CN115526503A

CN115526503A - Equipment inspection data processing method, device, equipment and readable storage medium

Info

Publication number: CN115526503A
Application number: CN202211210623.2A
Authority: CN
Inventors: 陈慧; 李博威; 吴广展; 罗涵; 王辉; 魏晖; 莫均波; 黄志�; 侯海敏; 龙正航; 罗绵辉; 林幼彬; 徐利
Original assignee: China United Network Communications Group Co Ltd
Current assignee: China United Network Communications Group Co Ltd
Priority date: 2022-09-30
Filing date: 2022-09-30
Publication date: 2022-12-27

Abstract

The application provides a method, a device and equipment for processing equipment inspection data and a readable storage medium. The method comprises the following steps: acquiring a data set of N characteristic parameters of target equipment, which is obtained by inspecting the target equipment in a target time period; preprocessing data in the data set of each characteristic parameter to obtain a data set preprocessed by each characteristic parameter; acquiring a health index observation sequence of the target equipment in a target time period by adopting the data set preprocessed by each characteristic parameter; and predicting the health state of the target equipment according to the health index observation sequence. According to the method, the health state of the equipment can be accurately predicted according to the inspection data of the equipment, so that data support is provided for inspection and operation maintenance of subsequent equipment.

Description

Equipment inspection data processing method, device, equipment and readable storage medium

Technical Field

The present application relates to the field of data processing, and in particular, to a method, an apparatus, a device, and a readable storage medium for processing device inspection data.

Background

In recent years, with the rapid development of new technologies such as cloud computing and big data, data center rooms can be built and operated on a large scale, the number of equipment in the rooms is increased, the operation and maintenance traffic volume matched with the rooms is increased greatly, and daily inspection of the equipment in the data center rooms is needed to ensure the normal operation of the data center rooms.

At present, the routing inspection of a data center machine room lacks the processing and analysis of routing inspection data of equipment obtained by routing inspection, and the health state of the equipment cannot be predicted. Therefore, how to predict the health state of the equipment based on the polling data of the equipment is an urgent problem to be solved.

Disclosure of Invention

The application provides a method, a device, equipment and a readable storage medium for processing equipment inspection data, which are used for solving the problem that the health state of the equipment cannot be predicted based on the equipment inspection data.

In a first aspect, the present application provides a device inspection data processing method, including:

acquiring N characteristic parameter data sets of target equipment obtained by polling the target equipment in a target time period, wherein the characteristic parameter data sets comprise M data of the characteristic parameters obtained by polling M times in the target time period; n is an integer greater than or equal to 1, and M is an integer greater than or equal to 2;

preprocessing data in the data set of each characteristic parameter to obtain a data set preprocessed by each characteristic parameter;

acquiring a health index observation sequence of the target equipment in the target time period by adopting the data set preprocessed by each characteristic parameter, wherein the health index observation sequence comprises M elements, each element corresponds to one inspection and is used for representing the health index of the target equipment acquired based on the data of the N characteristic parameters acquired by the inspection;

and predicting the health state of the target equipment according to the health index observation sequence.

Optionally, the obtaining a health index observation sequence of the target device in the target time period by using the data set preprocessed by each of the characteristic parameters includes:

acquiring the weight of each characteristic parameter;

adding products of the data of each characteristic parameter and the weight obtained by the same inspection in the data set after the pretreatment of each characteristic parameter to obtain a health index of the target equipment corresponding to the inspection;

and generating a health index observation sequence of the target equipment in the target time period by using the health index of the target equipment corresponding to each inspection.

Optionally, the obtaining the weight of each feature parameter includes:

generating a data matrix comprising data of all the characteristic parameters according to the data set of all the characteristic parameters;

and processing the data matrix by using a principal component analysis method to obtain the weight of each characteristic parameter.

Optionally, the predicting the health status of the target device according to the health index observation sequence includes:

predicting the health state of the target equipment by using the health index observation sequence and a hierarchical model for predicting the health state by adopting a Viterbi algorithm; the hierarchical model is represented by an initial health state vector, a health state transition probability matrix and an observation probability matrix.

Optionally, the hierarchical model is a hidden markov HMM based on an expectation-maximization algorithm, which is constructed according to a preset device health state and trained by using a sample health index observation sequence.

Optionally, after predicting the health status of the target device according to the health index observation sequence, the method further comprises:

updating the tag of the health status of the target device in the device information of the target device.

and if the health state of the target equipment does not accord with the preset health state, outputting prompt information of the health state of the target equipment to trigger an inspection task of inspecting the target equipment again.

In a second aspect, the present application provides a data processing apparatus for equipment inspection, the apparatus comprising:

the system comprises a first acquisition module, a second acquisition module and a third acquisition module, wherein the first acquisition module is used for acquiring a data set of N characteristic parameters of target equipment obtained by polling the target equipment in a target time period, and the data set of the characteristic parameters comprises M data of the characteristic parameters obtained by polling the target equipment M times in the target time period; n is an integer greater than or equal to 1, and M is an integer greater than or equal to 2;

the preprocessing module is used for preprocessing data in the data set of each characteristic parameter to obtain the data set preprocessed by each characteristic parameter;

a second obtaining module, configured to obtain a health index observation sequence of the target device in the target time period based on the data set preprocessed by each of the feature parameters, where the health index observation sequence includes M elements, each element corresponds to one inspection, and is used to represent a health index of the target device obtained based on data of N feature parameters obtained by the inspection;

and the prediction module is used for predicting the health state of the target equipment according to the health index observation sequence.

In a third aspect, the present application provides an electronic device, comprising: a processor, and a memory communicatively coupled to the processor;

the memory stores computer execution instructions;

the processor executes computer-executable instructions stored by the memory to implement the method of any of the first aspects.

