CN107607342A - The healthy efficiency detection method of Air Conditioning Facilities device cluster - Google Patents
The healthy efficiency detection method of Air Conditioning Facilities device cluster Download PDFInfo
- Publication number
- CN107607342A CN107607342A CN201710866822.1A CN201710866822A CN107607342A CN 107607342 A CN107607342 A CN 107607342A CN 201710866822 A CN201710866822 A CN 201710866822A CN 107607342 A CN107607342 A CN 107607342A
- Authority
- CN
- China
- Prior art keywords
- mrow
- msub
- msubsup
- signal
- mfrac
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Landscapes
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention discloses a kind of healthy efficiency detection method of Air Conditioning Facilities device cluster, installs vibrating sensor in computer room first, and the vibration signal received is passed into data collecting system, gathers out the vibration signal of equipment;Then disjunctive model W is calculated by the high-order accumulation blind source separation algorithm of mutual information under natural gradient1、W2.Blind source separating is carried out to migration fractionation signal using the disjunctive model calculated, so as to extract the time-frequency characteristics of equipment fault signal;Then signal characteristic identification is carried out according to the position of these characteristic parameters and sensor, judges fault type and equipment working state, respective handling is made according to the status information analyzed in time, so as to reach equipment health detection and improve the purpose of equipment operating efficiency.The present invention can when extracting individual equipment operation in the case of plurality of devices is operated together vibration signal characteristics, so as to correctly judging the health status of equipment and improve operational efficiency.
Description
Technical field
The invention belongs to Diagnosis Technique field, particularly relates to a kind of healthy energy of Air Conditioning Facilities device cluster
Imitate detection method.
Background technology
Complicated, under the trend of scale in equipment, Air Conditioning Facilities are in order to save space resources and be easy to maintenance management, often
Different types of unit is often installed concentratedly in the same space.Domestic Air-condition systematic energy efficiency ratio average value is 2.5 left at present
The right side, minimum are only 1.7, much fall the efficiency value less than single air conditioner unit.Compared with foreign countries, China utilizes in energy efficiency of equipment
Aspect is constantly in inferior position, and external energy efficiency of equipment utilization rate can reach more than 75%, but domestic only 30% or so.Cause
The main reason for this situation is that various kinds of equipment vibration is transmitted in computer room, and secondary vibration caused by noise makes under equipment operating efficiency
Drop, mechanical breakdown takes place frequently, so as to reduce the efficiency of unit.Therefore vibration monitoring is carried out to device cluster in Air Conditioning Facilities, in time
It was found that the initial failure of single device, is to ensure air-conditioner set health, the powerful guarantee of Effec-tive Function.
Equipment Condition Monitoring System is by analyzing all kinds of state characteristic informations, so as to know the running status of equipment, as far as possible
The damage of equipment is reduced, so as to avoid unnecessary efficiency waste and financial losses.Therefore, efficient, accurate equipment health energy
Effect monitoring system can not only ensure the normal operation of Air Conditioning Facilities, can also improve equipment operating efficiency, reduce maintenance shut-downs
Expense.But because the state monitoring method of sensor can only be monitored to single device, mutually it is concerned with for plurality of devices
Disturb, the influence that the factor such as unit operation efficiency change is brought can not be diagnosed accurately and real-time.Therefore, to more vibration source signals
It is unknown, and in the case of air-conditioner set power output dynamic change, can extract single fighter failure source signal characteristics turns into key
Technical barrier.
The content of the invention
The purpose of the present invention, which is that, overcomes deficiencies of the prior art, difficult for the mixed signal of multiple vibration sources
With the problem of quick and precisely identification equipment running status, it is proposed that a kind of air-conditioning of new migration fractionation signal fault identification technology
The healthy efficiency detection method of calculator room equipment group, both sensitive to faint initial failure signal, and can identification various faults feature;This
Invention can in the case of plurality of devices is operated together extract individual equipment operation when vibration signal characteristics, so as to correctly judge set
Standby health status simultaneously improves operational efficiency.And method proposed by the present invention improves the performance of blind separation algorithm, improves
Vibration signal separates precision, makes separating property in practical application more stable, reliable, accurate and efficient, so as to accurately have
Effect judges the health and energy efficiency state of unit in time, reaches early detection equipment fault, accurately identifies fault type and efficiency
Horizontal purpose.
