CN106228176A - A kind of method and system based on floor data detection equipment state - Google Patents

Abstract

The invention discloses a kind of method and system based on floor data detection equipment state, belong to equipment health state evaluation and electric powder prediction, described method includes: obtain the floor data of equipment and the importance degree that each floor data is corresponding；Floor data is normalized and obtains normalization floor data；The health degree that normalization floor data is corresponding is obtained by healthy SOM model；Importance degree and health degree to each floor data are weighted merging to obtain equipment health degree.The present invention proposes and applies floor data to the method detecting equipment state, solve and prior art uses emulation data equipment state carries out the problem big with reality application gap that detection exists；It addition, present invention further contemplates that different parameters is different to the contribution degree of equipment state, by being weighted merging by health degree and importance degree, obtain equipment health degree.

Description

A kind of method and system based on floor data detection equipment state

Method field

The present invention relates to equipment health state evaluation and electric powder prediction, particularly relate to a kind of based on floor data detection The method and system of equipment state.

Background technology

How airplane equipment being carried out health state evaluation, realizing condition maintenarnce for ground maintenance personnel provides decision-making to prop up Hold, be one of the study hotspot of civil aircraft health management arts.Currently for airplane equipment health evaluating also in rise a step Section, existing a few studies is only to utilize emulation data, bigger with practical engineering application gap.

QAR(Quick Access Recorder, quick access recorder) it is civil aircraft onboard flight data recording equipment, For recording aircraft floor data in day-to-day operation.QAR equipment is arranged in electronic compartment, can record aircraft continuously and be up to The original flying quality of 600 hours, gathers hundreds of to thousands of parameters simultaneously, comprehensively the work of each second in record aircraft flight Condition.After aircraft flight, QAR equipment is taken out by ground maintenance personnel, transfers to data analyst to be analyzed.From 20th century Since the seventies, the airline of developed country uses QAR data to carry out FOQA (such as AirFASE software), it is achieved Flight course playback and the inquiry of data time course.In recent years, CA and Beijing oceangoing voyage lead to information firm also develop QAR solve Analysis and application system, mainly solve QAR data decoding problem and provide support for pilot training, formulation flight plan.In view of QAR data have the advantages that data volume is big, empirical by force, temporal associativity is big, utilize QAR data to realize airplane equipment healthy State estimation has the highest technological value.

It addition, it is critical only that of airplane equipment health state evaluation, it is desirable to have the history decline data of equipment to be assessed, logical Cross and compare with history decline data, draw the current state situation of equipment to be assessed.Self-organizing Maps (SOM) is unsupervised learning A kind of important kind of the neutral net of method, it is possible to detect it by its input sample association regular mutual with input sample Between relation, and adjust network according to information self-adaptings of these input samples, make the later response of network and input sample This is consistent.Therefore, SOM is utilized can to realize the health state evaluation of equipment to be assessed.

Prior art 1(CN201110171401) multichannel transducing signal is carried out pretreatment and feature extraction, then Healthy quantitative evaluation and predicting residual useful life is realized based on SOM model modeling；Technical scheme 2(CN102606557A) pass through fault Observer obtains system and normally exports and the residual signals of output of degenerating, and utilizes residual error characteristic quantity to build self organizing maps model and obtains To corresponding hydraulic system health degree；Technical scheme 3(CN201410664864) to add residual error on the basis of scheme 2 adaptive Answer threshold generator, it is achieved thereby that Fault Isolation.

Existing 3 technical schemes use system emulation and test signal, is not for civil aircraft QAR data, its data Preprocess method is not suitable for QAR data characteristics.It addition, one-parameter, multiparameter all can be carried out certainly by technical scheme 1,2,3 Organising map modeling obtains its health degree, but its processing mode is for building parameter multi-C vector as input, does not considers different ginseng Several contribution degrees to system health state are different, the accuracy of possible impact evaluation.