In a fourth aspect, the present application provides a computer-readable storage medium having stored thereon computer-executable instructions for implementing the method according to any one of the first aspect when executed by a processor.

In a fifth aspect, the present application provides a computer program product comprising a computer program that, when executed by a processor, implements the method according to any one of the first aspect.

In a sixth aspect, the present application provides a chip having a computer program stored thereon, which, when executed by the chip, implements the method according to any of the first aspects.

According to the equipment inspection data processing method, the equipment inspection data processing device, the equipment and the readable storage medium, the health index of the target equipment is generated through multi-dimensional characteristic parameter data of the equipment which is acquired through inspection in the target time period and changes along with the health state of the target equipment, and the current health state of the target equipment can be reflected. The further generated health index observation sequence of the target equipment can reflect the change trend of the health state of the target equipment in the target time period, so that the health state of the equipment can be accurately predicted based on the change trend, and data support is provided for the inspection and operation maintenance of subsequent equipment.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and together with the description, serve to explain the principles of the application.

Fig. 1 is a schematic flow chart of a device inspection data processing method according to an embodiment of the present application;

fig. 2 is a schematic flow chart of another method for processing data of equipment inspection provided in the embodiment of the present application;

fig. 3 is a schematic diagram of an architecture of a device inspection system according to an embodiment of the present disclosure;

fig. 4 is a schematic diagram of another device inspection system architecture provided in the embodiment of the present application;

fig. 5 is a schematic structural diagram of an apparatus inspection data processing apparatus according to an embodiment of the present application;

fig. 6 is a schematic structural diagram of an electronic device 600 provided in the present application.

Specific embodiments of the present application have been shown by way of example in the drawings and will be described in more detail below. These drawings and written description are not intended to limit the scope of the inventive concepts in any manner, but rather to illustrate the inventive concepts to those skilled in the art by reference to specific embodiments.

Detailed Description

Reference will now be made in detail to the exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, like numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present application, as detailed in the appended claims.

At present, the routing inspection of a data center machine room lacks the processing and analysis of routing inspection data of equipment obtained by routing inspection, and the health state of the equipment cannot be predicted. In view of this, the present application provides a method for processing inspection data of a device, which can predict a health state of the device based on inspection data of the device within a period of time, so as to provide data support for inspection and operation maintenance of subsequent devices. For example, the equipment with poor health state is subjected to key inspection to avoid equipment failure and improve equipment utilization rate.

The execution main body of the application can be an inspection system or an inspection data analysis device in the inspection system. The patrol data analyzing apparatus may be an electronic device having a processing capability, such as a terminal, a server, or a computing device.

The following describes the technical solution of the present application and how to solve the above technical problems with specific embodiments by taking an execution subject as an example of the inspection data analysis device. The following several specific embodiments may be combined with each other, and details of the same or similar concepts or processes may not be repeated in some embodiments. Embodiments of the present application will be described below with reference to the accompanying drawings.

Fig. 1 is a schematic flow diagram of a device inspection data processing method according to an embodiment of the present disclosure. As shown in fig. 1, the method may include the steps of:

s101, acquiring a data set of N characteristic parameters of the target equipment, which is obtained by the target equipment in the target time period. The data set of the characteristic parameters comprises M data of the characteristic parameters obtained by M times of routing inspection in the target time period; n is an integer of 1 or more, and M is an integer of 2 or more.

The characteristic parameters of the target device refer to the operating parameters of the target device that can reflect the health state of the target device, or when the health state of the target device changes, the operating parameters also change correspondingly. Such as one or more of temperature, power, noise, etc. of the device, particularly with respect to the collection of inspection data, and the actual analytical requirements of the device.

For example, the data set of N characteristic parameters of the target device may be as shown in table 1 below:

TABLE 1

Data set	Target time period
		Characteristic parameter 1 dataset	Characteristic parameter 11, characteristic parameter 12,. And characteristic parameter 1M
Characteristic parameter 2 data set	Characteristic parameter 21, characteristic parameter 22,. And characteristic parameter 2M
		...	...
Characteristic parameter N data set	Characteristic parameter 31, characteristic parameter 32,. And characteristic parameter 3M

The above-mentioned feature data set may be obtained, for example, by calling a patrol record of a target time period in a database of the patrol system, or may be input to the patrol data analysis device by a user, which is not limited in this application.

S102, preprocessing the data in the data set of each characteristic parameter to obtain the data set after preprocessing of each characteristic parameter.

The preprocessing mentioned here is to perform a certain processing on the data in the data set of each of the characteristic parameters, and may include, for example: one or more processes such as data cleaning, data transformation, data reduction and the like are performed to improve the data quality, so that the subsequent data processing efficiency is higher, and the accuracy of the obtained result is higher.

S103, acquiring a health index observation sequence of the target equipment in a target time period by adopting the data set preprocessed by the characteristic parameters.

The health index observation sequence is a set of elements used for representing health indexes obtained after target equipment is patrolled and examined in a target time period, and comprises M elements, wherein each element corresponds to one patrol and is used for representing the health indexes of the target equipment obtained based on data of N characteristic parameters obtained by the patrol and examination.

The health index observation sequence can be expressed as: o = { O ₁ ，O ₂ ，...，O _M }. Wherein, aiming at any time of M times of polling M, corresponding to the M time polling O _m The following may be used:

for example, the sum of the N data of the characteristic parameters obtained in the mth round may be regarded as O _m Alternatively, a value obtained by appropriately modifying (for example, normalizing, and/or multiplying by a predetermined coefficient, and/or dividing by a predetermined coefficient, etc.) the sum after the addition is taken as O _m Or, taking the sum of each characteristic parameter obtained by the mth inspection and the corresponding weight as O _m 。

And S104, predicting the health state of the target equipment according to the health index observation sequence.