To achieve the above object, the present invention adopts the following technical scheme that.
A kind of healthy efficiency detection method of Air Conditioning Facilities device cluster of the present invention, it is characterised in that comprise the following steps:
Step 1:First, the key position such as the compressor of air-conditioner set, shaft coupling arrangement vibration measuring point in computer room, is used
Each part of vibration spectrum analyzer measuring unit it is axial, vertically to level rumble spectrum, signal is finally passed into number
According to acquisition system;
Step 2:Then, the vibration signal stored gathering is as input signal X (t), with three rank accumulationsWith
Quadravalence accumulationRepresent output signal y (t), edge entropy H (yi), and by three rank accumulationsAccumulated with quadravalence
AmountEdge entropy H (y are obtained with Edgeworth series expansionsi) display expression formula, using natural water surface coatings respectively to mutual
W in information representation formula1、W2Derivation, thus calculate separation matrix W1And W2, finally in non-linear source signal blind separation mould
On the basis of type, vibration source signal y (t) is isolated using the separation matrix solved;
Step 3:Finally, the source signal isolated is subjected to signature analysis, so that it is determined that the major failure source of computer room and root
Healthy energy efficiency monitoring system is fed back in time according to the signal of vibrating feature recognition running status of separation, and by equipment running status information
System, so as to reach monitoring health in real time and the purpose of energy efficiency state.
As a preferred embodiment of the present invention, the vibration measuring point arranged in the step 1 in equipment machine room is to shake
Dynamic sensor.
As another preferred scheme of the present invention, the step 2 falls into a trap point counting from matrix W1、W2It is divided into following steps:
(1) determines vibrating sensor best position according to air-conditioner set key position, is entered using data collecting system
Row air-conditioner set vibrating data collection, by the vibrating sensing for being arranged in the key positions such as compressor on air-conditioner set, shaft coupling
Device gathers the vibration signal X of air-conditioner setj(t), j=1,2 ..., n, wherein, j is channel number, and n is positive integer;
(2) air-conditioner set vibration signal Xs of the according to blind source separation algorithm of the invention to collectionj(t) blind source point is carried out
From so as to obtain original vibration signal source sj(t) approximate signal source yj(t), wherein, j=1,2 ..., n;
(3) is by yiEdge entropy with three rank accumulationsWith quadravalence accumulationExpression is used in combination
Edgeworth series expansions:
WhereinRespectively yiThree ranks and quadravalence accumulation;
(4) output y is expressed as mutual information by according to the definition of mutual information in information theory with separation parameter:
(5) it is nonlinear activation function that, which defines phasor function f (y), and calculates phasor function according to sigmoid functions:
(6) utilizes the gradient descent algorithm of Minimum mutual information, respectively to W in formula (b)1And W2Derivation:
Wherein D (u)=diag [P 'y1(y1) ..., P 'yn(yn)], diag [] is that vector is turned into diagonal matrix, diagonally
Element is the respective components of vector, and η is Learning Step;
(7) finally brings formula (c) in formula (d) and (e) into, solves separation matrix W1And W2:
W1(k+1)=W1(k)+η (I-f (y) yT)W1(k) (f),
W2(k+1)=W2(k)+η (I-f (y) yT- W1(k)T f(y)yT)W2(k) (g),
Wherein η is adaptive learning step-length;
(8) solves source signal y (t) using separation matrix separation collection signal X (t);
(9) draws;
(10) carries out signature analysis according to source signal y (t) is isolated;
(11) terminates.
As another preferred scheme of the present invention, the vibration spectrum analyzer in the step 1 is vibrated using CSI2310
Spectrum analyzer.
The Air Conditioning Facilities of B1 are chosen in the present invention, including two source pumps, three water circulating pumps and each
Class pipeline is as calculated examples.Because the multi-source Separation by vibration method of the present invention has good time-frequency locality and anti-interference,
Therefore migration fractionation signal can be separated to greatest extent, according to isolating the Energy distribution of signal, frequency composition and during pulse
Every out of order type can be judged, so that it is determined that in computer room equipment running status.