The application motion to solve the technical problem that and to include 2 points: one is aloft to obtain airplane equipment shape the most difficult In the case of state parameter, application aircraft QAR data are as the data source of health evaluating, true and reliable, improve data after boat Engineer applied is worth；Two propose based on self organizing maps model and level Weighted Fusion for aircraft equipment QAR data Method, can effectively realize multiparameter equipment health state evaluation.Comprehensive above 2 points, can build aircraft based on QAR data and set Standby health state evaluation system, for ground maintenance, personnel provide analytical tool.

Summary of the invention

It is an object of the invention to provide a kind of method and system based on floor data detection equipment state, the present invention proposes Utilize the floor data as data source, equipment state is detected；It addition, present invention further contemplates that system is good for by different parameters The contribution degree of health state is different, by being weighted merging by health degree and importance degree, obtains equipment health degree, improves equipment The accuracy of health evaluating.

For achieving the above object, provide one according to an aspect of the present invention and detect equipment state based on floor data Method, described method includes: obtain the floor data of equipment and importance degree corresponding to each floor data；Floor data is entered Row normalized obtains normalization floor data；The health degree that normalization floor data is corresponding is obtained by healthy SOM model； Importance degree and health degree to each floor data are weighted merging to obtain equipment health degree.

Wherein, the floor data of described acquisition equipment and the step of importance degree corresponding to each floor data, including: according to Type, equipment and parameter mapping table, determine parameter to be obtained；The floor data of the correspondence that gets parms；According to parameter with important Degree mapping table, the importance degree of the floor data of the correspondence that gets parms.

Wherein, described floor data is normalized the step obtaining normalization floor data, including: to operating mode Data are normalized so that the value of described normalization floor data is between [-1,1].

Wherein, described floor data is normalized the step obtaining normalization floor data before, also include: Floor data is carried out mean value smoothing process, by floor data abnormality value removing therein, and by before and after floor data exceptional value The average of 2 is as the value of floor data exceptional value.

Wherein, before the described step being obtained health degree corresponding to normalization floor data by healthy SOM model, also Including: the parameter sample input SOM model being under health status by described equipment, to be trained described SOM model, obtains To healthy SOM model.

Wherein, the described step being obtained health degree corresponding to normalization floor data by healthy SOM model, including: will Each described normalization floor data, as a healthy SOM model of input vector input, obtains a best weight value vector；Will Each described input vector is poor with described best weight value vector, obtains minimum quantization error MQE；To each described minimum quantization Error MQE is normalized and obtains health degree.

Wherein, based on following formula described minimum quantization error MQE is normalized and obtains health degree h:

$h=1-\frac{2}{\mathrm{\π}}\arctan (\sqrt{MQE}/2*\sqrt{Mean\left(MQE\right)}/\mathrm{\π}*\tan (1-h_s));$

Wherein, Mean (MQE) represents the average of minimum quantization error MQE, h_sRepresent the initial health degree of parameter sample.

Wherein, it is weighted merging to be set to the described importance degree to each floor data and health degree based on following formula Standby health degree:

$H=\underset{i=1}{\overset{i=n}{\Σ}}{\mathrm{\β}}_i{h}_i;$

Wherein, β₁,...β_i,...,β_nRepresent the importance degree of each floor data, h₁,...h_i,...,h_nRepresent each work The health degree of condition data, β₁+...+β_i+...+β_n=1, n represent the quantity of floor data.

Provide a kind of system based on floor data detection equipment state according to another aspect of the present invention, including: Data acquisition module, the first normalized module, health degree acquisition module and Weighted Fusion computing module；Data acquisition mould Block, for obtaining the floor data of equipment and the importance degree that each floor data is corresponding；First normalized module, for right Floor data is normalized and obtains normalization floor data；Health degree acquisition module, for obtaining by healthy SOM model Take the health degree that normalization floor data is corresponding；Weighted Fusion computing module, is used for the importance degree to each floor data and is good for Kang Du is weighted merging to obtain equipment health degree.

Wherein, described data acquisition module includes: parameter determination unit, parameter acquiring unit and importance degree acquiring unit； Parameter determination unit, for according to type, equipment and parameter mapping table, determining parameter to be obtained；Parameter acquiring unit, is used for The floor data of the correspondence that gets parms；Importance degree acquiring unit, for according to parameter and importance degree mapping table, get parms correspondence The importance degree of floor data.