The health state of the equipment is related to the subsequent operation maintenance mode and the inspection mode of the equipment, or the operation maintenance and the inspection modes of the equipment with different health states are different. For example, the health status of the device and the operation, maintenance and inspection modes have a mapping relationship as shown in the following table 2:

TABLE 2

Health status of equipment	Plant operation and maintenance recommendations	Inspection mode
			Is extremely good	Maintenance of normal operation	Normal inspection
Jia	Maintenance of normal operation	Normal inspection
			Good effect	Can continue to operate and strengthen supervision	Normal inspection
Is poor	Monitoring the operation and taking adjustment measures	Key inspection
			Extreme difference	Immediate shutdown for maintenance	Key inspection

Compared with normal inspection, the key inspection can increase the inspection times, the inspection items, the inspection times, the inspection items and the like.

It should be understood that the health status of the device shown in table 2 is only an example, and the present application does not specifically limit the health status. In addition, one health state may correspond to one device operation and maintenance suggestion, or one health state may correspond to a plurality of device operation and maintenance suggestions.

One possible implementation manner may be to predict the health status of the target device by using the health index observation sequence and a hierarchical model for predicting the health status by using a viterbi algorithm; the hierarchical model is represented by an initial health state vector, a health state transition probability matrix and an observation probability matrix.

In another possible implementation manner, the health index observation sequence may be input to a pre-trained prediction model to predict the health state of the target device. The prediction model can be obtained by training a health index observation sequence of the sample device in advance and a health state label of the sample device.

In another possible implementation manner, a mapping relationship between the health index observation sequence and the health status is preset, so that the health status of the target device can be obtained through the mapping relationship and the health index observation sequence of the target device.

By acquiring the health state of the target equipment based on the mode, data support can be provided for inspection and operation maintenance of subsequent equipment. For example, the equipment with poor health state is subjected to key inspection to avoid equipment failure and improve equipment utilization rate.

Optionally, after predicting the health status of the target device according to the health index observation sequence, the tag of the health status of the target device may also be updated in the device information of the target device.

The device information of the target device refers to basic information of the target device stored in the database of the inspection system, and includes, but is not limited to, tags such as a device name, a device ID, a home machine room, a device IP, a device location, and a device health status tag. After predicting the health status of the target device, the tag of the health status of the target device is updated in the device information of the target device. Therefore, when the follow-up inspection personnel acquire the equipment information of the target equipment from the database of the inspection system through the terminal equipment, the health state of the equipment can be inspected, so that the health state of the equipment can be visualized, and the follow-up inspection personnel can focus on the equipment when inspecting.

If the tag of the health state of the target device exists before the device information of the target device, the tag of the health state of the target device obtained by the prediction may be replaced with the tag of the original health state, or the tag of the health state of the target device obtained by the prediction may be stored in the device information of the target device in a time series manner after the tag of the health state of the target device obtained in the last time. As an alternative example, when the current predicted health state is inconsistent with the label of the health state of the original target device, the label of the health state of the target device is replaced; and when the predicted health state is consistent with the label of the original target equipment health state, keeping the predicted health state unchanged.

If the tag of the health status of the target device does not exist in the device information of the target device before, the tag of the health status of the target device predicted this time may be added to the device information of the target device as a new content.

Optionally, after the health state of the target device is predicted according to the health index observation sequence, if the health state of the target device is lower than the preset health state, prompt information of the health state of the target device may be output to trigger a polling task of polling the target device again.

For example, the device health status prompt information may be output to a terminal device of a person in charge of the target device, and the person in charge of the target device schedules an inspection task of performing inspection again on the target device to achieve the purpose of focused inspection.

According to the equipment inspection data processing method, the health index of the target equipment is generated based on the characteristic parameter data of the multi-dimensional equipment which is acquired by inspection in the target time period and changes along with the health state of the target equipment, and the current health state of the target equipment can be reflected. The health index observation sequence of the target equipment generated further can reflect the change trend of the health state of the target equipment in the target time period, so that the health state of the equipment can be accurately predicted based on the change trend, and data support is provided for the inspection and operation maintenance of subsequent equipment.

Fig. 2 is a schematic flow chart of another method for processing equipment inspection data according to the embodiment of the present application. As shown in fig. 2, the method may include:

s201, acquiring a data set of N characteristic parameters of the target equipment, which is obtained by the target equipment in the target time period.

Illustratively, the patrol records of the target device within the last month are retrieved from the database, and M times of patrol data of N characteristic parameters are respectively obtained from the patrol records to form a data set C of the N characteristic parameters _n Respectively including m times of polling data c of the target device in the last month _nm . Wherein, the value of M is 1,2, a., and M represents the mth inspection; the value of N is 1,2,.., N, corresponding to the N characteristic parameters, and forms a one-to-one mapping relationship with the characteristic parameters, for example, the temperature corresponds to N =1, the power corresponds to N =2, and the noise corresponds to N =3.

For example, the data set of N characteristic parameters of the target device is shown in table 3 below:

TABLE 3

Data set C _n	Target time period
		C ₁	Characteristic parameter c ₁₁ Characteristic parameter c ₁₂ A _1M
C ₂	Characteristic parameter c ₂₁ Characteristic parameter c ₂₂ A _2M
		...	...
C _N	Characteristic parameter c _N1 Characteristic parameter c _N2 A _NM

S202, carrying out normalization processing on data in the data set of each characteristic parameter to obtain the data set after preprocessing of each characteristic parameter.