In the engineer applied present invention, the long-term detection device of vibration equipment acquisition system is installed typically all in Air Conditioning Facilities
Running status, the vibration signal arrived according to actual acquisition, by the present invention signal processing method, isolate fault-signal
Time-frequency characteristics, you can the working condition of equipment is judged, so as to reach the purpose of equipment health energy efficiency monitoring.
The present invention in equipment by installing vibrating sensor and passing to data collecting system, to the migration fractionation of collection
High-order accumulation mutual information blind source separation algorithm of the signal based on natural gradient carries out signal blind source separating processing.At this signal
Reason method can reflect the time-frequency characteristics of equipment fault signal exactly, be monitoring device working condition and equipment fault early-warning
Effective means is provided.
The method that Higher Order Cumulants are used in the present invention, can reduce influence of the evaluated error to separating resulting, improve blind
The performance of separation algorithm, improves computational accuracy, can judge fault type by the time-frequency characteristics parameter of signal after processing
And equipment working state, the problem that General Troubleshooting method is difficult to be precisely separating source signal is avoided, therefore this method can
Strong operability, diagnosis is quick, precision is high.
Compared with prior art, present invention has the advantages that:Constantly maximize in plant equipment, diversification, group
Today, product of the domestic enterprise to device cluster detecting system without maturation, rely primarily on import, Some Domestic software with it is external
Compared to larger gap being also present;In addition, external complete equipment health monitoring systems price is expensive, and shape is run with domestic equipment
State is not consistent more, prevents it from playing good effect.Meanwhile the design environment of external product and the working environment of domestic equipment
Difference, operational reliability of the equipment in adverse circumstances can not be improved.The present invention can rapidly and accurately analyze the work of equipment
Make state, ensure the security and high efficiency of equipment, improve equipment unhealthy condition early warning effect, reduce maintenance cost.Simultaneously
The present invention is equally effective to the fault detect under the complex space environments such as computer room, device cluster, particularly suitable for water resource heat pump, water pump
The fault detect of the equipment such as computer room, air-conditioning box, wind energy conversion system, air blower, detection and the maintenance cost of various kinds of equipment can be substantially reduced,
Improve equipment maintenance mode, improves the security reliability and high efficiency of equipment operation, therefore the remarkable in economical benefits brought.
Brief description of the drawings
In order that technical problem solved by the invention, technical scheme and beneficial effect are more clearly understood, below in conjunction with
The drawings and the specific embodiments, the present invention will be described in further detail.It should be appreciated that specific embodiment party described herein
Formula only to explain the present invention, is not intended to limit the present invention.
Fig. 1 is a kind of source signal blind separation signal of healthy efficiency detection method of Air Conditioning Facilities device cluster of the present invention
Figure.
Fig. 2 is No. 3 measuring point radial vibration mixed signal figures of the invention.
Fig. 3 is No. 3 measuring point axial vibration mixed signal figures of the invention.
Fig. 4 is No. 3 measuring point radial vibration mixed signal separation figures of the invention.
Fig. 5 is No. 3 measuring point axial vibration mixed signal separation figures of the invention.
Fig. 6 is a kind of signal handler calculation block of the healthy efficiency detection method of Air Conditioning Facilities device cluster of the present invention
Figure.
Fig. 7 is a kind of health monitoring of equipment technology path of the healthy efficiency detection method of Air Conditioning Facilities device cluster of the present invention
Figure.
Embodiment
A kind of healthy efficiency detection method of Air Conditioning Facilities device cluster of the present invention, installs vibrating sensing in computer room first
Device, and the vibration signal received is passed into data collecting system, gather out the vibration signal of equipment;Then natural ladder is passed through
The high-order accumulation blind source separation algorithm of the lower mutual information of degree calculates disjunctive model W1、W2.Using the disjunctive model calculated to more
Source mixed signal carries out blind source separating, so as to extract the time-frequency characteristics of equipment fault signal;Then according to these characteristic parameters
Signal characteristic identification is carried out with the position of sensor, fault type and equipment working state is judged, is believed according to the state analyzed
Breath makes respective handling in time, so as to reach equipment health detection and improve the purpose of equipment operating efficiency.