Wherein, described first normalized module performs following operation: be normalized floor data so that The value of described normalization floor data is between [-1,1].

Wherein, described system also includes: mean value smoothing processing module, is used in the first normalized module operating mode number Before being normalized and obtaining normalization floor data, perform following operation: floor data is carried out at mean value smoothing Reason, by floor data abnormality value removing therein, and using the average of 2 before and after floor data exceptional value as described operating mode number Value according to exceptional value.

Wherein, described system also includes: healthy SOM model acquisition module, for passing through health at health degree acquisition module Before SOM model obtains the health degree that normalization floor data is corresponding, perform following operation: described equipment is in health status Under parameter sample input SOM model so that SOM model is trained, obtain healthy SOM model.

Wherein, described health degree acquisition module includes: best weight value vector acquiring unit, minimum quantization error calculate single Unit, the second normalized unit；Best weight value vector acquiring unit, is used for each described normalization floor data as one The healthy SOM model of individual input vector input, obtains a best weight value vector；Minimum quantization error computing unit, for by every Individual described input vector is poor with described best weight value vector, obtains minimum quantization error MQE；Second normalized unit, Health degree is obtained for each described minimum quantization error MQE is normalized.

Wherein, described second normalized unit is normalized place based on following formula to described minimum quantization error MQE Manage and obtain health degree h:

$h=1-\frac{2}{\mathrm{\π}}\arctan (\sqrt{MQE}/2*\sqrt{Mean\left(MQE\right)}/\mathrm{\π}*\tan (1-h_s));$

Wherein, Mean (MQE) represents the average of minimum quantization error MQE, h_sRepresent the initial health degree of parameter sample.

Wherein, described Weighted Fusion computing module based on following formula to the described importance degree to each floor data and health degree It is weighted merging to obtain equipment health degree:

$H=\underset{i=1}{\overset{i=n}{\Σ}}{\mathrm{\β}}_i{h}_i;$

Wherein, β₁,...β_i,...,β_nRepresent the importance degree of each floor data, h₁,...h_i,...,h_nRepresent each work The health degree of condition data, β₁+...+β_i+...+β_n=1, n represent the quantity of floor data.

A kind of method and system based on floor data detection equipment state provided by the present invention, the present invention proposes should With floor data as the data source of detection equipment state, first, choose and type, parameter that equipment state is relevant, according to often The significance level distribution importance degree of individual parameter；Then, parameter normal value and decline value are separately input to healthy SOM model, Health degree under corresponding states；Finally, by relevant parameter health degree belonging to equipment and importance degree Weighted Fusion, currently set Standby health status.

Accompanying drawing explanation

Fig. 1 is SOM model schematic of the prior art；

Fig. 2 is the schematic flow sheet of the method based on floor data detection equipment state according to the present invention；

Fig. 3 is the schematic flow sheet of step S100 of the present invention；

Fig. 4 is the schematic flow sheet of step S300 of the present invention；

Fig. 5 is the structural representation of the system based on floor data detection equipment state of the present invention；

Fig. 6 is the structural representation of the data acquisition module of the present invention；

Fig. 7 is the structural representation of the health degree acquisition module of the present invention.

Detailed description of the invention

For making the purpose of the present invention, method scheme and advantage of greater clarity, below in conjunction with detailed description of the invention and join According to accompanying drawing, the present invention is described in more detail.It should be understood that these describe the most exemplary, and it is not intended to limit this Bright scope.Additionally, in the following description, eliminate the description to known features and method, to avoid unnecessarily obscuring this The concept of invention.