For example, the data in the feature parameter data set is normalized by using the following formula (1):

wherein, the first and the second end of the pipe are connected with each other,

the data corresponding to the preprocessed data of each characteristic parameter obtained by the mth patrol and the data obtained by preprocessing the data of each characteristic parameter obtained by the M patrols in the latest month form a data set preprocessed by each characteristic parameter

c _nmmax And c _nmmin The characteristic parameters are the maximum value and the minimum value of the data in the data set. Taking n as 1 as an example, c _nmmax Is C ₁ Maximum value of data in (1), c _nmmin Is C ₁ Of (2) is determined.

And S203, acquiring the weight of each characteristic parameter.

In a possible implementation manner, the weight of each feature parameter may be, for example, a preset value. The sum of the weights of all the characteristic parameters may be 1, for example.

In another possible implementation, the weights of the N characteristic parameters are determined by principal component analysis. The method comprises the following specific steps:

(1) And generating a data matrix comprising the data of all the characteristic parameters according to the data set of all the characteristic parameters.

Namely, M times of patrol data of N characteristic parameters are combined into a data matrix with M multiplied by N dimensions.

(2) And processing the data matrix by using a principal component analysis method to obtain the weight of each characteristic parameter. The method comprises the following specific steps:

1) And normalizing the data matrix data of the dimension M multiplied by N to obtain a normalized data matrix D = (D) ₁ ，D ₂ ，...，D _N ). Wherein D is ₁ ，D ₂ ，...，D _N Namely N variable indexes of the principal component analysis method.

Further, assume a new set of variable indices Z ₁ ，Z ₂ ，...，Z _p Wherein p is 1 to N, such that it satisfies Z _t ＝σ _1t D ₁ +σ _2t D ₂ +...+σ _Nt D _N . Where t is 1 to p, coefficient matrix σ = (σ) _N1 ，σ _N2 ，...，σ _Np ) Is a matrix of dimension N × p.

Then the new variable index Z ₁ ，Z ₂ ，...，Z _p Are respectively called original variable indexes D ₁ ，D ₂ ，...，D _N The first, second, right, pth principal component of (a).

2) And calculating a covariance matrix E of the normalized data matrix D. The covariance matrix E is a matrix of dimension N × N.

3) Obtaining N eigenvalues alpha of covariance matrix E _n And a feature vector beta corresponding to each feature value _n 。

Wherein alpha is _n Can be expressed as alpha ₁ ，α ₂ ，...，α _N A feature vector β corresponding to it one by one _n Can be expressed as beta ₁ ，β ₂ ，...，β _N . To be noted, the characteristic value α ₁ ﹥α ₂ ﹥...﹥α _N Is more than 0, and the corresponding feature vectors are beta in sequence ₁ ＝[β ₁₁ ，β ₂₁ ，...，β _N1 ] ^T ,β ₂ ＝[β ₁₂ ，β ₂₂ ，...，β _N2 ] ^T ，...，β _N ＝[β _1N ，β _2N ，...，β _NN ] ^T 。

4) And determining the value of the number p of the main components.

A contribution rate threshold is preset, for example, 0.8, and the value of p is obtained by using the following formula (2).

5) And acquiring the former p main components according to the value of p.

After the first p principal components are determined by the formula (2), the first p principal components sequentially correspond to the first p eigenvalues, and the eigenvector corresponding to each eigenvalue. For example, the first principal component corresponds to the eigenvalue α ₁ And a feature vector beta ₁ 。

Then, the tth principal component can be expressed as: z is a linear or branched member _t ＝β _1t D ₁ +β _2t D ₂ +...+β _Nt D _N 。

6) Determining the corresponding variable index (i.e. D) in each principal component ₁ ，D ₂ ，...，D _N ) The coefficient matrix σ of (a) is, for example, as shown in the following formula (3):

7) Determining the variance contribution rate of each characteristic value by using the following formula (4)

8) Determining a comprehensive score model coefficient gamma by using the following formula (5) _n 。

Based on equation (3), each principal component in the first p principal components can be represented by Z _t ＝σ _1t D ₁ +σ _2t D ₂ +...+σ _Nt D _N And (4) showing. For example, the first principal component Z ₁ ＝σ ₁₁ D ₁ +σ ₂₁ D ₂ +...+σ _N1 D _N 。

The comprehensive score model Y = gamma can be obtained by synthesizing the first p principal components ₁ D ₁ +γ ₂ D ₂ +...+γ _N D _N . The comprehensive score model is a health state model of the equipment, namely the health state of the equipment is described by using N variable indexes. Gamma ray _n To synthesize the score model coefficients, the following equation (5) can be used in relation to the coefficient σ matrix.

9) And calculating the weight normalization of the indexes by adopting the following formula (6) according to the comprehensive score model coefficient to obtain the weight corresponding to each index:

where ρ is _n For each of the above characteristic parametersThe weight of (c).

And S204, adding products of the data of the characteristic parameters and the weight obtained by the same inspection in the data set after the pretreatment of the characteristic parameters to obtain the health index of the target equipment corresponding to each inspection. That is, the following formula (7) is adopted:

s205, generating a health index observation sequence O = { O ] of the target device in the last month by using the health index of the target device corresponding to each inspection ₁ ，O ₂ ，...，O _M }。

And S206, predicting the health state of the target equipment by using the health index observation sequence and a hierarchical model for predicting the health state by adopting a Viterbi algorithm.

The hierarchical model is represented by an initial health state vector, a health state transition probability matrix and an observation probability matrix.

For example, the hierarchical Model is a Model trained by using a sample health index observation sequence according to a preset device health state, for example, a Hidden Markov Model (HMM) based on an expectation-maximization algorithm.

Next, a procedure of predicting the health state of the target device using the viterbi algorithm will be described by taking an HMM of the expectation-maximization algorithm as an example.