The present invention's concretely comprises the following steps:
1) blind source separating of migration fractionations signal
Due to the complexity and number of devices, huge number of equipment working environment, therefore classical signal blind source separating is not
It is applicable this complicated, severe working environment.The target of blind source separating is exactly to seek one properly by observation signal x (t)
Nonlinear function g (x) so that source signal s (t) is recovered by it, i.e.,:
S (t)=[y1(t),y2(t),…ym(t)]=g [(x (t))] (1)
In formula, y (t) is referred to as s (t) estimated vector.
The present invention considers interference of the migration fractionation signal to fault-signal feature recognition, the model of the piece-rate system of use
For:
Y (t)=W1g(W2X (t))=s (t) (2)
Here g (u) acts on vector u each component, i.e. g (u)={ g (u as a scalar function1),g(u2),…,
g(un)}T;W1、W2For separation matrix, i.e., all it is reversible.
Formula (2) illustrates that under the conditions of mixed signal blind separation can be directly using the unknown mixed signal received, without elder generation
Test under knowledge and recover source signal from observation data, frequently as a kind of Signal De-noising Method;As shown in figure 1, X (t) is to pass
The signal phasor that sensor is gathered, S (t) are source signal vectors, and A is unknown hybrid matrix, and output Y (t) is S (t) estimation arrow
Amount.
2) equipment vibrating signals gather
With B1 equipment machine room, including two source pumps, three water circulating pumps and various pipes, unit is strong
Vibration is delivered to upstairs by building;In computer room, No. 1 each portion of source pump is measured using CSI2310 vibration spectrum analyzers
Part it is axial, vertically to level rumble spectrum;Data collecting system is by vibrating sensor, data collecting card, data signal
Simulator and computer are formed, and vibration signal is after data collecting card is changed, the letter that is exported sensor by data collecting card
Number it is sent in display and stores.
3) minimum mutual information in Algorithm for nonlinear blind source separation
Nonlinear blind signalseparation method mainly includes maximum entropy method (MEM) and minimum mutual information, is linear when mixing model
When, maximum entropy method (MEM) and Minimum mutual information are of equal value;When it is non-linear to mix model, have on both algorithm performances larger
Difference;Blind separation algorithm based on Minimum mutual information can separate multiple signal sources with less sensor, be suitable for more equipment
And the actual conditions that number of faults is unknown;According to the definition of mutual information in information theory, being represented with separation parameter for y can must be exported
Mutual information is as follows:
Wherein G=WA is global matrix, and H (y)=- E [log (Py(y) it is)] the edge entropy of piece-rate system output signal,
Therefore formula (3) can be write as:
The defects of with the increase of source number, the amount of calculation of the algorithm is excessive, has and is difficult to restrain, and stability is bad;Therefore
It is extremely important for improving algorithm performance to find effective method of estimation.
4) separation matrixes computational methods
In order to unknown in source of trouble quantity and in the case of being likely larger than number of sensors, according to the independence between information source
Property estimate relation and extract most significant characteristic value successively, the present invention utilizes on the basis of Minimum mutual information Natural Gradient Algorithm
The random character of source signal improves its algorithm.
Piece-rate system output signal y (t) and observation signal x (t) probability density expression formula is:
Wherein Py(y) be Y probability density function, because the probability density function of output signal belongs to unknown quantity, therefore
Firstly the need of calculating edge entropy H (yi), it otherwise can not obtain the explicit expression of mutual information.
Therefore by y in formula (b)iEdge entropy item with three rank accumulationsWith quadravalence accumulationRepresent simultaneously
It is used into Edgeworth series expansions:
WhereinRespectively yiThree ranks and quadravalence accumulation.
Then under natural gradient respectively to formula (b) in W1And W2Differentiate:
Phasor function f (y) is nonlinear activation function in formula (e), and calculates phasor function according to sigmoid functions:
Wherein D (u)=diag (P 'y1(y1) ..., P 'yn(yn)), diag [] is that vector is turned into diagonal matrix, diagonally
Element is the respective components of vector, and η is adaptive learning step-length.