It should be understood that Self-organizing Maps SOM(self-organizing map) it is that Helsinki, Netherlands university is neural Digerait Kohonen teaches the competitive mode neutral net proposed in nineteen eighty-two, and the most also referred to as Kohonen self-organizing feature reflects Penetrate.The learning process of its simulation cerebral neuron cell, is a kind of unsupervised learning method.SOM algorithm is to data vector While quantization, additionally it is possible to the dimensionality reduction realizing data maps, this mapping has the premium properties that topological relation keeps, thus extensively It is applied to dimension reduction and visualization field.As it is shown in figure 1, SOM network structure only has input layer and output layer two-layer.If input vector X Dimension be n dimension, then input layer has n node, n node to be respectively each component of input vector X；Output layer is by m god Forming two-dimensional planar array through unit, its arrangement mode can be rectangular arranged, hexagonal array or random alignment；Input layer is each It is fully connected between node and each neuron of output layer.Briefly, SOM is aiming at each input vector, all in low-dimensional On output plane (generally two dimension) to find out a point the most corresponding, after this corresponding relation is set up, just claim this input vector Have activated an output neuron.

Fig. 2 is the schematic flow sheet of the method based on floor data detection equipment state according to the present invention.

As in figure 2 it is shown, the method based on floor data detection equipment state of the present invention, described method includes walking as follows Rapid:

Step S100, obtains the floor data of equipment and the importance degree that each floor data is corresponding.

In this step, first, select equipment to be assessed, by QAR(Quick Access Recorder, quickly access Recorder) floor data of collecting device, therefrom obtain the floor data of described equipment and the weight that each floor data is corresponding Spend.

Here, floor data refers to that equipment is in the actual operating data of mission phase, when described floor data is one section Interior floor data, for example, it may be equipment floor data being in steady state phase etc..

Step S200, is normalized floor data and obtains normalization floor data.

In this step, each described floor data is normalized and obtains normalization floor data；Concrete, right Each described floor data is normalized so that the value of each described normalization floor data is between [-1,1].

In an optional embodiment, before step S200, also include: floor data is carried out mean value smoothing process, By floor data abnormality value removing therein, and using the average of 2 before and after floor data exceptional value as floor data exceptional value Value.In this step, owing to, in floor data, occasional occurs that limited pulses disturbs, i.e. there is floor data exceptional value, for Ensure the effectiveness of floor data to greatest extent, before floor data is normalized, need to floor data first Carry out mean value smoothing process, floor data exceptional value therein is rejected, with 2 equal before and after operating mode data outliers It is worth the value as floor data exceptional value.

Step S300, obtains, by healthy SOM model, the health degree that normalization floor data is corresponding.

In this step, by healthy for the input of each described normalization floor data SOM model, obtain each normalization operating mode number According to corresponding health degree.

In an optional embodiment, before step S300, also include: described equipment is in the ginseng under health status Numerical example input SOM model, to be trained described SOM model, obtains healthy SOM model.In this step, choose with aforementioned Multiple parameter samples (n parameter alpha altogether that multiple parameters are corresponding₁,α₂,α₃,...α_i,..α._n), it is designated as respectively: α₁'(t),α₂' (t),α₃'(t),...α_n(t) ', (wherein, α_i' (t) be i-th parameter alpha_iParameter sample).By defeated for each described parameter sample Enter SOM model so that described SOM model to be trained, respectively obtain multiple healthy SOM model, be designated as SOM₁,SOM₂, ...SOM_n。

Here, the health degree h of described parameter sample_sCan be 0.99, or be other numerical value close to 1.

Step S400, importance degree and health degree to each floor data are weighted merging to obtain equipment health degree.

In this step, importance degree and health degree to each floor data are weighted merging, and obtain equipment health degree.

Concrete, it is weighted merging to obtain to the described importance degree to each floor data and health degree based on following formula Equipment health degree:

$H=\underset{i=1}{\overset{i=n}{\Σ}}{\mathrm{\β}}_i{h}_i;$

Wherein, β₁,...β_i,...,β_nRepresent the importance degree of each floor data, h₁,...h_i,...,h_nRepresent each work The health degree of condition data, β₁+...+β_i+...+β_n=1, n represent the quantity of floor data.

Fig. 3 is the schematic flow sheet of step S100 of the present invention.

As it is shown on figure 3, described step S100 farther includes following steps:

Step S110, according to type, equipment and parameter mapping table, determines parameter to be obtained.