It should be understood that when the health index observation sequence is used to predict the health state of the target device, the HMM based on the expectation-maximization algorithm described above should be trained in advance based on the sample health index observation sequence. The following steps of constructing an HMM based on an expectation-maximization algorithm trained by using an observation sequence of sample health indexes are described, including the following steps:

preset, the device is divided into L health states, and the L health states form a one-to-one mapping relationship with the hidden state set R = {1,2.Taking at least one sample device to inspect corresponding health index observation sequence F = { F) for H times in sample duration ₁ ，F ₂ ，...，F _H As a sample health index observation sequence corresponding to hidden states I = { I = ₁ ，I ₂ ，...，I _H And taking values from the hidden state set R. Wherein H represents the number of polling times.

(1) Definition HMM

And (3) representing the HMM as lambda = (A, B, pi), and obtaining an initial state probability distribution pi of the equipment, an equipment state transition probability matrix A and an equipment observation state probability matrix B based on the sample health index observation sequence of the equipment.

The above-described device initial state probability distribution pi can be expressed as pi = { pi = { [ pi ] _i Represents the first inspection in the first two months, i.e. when h =1, F in the sample health index observation sequence ₁ Corresponding hidden state I ₁ Probability of = i, where H represents a certain of H rounds of inspection, and i takes a value from R.

The device state transition probability matrix a may be represented as a = [ a = _ij ] _L×L Indicates that the inspection is performed in a hidden state I when the number of times of inspection is h _h Value is I, and hidden state I is generated when h +1 times of routing inspection is carried out _h+1 The probability value of j is taken as i and j are taken from R.

The above-described device observation state probability matrix B may be represented as B = [ B ] _j (k)] _L×H Indicates that if the number of rounds is h, the hidden state is I _h = j, the corresponding observation sequence is F _h ＝ν _k Then b _j (k) Shows the observation sequence v at the h-th inspection _k In the hidden state is I _h Probability of generation under = j.

(2) Training to obtain HMM parameters

Initializing lambda = (A, B, pi) to obtain initialized HMM parameter lambda ₀ ＝(A ₀ ，B ₀ ，π ₀ ). The initialization may be performed by randomly substituting a sample health index observation sequence, or by substituting each parameter value with a respective uniform value, such as pi _i =1/L, i.e.The probability of each health state occurring when the first inspection is carried out in the target time period is equal. The initialization should satisfy the following conditions:

π _i ≥0，a _ij ≥0，Σ _i π _i ＝1,Σ _j a _ij ＝1。

and (3) learning and training the initialized HMM by using the sample health index observation sequence to obtain an HMM parameter lambda = (A, B, pi).

How to obtain the HMM parameters A, B, pi is described below, taking as an example the Baum-Welch algorithm based on the expectation-maximization (EM) algorithm. The specific process is as follows:

1) Determining the conditional probability P (F | Lambda) of the observation sequence F appearing under the HMM parameter Lambda.

In a known observation sequence F = { F = { (F }) ₁ ，F ₂ ，...，F _H H and initial HMM parameter λ ₀ ＝(A ₀ ，B ₀ ，π ₀ ) Based on a forward-backward algorithm, a conditional probability P (F | λ) of the occurrence of the observation sequence F under the HMM parameter λ is calculated. The method comprises the following specific steps:

1a) According to the forward algorithm, the hidden state is I when the polling times is defined as h _h = i, sequence of observed states is F ₁ ，F ₂ ，...，F _h The probability of (d) is a forward probability. The forward probability can be expressed as: alpha is alpha _h (i)＝P(F ₁ ，F ₂ ，...，F _h ，I _h = i | λ), obtained recursively from the value of h, exemplarily:

when h =1, α ₁ (i)＝π _i b _i (F ₁ )；

When H =2,3,.. H,

to finally obtain

1b) According to a backward algorithm, defining the hidden state as I when the polling frequency is h _h = i, and the sequence of observation states from the time when the number of rounds is H to the last round H is F _h+1 ，F _h+2 ，...，F _H The probability of (d) is a backward probability, which can be expressed as:

β _h (i)＝P(F _h+1 ，F _h+2 ，...，F _H ，|I _h ＝i，λ)，

and (4) recurrently obtaining the value of h according to an example:

initializing each hidden state backward probability when the last polling frequency is H: beta is a _H (i)＝1；

Recursion patrol frequency H = H-1,H-2, a backward probability of 1:

to obtain finally

2) Determining two sets of probability variables epsilon _h (i, j) and γ _h (i)。

Above epsilon _h (I, j) represents that the system is in a hidden state I when the polling frequency is h _h Is in a hidden state I when the number of patrolling times is h +1 _h+1 The probability of = j, which can be calculated using the following equation (8).

The above gamma-ray _h (i) Indicates that the system is in the state I when the polling times are h _h The probability of = i can be calculated using the following equation (9).

3) And determining the expected value of the state.

Will be based on epsilon of each round inspection _h (i，j) And gamma _h (i) Summing, one can get:

expected value of occurrence of state i under observation sequence F:

expectation of transition from state i under observation sequence F:

expected value for transition from state i to state j under observation sequence F:

4) And obtaining a model parameter reestimation formula according to the calculation result, wherein the model parameter reestimation formula is described in the following formula (10) to formula (15):

π _i ′＝γ ₁ (i) Formula (10)

π′＝{π _i ' } formula (11)

A′＝[a _ij ′] _L×L Formula (13)

And is

B′＝[b _j ′(k)] _L×H Formula (15)

5) Updating HMM parameters lambda '= (A', B ', pi') according to the reestimation formula, and performing updating iteration until P (F | lambda) is not obviously increased any more, and obtaining an optimal model lambda = (A, B, pi) to terminate iteration; otherwise, repeating the training steps.

Through the steps, HMM parameters lambda = (A, B, pi) based on the sample health index observation sequence can be obtained through training.