Finally, formula (c) nonlinear activation function is brought into formula (e), just obtains the mutual information based on natural gradient
High-order accumulation blind separation algorithm:
W1(k+1)=W1(k)+η (I-f (y) yT)W1(k) (f),
W2(k+1)=W2(k)+η (I-f (y) yT- W1(k)Tf(y)yT)W2(k) (g),
Wherein η is adaptive learning step-length.
Edge entropy in mutual information is calculated using high-order accumulation, the signal of disjunctive model is calculated finally by natural water surface coatings
Processing routine computing block diagram is as shown in fig. 6, the calculation procedure has following steps:
1. is using the vibration signal of certain measuring point on the unit of collection as observation signal X (t);
2. is expressed as mutual information form by y is exported according to formula (b) with separation parameter;
3. calculates output y according to formula (a) with three ranks and quadravalence accumulationiEdge entropy, and with Edgeworth series exhibitions
Open;
4. is according to formula (d) and (e), using the Natural Gradient Algorithm of Minimum mutual information, respectively to W1、W2Derivation;
5. solves phasor function f (y) according to sigmoid functions
6. brings formula (c) into formula (e) and calculates disjunctive model W1, W2;
7. is by separation matrix W1、W2Bring formula (2) into and solve source signal;
8. draws
9. terminates.
5) separates signature analysis
Disjunctive model is recalculated with high-order accumulation after optimization, on the basis of blind separation theory, utilizes new separation
Model, so as to eliminate the interference in fault-signal identification, fault-signal characteristic parameter is excavated, it is last according to the failure isolated
Signal identification is out of order type and equipment working state, respectively as shown in Figure 4, Figure 5.As shown in figure 4, from the point of view of whole frequency spectrum,
Value of shaking maximum appears in 4 frequencys multiplication (200Hz), and other frequency component vibration velocities are almost negligible disregards, due to male rotor
Number is 4, and a rotation period can touch mill 4 times, big equivalent to 1 frequency multiplication, illustrates that rotor caused by Gear Processing manufacture assembling engages
It is bad;As shown in figure 5, occurring harmonic wave in 150-250Hz frequency bands, and there is peak-peak at 4 frequencys multiplication, the symmetrical width in both sides
Value is successively decreased, and is found after tested, and intrinsic frequency of the unit when shutting down is 203Hz, and frequency spectrum when running has peak value in 4 frequencys multiplication,
Illustrate that frequency has approached intrinsic frequency during unit operation, thus resonate.Equipment resonance can cause the deformation of internal structure with
Unbalanced stress, rotor part especially higher to requirement for dynamic balance, this forced vibration can influence the general level of the health of unit, from
And cause unit work efficiency drop.
6) equipment health status monitoring is with improving equipment operating efficiency
The device signal isolated according to signal separation system, after carrying out labor, by the equipment running status of analysis
Signal passes to equipment running status and assesses area, as shown in Figure 7.According to the status information assessed in area, on the one hand can make
Corresponding status tracking is suggested with maintenance, makes the timely discovering device unhealthy condition of staff, and take according to system recommendations
Rational maintenance measures.Another aspect can also make maintenance and remind and maintain guidance, reach equipment health forecast and management
Effect, periodically rational progress plant maintenance is instructed according to maintenance.In addition, by the equipment operating data analyzed, fault message
Data, algorithm data are stored in corresponding database in time, quickly and easily to be made in time to the running status of equipment
Reflection.According to above method, equipment running status can be adjusted optimally by we at any time, reduce equipment inferior health shape
State, so as to improve equipment operational energy efficiency.
It is understood that above with respect to the specific descriptions of the present invention, it is merely to illustrate the present invention and is not limited to this
Technical scheme described by inventive embodiments, it will be understood by those within the art that, still the present invention can be carried out
Modification or equivalent substitution, to reach identical technique effect;As long as meet use needs, all protection scope of the present invention it
It is interior.