In this step, concrete, for different aircraft models, different unit types, parameter to be obtained is different, therefore Need to previously generate type, equipment and the mapping table of parameter, based on described mapping table, determine multiple parameters to be obtained (n altogether Parameter), it is set to: α₁,α₂,α₃,...α_i,...α_n。

Step S120, the floor data of the correspondence that gets parms.

In this step, based on each the described parameter got, further, obtain the work that each the plurality of parameter is corresponding Condition data.

Step S130, according to parameter and importance degree mapping table, the importance degree of the floor data of the correspondence that gets parms.

In this step, different parameters, its importance degree is different, therefore needs to previously generate parameter and importance degree mapping table, base In mapping table, determine the importance degree (n importance degree altogether) of each floor data corresponding to each parameter, be respectively as follows: β₁,β₂, β₃,...β_n。

Fig. 4 is the schematic flow sheet of step S300 of the present invention.

As shown in Figure 4, described step S300 farther includes following steps:

Step S310, using each described normalization floor data as a healthy SOM model of input vector input, obtains One best weight value vector.

In this step, by each described parameter sample α₁'(t),α₂'(t),α₃'(t),...α_nT () ' is as an input Vector inputs SOM respectively₁,SOM₂,...SOM_n, obtaining a corresponding best weight value vector, described best weight value vector is The weight vector the shortest with described input vector distance.

Step S320, by poor to each described input vector and described best weight value vector, obtains minimum quantization error MQE。

In this step, by poor to each described input vector and described best weight value vector, it is calculated minimum quantization by mistake Difference MQE.Concrete, based on following formula minimum quantization error MQE:

MQE=| | X_input-W_bmu||

Wherein, X_inputFor normalization floor data, W_bmuIt it is best weight value vector.

Step S330, is normalized each described minimum quantization error MQE and obtains health degree.

In this step, based on following formula, best weight value vector M QE is normalized between [0,1]:

$h=1-\frac{2}{\mathrm{\π}}\arctan (\sqrt{MQE}/2*\sqrt{Mean\left(MQE\right)}/\mathrm{\π}*\tan (1-h_s));$

Wherein, Mean (MQE) represents the average of minimum quantization error MQE, h_sRepresent the initial health degree of parameter sample.

As it has been described above, describe the method based on floor data detection equipment state of the present invention in detail, the present invention utilizes Equipment state, as data source, is detected by floor data；It addition, the contribution that different parameters is to the evaluation of equipment health status Degree is different, for the importance degree that different parametric distributions is different, it is proposed that health degree and the method for importance degree Weighted Fusion, carries The high accuracy of equipment health state evaluation.

Fig. 5 is the structural representation of the system based on floor data detection equipment state of the present invention.

As it is shown in figure 5, the system based on floor data detection equipment state of the present invention, described system includes: data obtain Delivery block the 100, first normalized module 200, health degree acquisition module 300 and Weighted Fusion computing module 400.

Data acquisition module 100, for obtaining the floor data of equipment and the importance degree that each floor data is corresponding.

First normalized module 200 is connected with described data acquisition module 100, for floor data is carried out normalizing Change processes and obtains normalization floor data.Described first normalized module 200 performs following operation: carry out floor data Normalized so that the value of described normalization floor data is between [-1,1].

Health degree acquisition module 300 is connected with described first normalized module 200, for by healthy SOM model Obtain the health degree that normalization floor data is corresponding.

Weighted Fusion computing module 400 is connected with described data acquisition module 100 and health degree acquisition module 300 respectively, For the importance degree of each floor data and health degree being weighted fusion to obtain equipment health degree.

Concrete, Weighted Fusion computing module 400 based on following formula to the described importance degree to each floor data and health Degree is weighted merging to obtain equipment health degree:

$H=\underset{i=1}{\overset{i=n}{\Σ}}{\mathrm{\β}}_i{h}_i;$

Wherein, β₁,...β_i,...,β_nRepresent the importance degree of each floor data, h₁,...h_i,...,h_nRepresent each work The health degree of condition data, β₁+...+β_i+...+β_n=1, n represent the quantity of floor data.