The target health index observation sequence O of the target equipment is brought into the trained HMM parameter lambda = (A, B, pi), and the health state of the target equipment is predicted by utilizing a Viterbi algorithm, and the method comprises the following steps:

(1) Inputting: HMM parameters λ = (a, B, pi), target health index observation sequence O = { O ₁ ，O ₂ ，...，O _M }。

(2) When the number of times of inspection is m, and the observation sequence is O _m Hidden state is I _m When = i, the maximum value of the joint probability in all the state transition paths is δ _m (i) At m =1, δ ₁ (i)＝π _i b _i (O ₁ ) I is taken from R; recursion yields δ when M =2,3, M _m (i)＝max _1≤j≤M [δ _m-1 (j)a _ji ]b _i (O _m ) (ii) a Calculating delta when the polling times is M _M (i) Therefore, the maximum probability value in all the state transition paths at the moment after the Mth patrol is further calculated. The state transition path corresponding to the maximum probability value obtained at this time is the hidden state when the mth inspection is performed, which is the predicted health state of the target device.

Based on the steps, the health state of the target equipment can be predicted by utilizing a health index observation sequence obtained by the characteristic parameters of the target equipment, and data support is provided for inspection and operation maintenance of subsequent equipment.

Exemplarily, if the health state of the target device is predicted to be poor, prompt information of the health state of the target device is output to trigger important inspection of the target device, so as to avoid equipment failure and improve equipment utilization rate.

According to the equipment inspection data processing method provided by the embodiment of the application, based on the characteristic parameter data of the multi-dimensional equipment acquired by inspection in the target time period, the weight of each characteristic parameter is acquired by means of principal component analysis and the like, multi-dimensional dimensionality reduction of the characteristic parameters is realized, the health index of the multi-dimensional fused target equipment is further acquired, the accuracy of reflecting the health state of the target equipment by the health index is improved, and the health index observation sequence of the target equipment with higher accuracy is further acquired. And dynamically planning possible health state transition paths of the target equipment based on the transition probability among the historical health states by adopting an HMM (hidden Markov model) and a Viterbi algorithm based on an expectation-maximization algorithm trained by a sample health index observation sequence generated by the characteristic parameter data of the historical equipment, so that the health state of the target equipment is predicted, and the prediction accuracy is improved.

The above describes how to predict the health status of the equipment by using the inspection data of the equipment, and there is no limitation on how to inspect the equipment to obtain the inspection data.

Illustratively, the following equipment inspection method can be used to inspect equipment, for example.

Fig. 3 is a schematic diagram of an architecture of a device inspection system according to an embodiment of the present application. As shown in fig. 3, the inspection system is functionally divided, and may include, for example, an information entry device, a two-dimensional code generation device, a mobile intelligent terminal, an applet terminal, a data storage device, a message push device, and an inspection data analysis device.

The information input device is used for inputting basic information of the equipment.

The two-dimensional code generating device is used for generating the two-dimensional code which has the unique identifier and has a one-to-one mapping relation with the equipment ID.

The mobile intelligent terminal is used for checking equipment information and filling and submitting inspection data when an inspector inspects equipment on site. For example, the mobile phone may be a mobile phone, a tablet computer, a handheld device with scanning and communication functions, a wearable device (a watch, a sports bracelet, a sports foot ring, etc.), etc., and the application is not limited herein.

The small program end refers to a routing inspection small program and is used for achieving functions of checking equipment information, filling and submitting routing inspection data and the like.

The data storage device is used for storing data related to the inspection of the equipment, and can be equipment basic information, equipment inspection records and the like.

The message pushing device is used for pushing messages to terminal equipment of related personnel. Wherein the person of interest, e.g. the person in charge of the device; the message may be, for example, a message with a poor health status of the device, or a message with an abnormal inspection result; the message pushing mode may be through a wechat public number, a small program, or a short message, and the application is not limited herein.

The inspection data analysis device is used for predicting the health state of equipment by the method provided by the embodiment.

The division of the devices of the equipment inspection system shown in fig. 3 is merely an example, and the present application does not limit the division of the devices and the names of the devices. The above-mentioned devices may be independent entities, or some devices may be integrated in the same entity.

The patrol system is illustrated by a specific hardware architecture.

Fig. 4 is a schematic diagram of another device inspection system architecture provided in the embodiment of the present application. As shown in fig. 4, the inspection system includes a database, a handheld end, a server end, a data acquisition server end, and a Personal Computer (PC) end.

The devices shown in the figure may be, for example, inspection objects such as switches, servers, etc.

The database is a data storage device. The handheld terminal can be, for example, the mobile intelligent terminal.

The server is in communication connection with the equipment, the database, the handheld end, the data acquisition server and the PC end and is used for realizing data interaction among all parties.

The data acquisition server side can be integrated with a message pushing device, a routing inspection data analysis device and the like.

The PC terminal may be integrated with an information input device and a two-dimensional code generation device, for example. The PC terminal can be used for realizing operations such as generation of two-dimensional codes of equipment, inquiry of historical data, system management, model display and the like.

It should be noted that, the communication modes between the devices and/or apparatuses may refer to the prior art, and the application is not limited herein.

The following description will be made by taking an architecture diagram of the inspection system shown in fig. 4 as an example, where the method includes the following steps:

s501, the patrol personnel scans the two-dimensional code of the target equipment of the machine room by using the mobile intelligent terminal to trigger a login request of entering a patrol page of the target equipment in the applet to be sent to the data acquisition server, wherein the login request carries the position information of the mobile intelligent terminal. The patrol page may, for example, fill in a patrol form of the target device, view information of the target device, and so on.

