the training features comprises:

establishing a machine learning model and training the machine learning model based on the training features to obtain the board detection model.

4. The method of claim 3 , wherein the predetermined converting rule comprises:

converting a feature value of the features list to a truth-value of “1” in response to the feature value of the features list existing in the predetermined features; and

converting the feature value of the features list to a truth-value of “0” in response to the feature value of the features list not existing in the predetermined features.

5. The method of claim 1 , wherein establishing and training the board detection model based on the training features comprises:

dividing the training features into a training set and a verifying set;

establishing a machine learning model and training the machine learning model according to the training set, wherein the machine learning model comprises an input layer, a plurality of hidden layers, and an output layer;

applying the verifying set to verify a trained machine learning model and counting a prediction accuracy of the trained machine learning model based on each verification result;

determining whether the prediction accuracy of the trained machine learning model is less than a predetermined value; and

setting the trained machine learning model as the board detection model in response to the prediction accuracy of the trained machine learning model being not less than the predetermined value.

6. The method of claim 5 , further comprising:

adjusting parameters of the machine learning model in response to the prediction accuracy being less than the predetermined value;

retraining an adjusted machine learning model according to the training set;

applying the verifying set to verify the adjusted machine learning model and regaining a prediction accuracy of the adjusted machine learning model based on each verification result;

determining whether the prediction accuracy of the adjusted machine learning model is less than the predetermined value;

setting the adjusted machine learning model as the board detection model in response to the prediction accuracy of the adjusted machine learning model being not less than the predetermined value; and

repeating adjustment and retraining steps until the prediction accuracy is not less than the predetermined value in response to the prediction accuracy of the adjusted machine learning model being less than the predetermined value.

7. The method of claim 1 , wherein receiving the repair-relevant information of the repairable board and transmitting the repair-relevant information of the repairable board to the board detection model to obtain the detection result of the repairable board comprises:

receiving the repair-relevant information of the repairable board and transmitting the repair-relevant information of the repairable board to the board detection model to obtain a probability value being a normal board or a probability value being an abnormal board;

determining whether the probability value is greater than a predetermined probability; and

obtaining the detection result of the repairable board based on a determined result.

8. A repairable board detection device comprising:

at least one processor;

a storage; and

one or more programs that are stored in the storage and executed by the at least one processor, the one or more programs comprising instructions for:

obtaining repair-relevant information of a plurality of sample repairable boards;

extracting predetermined features from the repair-relevant information of the plurality of the sample repairable boards;

analyzing the predetermined features to obtain one or more training features;

establishing and training a board detection model based on the training features; and

receiving repair-relevant information of a repairable board and transmitting the repair-relevant information of the repairable board to the board detection model to obtain a detection result of the repairable board;

wherein the predetermined features are selected from the group consisting of agent ID, category, severity, timestamp, message, message ID, FQDD, ARG, and raw event data.

9. The repairable board detection device of claim 8 , wherein the instruction of obtaining the repair-relevant information of the plurality of sample repairable boards comprises:

obtaining the repair-relevant information of the plurality of the sample repairable boards from a BMC of each sample repairable board.

10. The repairable board detection device of claim 8 , wherein the instruction of analyzing the predetermined features to obtain the one or more training features comprises:

obtaining a features list which comprises the predetermined features;

encoding a value of each feature of the features list and converting feature values of the predetermined features to truth-values based on a predetermined converting rule to establish a features truth-value list; and

applying the features truth-value list and detection results of the plurality of the sample repairable boards as the training features.

11. The repairable board detection device of claim 10 , wherein the instruction of establishing and training the board detection model based on the training features comprises:

establishing a machine learning model and training the machine learning model based on the training features to obtain the board detection model.

12. The repairable board detection device of claim 11 , wherein each feature value is trained to obtain a weighted value, and the board detection model calculates the detection result of the repairable board according to the weighted value of each feature value.

13. The repairable board detection device of claim 11 , wherein the predetermined converting rule comprises:

converting a feature value of the features list to a truth-value of “1” in response to the feature value of the features list existing in the predetermined features; and

converting the feature value of the features list to a truth-value of “0” in response to the feature value of the features list not existing in the predetermined features.

14. The repairable board detection device of claim 8 , wherein the instruction of establishing and training the board detection model based on the training features comprises:

dividing the training features into a training set and a verifying set;

establishing a machine learning model and training the machine learning model according to the training set, wherein the machine learning model comprises an input layer, a plurality of hidden layers, and an output layer;

applying the verifying set to verify a trained machine learning model and counting a prediction accuracy of the trained machine learning model based on each verification result;

determining whether the prediction accuracy of the trained machine learning model is less than a predetermined value; and

setting the trained machine learning model as the board detection model in response to the prediction accuracy of the trained machine learning model being not less than the predetermined value.

15. The repairable board detection device of claim 14 , wherein the one or more programs further comprise:

adjusting parameters of the machine learning model in response to the prediction accuracy being less than the predetermined value;

retraining an adjusted machine learning model according to the training set;

applying the verifying set to verify the adjusted machine learning model and regaining a prediction accuracy of the adjusted machine learning model based on each verification result;

determining whether the prediction accuracy of the adjusted machine learning model is less than the predetermined value;

setting the adjusted machine learning model as the board detection model in response to the prediction accuracy of the adjusted machine learning model being not less than the predetermined value; and

repeating adjustment and retraining steps until the prediction accuracy is not less than the predetermined value in response to the prediction accuracy of the adjusted machine learning model being less than the predetermined value.

16. The repairable board detection device of claim 8 , wherein the instruction of receiving the repair-relevant information of the repairable board and transmitting the repair-relevant information of the repairable board to the board detection model to obtain the detection result of the repairable board comprises:

receiving the repair-relevant information of the repairable board and transmitting the repair-relevant information of the repairable board to the board detection model to obtain a probability value being a normal board or a probability value being an abnormal board;

determining whether the probability value is greater than a predetermined probability; and

obtaining the detection result of the repairable board based on a determined result.