Fig. 6 is the structural representation of the data acquisition module 100 of the present invention.

Wherein, described data acquisition module 100 includes: parameter determination unit 110, parameter acquiring unit 120 and importance degree Acquiring unit 130.

Parameter determination unit 110, for according to type, equipment and parameter mapping table, determining parameter to be obtained.

Parameter acquiring unit 120 is connected with described parameter determination unit 110, for the floor data of the correspondence that gets parms.

Importance degree acquiring unit 130 is connected with described parameter acquiring unit 120, for mapping with importance degree according to parameter Table, the importance degree of the floor data of the correspondence that gets parms.

In an optional embodiment, described system also includes: mean value smoothing processing module 500, at the first normalizing Floor data is normalized before obtaining normalization floor data by change processing module 200, performs following operation: to work Condition data carry out mean value smoothing process, by floor data abnormality value removing therein, and by 2 points before and after floor data exceptional value Average as the value of described floor data exceptional value.

In an optional embodiment, described system also includes: healthy SOM model acquisition module 600, at health degree Before acquisition module 300 obtains, by healthy SOM model, the health degree that normalization floor data is corresponding, perform following operation: use In the parameter sample input SOM model being under health status by described equipment so that SOM model to be trained, obtain health SOM model.

Fig. 7 is the structural representation of the health degree acquisition module 300 of the present invention.

As it is shown in fig. 7, described health degree acquisition module 300 includes: best weight value vector acquiring unit 310, minimum quantization Error calculation unit 320 and the second normalized unit 330.

Best weight value vector acquiring unit 310, is used for each described normalization floor data as an input vector The healthy SOM model of input, obtains a best weight value vector.

Minimum quantization error computing unit 320 is connected with described best weight value vector acquiring unit 310, for by each institute State input vector poor with described best weight value vector, obtain minimum quantization error MQE.

Second normalized unit 330 is connected with described minimum quantization error computing unit 320, for each described Minimum quantization error MQE is normalized and obtains health degree.

Wherein, described minimum quantization error MQE is normalized by the second normalized unit 330 based on following formula Obtain health degree h:

$h=1-\frac{2}{\mathrm{\π}}\arctan (\sqrt{MQE}/2*\sqrt{Mean\left(MQE\right)}/\mathrm{\π}*\tan (1-h_s));$

Wherein, Mean (MQE) represents the average of minimum quantization error MQE, h_sRepresent the initial health degree of parameter sample.

As it has been described above, describe the system based on floor data detection equipment state of the present invention in detail, the present invention utilizes Equipment state, as data source, is detected by floor data；It addition, the contribution that different parameters is to the evaluation of equipment health status Degree is different, for the importance degree that different parametric distributions is different, it is proposed that health degree and the method for importance degree Weighted Fusion, carries The high accuracy of equipment health state evaluation.

In sum, the invention provides a kind of method and system based on floor data detection equipment state, the present invention Middle propose application floor data detection equipment state, first, choose and type, parameter that equipment state is relevant, according to each ginseng The significance level distribution importance degree of number；Then, parameter normal value and decline value being separately input to healthy SOM model, it is right to obtain Answer the health degree under state；Finally, by relevant parameter health degree belonging to equipment and importance degree Weighted Fusion, current device is obtained Health status.

It should be appreciated that the above-mentioned detailed description of the invention of the present invention is used only for exemplary illustration or explains the present invention's Principle, and be not construed as limiting the invention.Therefore, that is done in the case of without departing from the spirit and scope of the present invention is any Amendment, equivalent, improvement etc., should be included within the scope of the present invention.Additionally, claims purport of the present invention Whole within containing the equivalents falling into scope and border or this scope and border change and repair Change example.

Claims (10)

1. a method based on floor data detection equipment state, including:

The floor data of acquisition equipment and importance degree corresponding to each floor data；

Floor data is normalized and obtains normalization floor data；

The health degree that normalization floor data is corresponding is obtained by healthy SOM model；

Importance degree and health degree to each floor data are weighted merging to obtain equipment health degree.