S502, the data acquisition server side judges whether the mobile intelligent terminal is located at the target equipment according to the acquired position information of the mobile intelligent terminal. If the target equipment is located, returning data for displaying the inspection page to the mobile intelligent terminal, namely executing step S503; if the target equipment is not located, outputting prompt information for prompting the user to go to the target equipment to the mobile intelligent terminal, so that the patrol personnel can re-trigger the login request after moving to the target equipment.

The above manner of determining whether the obtained location information of the mobile intelligent terminal is located at the target device may be, for example, determining whether the obtained location information of the mobile intelligent terminal is smaller than a preset effective inspection distance, where the preset effective inspection distance may be, for example, 100 meters around the target device. Specifically, if the position information of the mobile intelligent terminal acquired by the data acquisition server is 50 meters away from the target device and is less than the preset effective inspection distance, the step S503 is continuously executed.

And S503, the patrol personnel fills the patrol form of the target equipment in the patrol page of the target equipment in the applet and sends the patrol form to the data acquisition server.

It should be understood that a series of operations may be performed in the applet described above with respect to the target device, including but not limited to viewing basic information of the target device, filling in a patrol form or log of the target device, taking a live photograph, authorizing a location, submitting a patrol form.

The routing list includes, but is not limited to, characteristic parameters of the target device, such as temperature, power, noise, and the like, and an operation status label of the device. The operation state label of the device refers to the operation state of the target device, such as normal or abnormal, judged by the inspection staff after the inspection of the target device is finished.

It should be understood that the basic information of the target device is entered in advance through the PC and stored in the database.

S504, the data acquisition server side archives the inspection record of the target equipment according to the acquired inspection form of the target equipment, and sends the inspection record to a database for storage. In addition, the data acquisition server side can judge whether the polling result is abnormal or not based on the content in the polling list, and if the polling result is abnormal, the message pushing device is used for sending the polling abnormal message to the mobile intelligent terminal device of the responsible person of the target device.

The determination as to whether the polling result is abnormal may be made, for example, from the following two points. Judging whether the characteristic parameter value of the target equipment in the submitted routing inspection form is in a preset normal range, judging whether the operation label of the equipment in the routing inspection form of the target equipment is normal, and obtaining an abnormal routing inspection result if the operation label is not normal.

And S505, after the mobile intelligent terminal device of the person in charge of the target device receives the routing inspection abnormal message, the person in charge of the target device arranges maintenance personnel to perform maintenance processing on the target device.

And S506, after the maintainer arrives at the target equipment to perform maintenance processing, and after the normal state of the target equipment is recovered, the mobile intelligent terminal of the maintainer can be used for scanning the two-dimensional code of the target equipment to enter a small program, filling the processing log of the target equipment, and sending the processing log to the data acquisition server.

The contents of the processing log include, but are not limited to, a fault cause tag, characteristic parameters of the target device after processing, such as temperature, power, noise, etc. of the processed device.

And S507, the data acquisition server archives the acquired processing log of the target equipment to form an inspection record of the target equipment, and sends the inspection record to a database for storage.

S508, the data acquisition server side obtains the routing inspection record of the target equipment in the database, analyzes and processes the routing inspection data of the target equipment, and predicts the health state of the target equipment.

The data acquisition server side obtains N pieces of characteristic parameter information in the patrol records of the target equipment in the database within the target time period, and the health state of the target equipment is predicted by using the equipment patrol data processing method provided by the application.

And after the health state of the target equipment is predicted, the data acquisition server sends the health state of the target equipment to a database through the server for updating the health state label of the target equipment.

After the health state of the target equipment is predicted, if the predicted health state of the target equipment is lower than a preset health state, the data acquisition server side sends a prompt message of the health state of the target equipment to the mobile terminal equipment of a person in charge of the target equipment by using a message pushing device. And after the responsible person of the target equipment receives the prompt message, arranging to inspect the target equipment again.

It should be understood that the above step S508 may be performed in an asynchronous manner with the steps S501 to S507.

The equipment inspection method provided by the embodiment of the application carries out inspection in a mode of scanning the two-dimensional code of the target equipment, sets position judgment and abnormal message pushing, predicts the health state of the target equipment through analysis of obtained inspection data, and visually displays the predicted health state of the target equipment. If the predicted health state of the target equipment is lower than the preset health state, the target equipment is patrolled once more, equipment faults are avoided, the equipment utilization rate is improved, and the operation reliability of the data center machine room is further improved.

Fig. 5 is a schematic structural diagram of an apparatus inspection data processing apparatus according to an embodiment of the present application. As shown in fig. 5, the apparatus includes: a first obtaining module 11, a preprocessing module 12, a second obtaining module 13, and a prediction module 14. Optionally, the apparatus may further include: an update module 15 and/or an output module 16.

A first obtaining module 11, configured to obtain data sets of N feature parameters of a target device obtained by polling the target device within a target time period, where the data sets of the feature parameters include M data of the feature parameters obtained by M times of polling within the target time period; n is an integer of 1 or more, and M is an integer of 2 or more;

a preprocessing module 12, configured to preprocess data in the data set of each feature parameter to obtain a data set after preprocessing of each feature parameter;

a second obtaining module 13, configured to obtain a health index observation sequence of the target device in the target time period based on the data set after the preprocessing of each feature parameter, where the health index observation sequence includes M elements, each element corresponds to one polling, and is used to represent a health index of the target device obtained based on data of N feature parameters obtained by the polling;

and the predicting module 14 is used for predicting the health state of the target equipment according to the health index observation sequence.

In a possible implementation manner, the second obtaining module 13 is specifically configured to obtain a weight of each of the feature parameters; adding products of the data of each characteristic parameter and the weight obtained by the same inspection in the data set after the pretreatment of each characteristic parameter to obtain a health index of the target equipment corresponding to the inspection; and generating a health index observation sequence of the target equipment in the target time period by using the health index of the target equipment corresponding to each inspection.

