Method, apparatus and device for predicting fault of optical module

Patent metadata

Field	Value
Publication number	US-10944473-B2
Application number	US-202016750611-A
Country	US
Kind code	B2
Filing date	Jan 23, 2020
Priority date	Jan 30, 2019
Publication date	Mar 9, 2021
Grant date	Mar 9, 2021

Abstract

A method and an apparatus for predicting a fault of an optical circuit includes determining a classification threshold of an operating parameter based on a classification sample set corresponding to the operating parameter of optical circuit and predicting, based on comparison results between the classification threshold and a plurality of measured values in a sequence, whether a fault occurs in the future on the optical circuit corresponding to the sequence.

First claim

What is claimed is:

1. A method for predicting a fault of an optical module, comprising:
obtaining a first sequence of the optical module in a preset time period, wherein the first sequence comprises a plurality of measured values of an operating parameter of the optical module;
obtaining a classification threshold corresponding to the operating parameter, wherein the classification threshold is based on a classification sample set corresponding to the operating parameter; and
determining a first prediction result of the optical module based on comparison results between the classification threshold and the plurality of measured values in the first sequence,
wherein the first prediction result indicates that the optical module has entered a faulty mode when the operating parameter is a bias current, the classification threshold is a bias current classification threshold, and at least one of the plurality of measured values of the operating parameter is greater than or equal to the bias current classification threshold, or
wherein the first prediction result indicates that the optical module has entered the faulty mode when the operating parameter is a receive power, the classification threshold is a receive power classification threshold, and the at least one of the plurality of measured values of the operating parameters is less than or equal to the receive power classification threshold.

2. The method of claim 1 , wherein when the first prediction result indicates that the optical module enters the faulty mode, the method further comprises:
generating a feature set based on the first sequence; and
inputting the feature set into a fault prediction model to obtain a second prediction result, wherein the second prediction result indicates an urgency level at which a fault is expected to occur on the optical module.

3. The method of claim 2 , further comprising:
obtaining a sample sequence corresponding to the operating parameter;
determining a tag of the sample sequence based on the classification threshold and the sample sequence, wherein either the tag indicates whether an optical module corresponding to the sample sequence is a non-faulty optical module, or the tag indicates an urgency level at which a fault occurs on the optical module corresponding to the sample sequence;
generating a fault prediction sample corresponding to the sample sequence, wherein the fault prediction sample comprises a feature set of the sample sequence and the tag of the sample sequence; and
generating the fault prediction model based on the fault prediction sample.

4. The method of claim 3 , further comprising updating the fault prediction model based on the feature set of the first sequence and the second prediction result.

5. The method of claim 1 , wherein before obtaining the classification threshold corresponding to the operating parameter, the method further comprises:
obtaining the classification sample set corresponding to the operating parameter, wherein the classification sample set comprises a plurality of classification samples, and wherein each classification sample comprises one measured value of the operating parameter and one first classification identifier; and
determining, based on the classification sample set, the classification threshold corresponding to the operating parameter.

6. The method of claim 5 , wherein determining, based on the classification sample set, the classification threshold corresponding to the operating parameter comprises:
determining a plurality of loss values based on the classification sample set and a plurality of reference classification thresholds, wherein each loss value corresponds to one reference classification threshold, and wherein either the plurality of reference classification thresholds are preset, or the plurality of reference classification thresholds are determined based on the measured value of each classification sample in the classification sample set;
determining a reference classification threshold corresponding to a smallest loss value in the plurality of loss values as a classification threshold corresponding to the classification sample set; and
determining, based on the classification threshold corresponding to the classification sample set, the classification threshold corresponding to the operating parameter.

7. The method of claim 6 , wherein determining the plurality of loss values based on the classification sample set and the plurality of reference classification thresholds comprises:
reclassifying the classification samples in the classification sample set based on each of the plurality of reference classification thresholds;
determining a second classification identifier of each classification sample in the classification sample set based on a classification result; and
determining, based on a loss function and the second classification identifier and the first classification identifier of each classification sample, a loss value corresponding to the reference classification threshold, wherein the loss function comprises:

loss
=

1
N

⁢

∑

q
=
1

N

⁢

(

O
q

-

P
q

)

2

,

wherein loss represents a mean square error of all the first classification identifiers in the classification sample set and the corresponding second classification identifiers, wherein N is a quantity of classification samples in the classification sample set, wherein N is greater than or equal to 2, wherein O q is a first classification identifier of a classification sample whose index value is q in the N classification samples, and wherein P q is a second classification identifier of the classification sample whose index value is q.

8. The method of claim 6 , wherein the classification sample set comprises R positive samples and S negative samples, and wherein determining the plurality of loss values based on the classification sample set and the plurality of reference classification thresholds comprises:
reclassifying the classification samples in the classification sample set based on each of the plurality of reference classification thresholds;
determining a second classification identifier of each classification sample in the classification sample set based on a classification result; and
determining, based on a loss function and the second classification identifier and the first classification identifier of each classification sample, a loss value corresponding to the reference classification threshold, wherein the loss function comprises:

loss
=

c
×

Assignees

• Huawei Tech Co Ltd

Inventors

• Yu Fangfang
• Dong Feng
• Song Wei
• Yang Qingping

Classifications

• G06F18/2415 Primary

  based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate · CPC title

• H04L43/16 Primary

  Threshold monitoring · CPC title

• H04L41/16

  using machine learning or artificial intelligence · CPC title

• H04B10/50

  Transmitters · CPC title

• H04B10/0791 Primary

  Fault location on the transmission path · CPC title