Method the most according to claim 1, wherein, floor data and each floor data of described acquisition equipment are corresponding The step of importance degree, including:

According to type, equipment and parameter mapping table, determine parameter to be obtained；

The floor data of the correspondence that gets parms；

According to parameter and importance degree mapping table, the importance degree of the floor data of the correspondence that gets parms.

3. according to the method described in any one of claim 1-2, wherein, described by healthy SOM model acquisition normalization operating mode The step of the health degree that data are corresponding, including:

Using each described normalization floor data as a healthy SOM model of input vector input, obtain a best weight value Vector；

By poor to each described input vector and described best weight value vector, obtain minimum quantization error MQE；

Each described minimum quantization error MQE is normalized and obtains health degree.

Method the most according to claim 3, wherein, is normalized place based on following formula to described minimum quantization error MQE Manage and obtain health degree h:

$h=1-\frac{2}{\mathrm{\π}}\arctan (\sqrt{MQE}/2*\sqrt{Mean\left(MQE\right)}/\mathrm{\π}*\tan (1-h_s\left)\right);$

Wherein, Mean (MQE) represents the average of minimum quantization error MQE, h_sRepresent the initial health degree of parameter sample.

5. according to the method described in any one of claim 1-2, wherein, based on following formula to described important to each floor data Degree and health degree are weighted merging to obtain equipment health degree:

$H=\underset{i=1}{\overset{i=n}{\Σ}}{\mathrm{\β}}_i{h}_i;$

Wherein, β₁,...β_i,...,β_nRepresent the importance degree of each floor data, h₁,...h_i,...,h_nRepresent each operating mode number According to health degree, β₁+...+β_i+...+β_n=1, n represent the quantity of floor data.

6. a system based on floor data detection equipment state, including:

Data acquisition module (100), for obtaining the floor data of equipment and the importance degree that each floor data is corresponding；

First normalized module (200), obtains normalization floor data for being normalized floor data；

Health degree acquisition module (300), for obtaining, by healthy SOM model, the health degree that normalization floor data is corresponding；

Weighted Fusion computing module (400), for being weighted fusion to obtain to the importance degree of each floor data and health degree To equipment health degree.

System the most according to claim 6, wherein, described data acquisition module (100) including:

Parameter determination unit (110), for according to type, equipment and parameter mapping table, determining parameter to be obtained；

Parameter acquiring unit (120), for the floor data of the correspondence that gets parms；

Importance degree acquiring unit (130), for according to parameter and importance degree mapping table, the weight of the floor data of the correspondence that gets parms Spend.

8. according to the system described in any one of claim 6-7, wherein, described health degree acquisition module (300) including:

Best weight value vector acquiring unit (310), for defeated as an input vector using each described normalization floor data Enter healthy SOM model, obtain a best weight value vector；

Minimum quantization error computing unit (320), for by poor to each described input vector and described best weight value vector, obtaining To minimum quantization error MQE；

Second normalized unit (330), is good for for being normalized each described minimum quantization error MQE Kang Du.

System the most according to claim 8, wherein, described second normalized unit (330) based on following formula to described Minimum quantization error MQE is normalized and obtains health degree h:

$h=1-\frac{2}{\mathrm{\π}}\arctan (\sqrt{MQE}/2*\sqrt{Mean\left(MQE\right)}/\mathrm{\π}*\tan (1-h_s\left)\right);$

Wherein, Mean (MQE) represents the average of minimum quantization error MQE, h_sRepresent the initial health degree of parameter sample.

10. according to the system described in claim 6-7,9 any one, wherein, described Weighted Fusion computing module (400) based under The described importance degree to each floor data and health degree are weighted merging to obtain equipment health degree by formula:

$H=\underset{i=1}{\overset{i=n}{\Σ}}{\mathrm{\β}}_i{h}_i;$

Wherein, β₁,...β_i,...,β_nRepresent the importance degree of each floor data, h₁,...h_i,...,h_nRepresent each operating mode number According to health degree, β₁+...+β_i+...+β_n=1, n represent the quantity of floor data.