For example, the second obtaining module 13 is specifically configured to generate a data matrix including data of all the characteristic parameters according to the data set of all the characteristic parameters; and processing the data matrix by using a principal component analysis method to obtain the weight of each characteristic parameter.

In one possible implementation, the prediction module 14 is specifically configured to predict the health status of the target device by using the health index observation sequence and a hierarchical model for predicting the health status by using a viterbi algorithm; the hierarchical model is represented by an initial health state vector, a health state transition probability matrix and an observation probability matrix. Illustratively, the hierarchical model is a hidden markov HMM based on an expectation-maximization algorithm, which is constructed according to a preset device health state and trained by using a sample health index observation sequence.

In a possible implementation manner, the updating module 15 is configured to update the tag of the health status of the target device in the device information of the target device after the predicting module 14 predicts the health status of the target device according to the health index observation sequence.

In a possible implementation manner, the output module 16 is configured to, after the predicting module 14 predicts the health state of the target device according to the health index observation sequence, output a prompt message of the health state of the target device if the health state of the target device is lower than a preset health state, so as to trigger a polling task of polling the target device again.

The device inspection data processing device provided by the application can execute the device inspection data processing method in the method embodiment, and the implementation principle and the technical effect are similar, and are not repeated here.

Fig. 6 is a schematic structural diagram of an electronic device 600 provided in the present application. As shown in fig. 6, the electronic device 600 may include: at least one processor 601, a memory 602. The electronic device may be a device with processing capability, such as a terminal, a server, a computer device, or the like, or may be a chip, a chip module, or the like with processing capability.

A memory 602 for storing programs. In particular, the program may include program code including computer operating instructions.

The memory 602 may comprise high-speed RAM memory, and may also include non-volatile memory (non-volatile memory), such as at least one disk memory.

The processor 601 is configured to execute computer-executable instructions stored in the memory 602 to implement the device inspection data processing method described in the foregoing method embodiments. The processor 601 may be a Central Processing Unit (CPU), an Application Specific Integrated Circuit (ASIC), or one or more Integrated circuits configured to implement the embodiments of the present Application.

The electronic device 600 may also include a communication interface 603 so that external devices may be communicatively interacted with via the communication interface 603. The external device may be, for example, a cell phone, tablet, etc.

In a specific implementation, if the communication interface 603, the memory 602, and the process 601 are implemented independently, the communication interface 603, the memory 602, and the process 601 may be connected to each other through a bus and perform communication with each other. The bus may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, an Extended ISA (EISA) bus, or the like. Buses may be classified as address buses, data buses, control buses, etc., but do not represent only one bus or type of bus.

Alternatively, in a specific implementation, if the communication interface 603, the memory 602, and the process 601 are integrated into a chip, the communication interface 603, the memory 602, and the process 601 may complete communication through an internal interface.

The present application also provides a computer-readable storage medium, which may include: various media capable of storing program codes, such as a usb disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and in particular, the computer-readable storage medium stores program instructions, and the program instructions are used in the method in the foregoing embodiments.

The present application also provides a program product comprising execution instructions stored in a readable storage medium. The at least one processor of the electronic device 600 may read the execution instructions from the readable storage medium, and the execution of the execution instructions by the at least one processor causes the electronic device 600 to implement the device inspection data processing method provided by the various embodiments described above.

The application also provides a chip, wherein a computer program is stored on the chip, and when the computer program is executed by the chip, the equipment inspection data processing method provided by various embodiments is realized.

Other embodiments of the present application will be apparent to those skilled in the art from consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the invention following, in general, the principles of the application and including such departures from the present disclosure as come within known or customary practice within the art to which the invention pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the application being indicated by the following claims.

It will be understood that the present application is not limited to the precise arrangements described above and shown in the drawings and that various modifications and changes may be made without departing from the scope thereof. The scope of the application is limited only by the appended claims.

Claims

1. A data processing method for equipment inspection is characterized by comprising the following steps:

acquiring a health index observation sequence of the target equipment in the target time period by adopting the data set preprocessed by each characteristic parameter, wherein the health index observation sequence comprises M elements, each element corresponds to one inspection and is used for representing the health index of the target equipment acquired on the basis of the data of the N characteristic parameters acquired by the inspection;

2. The method of claim 1, wherein the obtaining a health index observation sequence of the target device over the target time period using the pre-processed data set of each of the characteristic parameters comprises:

acquiring the weight of each characteristic parameter;

adding products of the data of the characteristic parameters and the weights, which are obtained by the same inspection in the data set after the pretreatment of the characteristic parameters, to obtain the health index of the target equipment corresponding to the inspection;

3. The method of claim 2, wherein the obtaining the weight of each of the feature parameters comprises:

4. The method of any one of claims 1-3, wherein predicting the health status of the target device based on the health index observation sequence comprises:

5. The method of claim 4, wherein the hierarchical model is a hidden Markov HMM based on an expectation-maximization algorithm constructed and trained using a sample health index observation sequence according to a preset device health state.

6. The method of claim 4, wherein after predicting the health status of the target device based on the health index observation sequence, the method further comprises:

7. The method of claim 4, wherein after predicting the health status of the target device based on the health index observation sequence, the method further comprises:

and if the health state of the target equipment is lower than the preset health state, outputting prompt information of the health state of the target equipment to trigger an inspection task for inspecting the target equipment again.

8. An apparatus inspection data processing apparatus, the apparatus comprising:

9. An electronic device, comprising: a processor, and a memory communicatively coupled to the processor;

the memory stores computer-executable instructions;

the processor executes computer-executable instructions stored by the memory to implement the method of any of claims 1-7.

10. A computer readable storage medium having computer executable instructions stored thereon, which when executed by a processor, are configured to implement the device inspection data processing method according to any one of claims 1 to 7.