Claims (4)
1. the healthy efficiency detection method of Air Conditioning Facilities device cluster, it is characterised in that comprise the following steps:
Step 1:First, the key position such as the compressor of air-conditioner set, shaft coupling arrangement vibration measuring point in computer room, uses vibration
Each part of spectrum analyzer measuring unit it is axial, vertically to level rumble spectrum, signal is finally passed into data and adopted
Collecting system;
Step 2:Then, the vibration signal stored gathering is as input signal X (t), with three rank accumulationsAnd quadravalence
AccumulationRepresent output signal y (t), edge entropy H (yi), and by three rank accumulationsWith quadravalence accumulationEdge entropy H (y are obtained with Edgeworth series expansionsi) display expression formula, using natural water surface coatings respectively to mutual trust
Cease the W in expression formula1、W2Derivation, thus calculate separation matrix W1And W2, finally in non-linear source signal blind separation model
On the basis of, isolate vibration source signal y (t) using the separation matrix solved;
Step 3:Finally, the source signal isolated is subjected to signature analysis, so that it is determined that the major failure source of computer room and basis point
From signal of vibrating feature recognition running status, and equipment running status information is fed back into healthy energy efficiency monitoring system in time,
So as to reach monitoring health in real time and the purpose of energy efficiency state.
2. the healthy efficiency detection method of Air Conditioning Facilities device cluster according to claim 1, it is characterised in that the step
The vibration measuring point arranged in 1 in equipment machine room is vibrating sensor.
3. the healthy efficiency detection method of Air Conditioning Facilities device cluster according to claim 1, it is characterised in that:The step
2 fall into a trap point counting from matrix W1、W2It is divided into following steps:
(1) determines vibrating sensor best position according to air-conditioner set key position, is carried out using data collecting system empty
Unit vibration data acquisition is adjusted, is adopted by the vibrating sensor for being arranged in the key positions such as compressor on air-conditioner set, shaft coupling
Collect the vibration signal X of air-conditioner setj(t), j=1,2 ..., n, wherein, j is channel number, and n is positive integer;
(2) air-conditioner set vibration signal Xs of the according to blind source separation algorithm of the invention to collectionj(t) blind source separating is carried out, so as to
Obtain original vibration signal source sj(t) approximate signal source yj(t), wherein, j=1,2 ..., n;
(3) is by yiEdge entropy with three rank accumulationsWith quadravalence accumulationRepresent and used Edgeworth
Series expansion:
<mrow>
<mi>H</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>y</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mn>2</mn>
</mfrac>
<mi>log</mi>
<mrow>
<mo>(</mo>
<mn>2</mn>
<mi>&pi;</mi>
<mi>e</mi>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mfrac>
<mn>1</mn>
<mn>12</mn>
</mfrac>
<msubsup>
<mi>C</mi>
<mn>3</mn>
<mn>2</mn>
</msubsup>
<msubsup>
<mi>y</mi>
<mi>i</mi>
<mn>3</mn>
</msubsup>
<mo>-</mo>
<mfrac>
<mn>1</mn>
<mn>48</mn>
</mfrac>
<msubsup>
<mi>C</mi>
<mn>4</mn>
<mn>2</mn>
</msubsup>
<msubsup>
<mi>y</mi>
<mi>i</mi>
<mn>4</mn>
</msubsup>
<mo>-</mo>
<mfrac>
<mn>1</mn>
<mn>48</mn>
</mfrac>
<msubsup>
<mi>C</mi>
<mn>3</mn>
<mn>1</mn>
</msubsup>
<msubsup>
<mi>y</mi>
<mi>i</mi>
<mn>3</mn>
</msubsup>
<mo>+</mo>
<mfrac>
<mn>1</mn>
<mn>8</mn>
</mfrac>
<msubsup>
<mi>C</mi>
<mn>3</mn>
<mn>2</mn>
</msubsup>
<msubsup>
<mi>y</mi>
<mi>i</mi>
<mn>3</mn>
</msubsup>
<msubsup>
<mi>C</mi>
<mn>4</mn>
<mn>1</mn>
</msubsup>
<msubsup>
<mi>y</mi>
<mi>i</mi>
<mn>4</mn>
</msubsup>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mi>a</mi>
<mo>)</mo>
</mrow>
<mo>,</mo>
</mrow>
WhereinRespectively yiThree ranks and quadravalence accumulation;
(4) output y is expressed as mutual information by according to the definition of mutual information in information theory with separation parameter:
<mrow>
<mtable>
<mtr>
<mtd>
<mrow>
<mi>I</mi>
<mrow>
<mo>(</mo>
<mi>y</mi>
<mo>,</mo>
<msub>
<mi>W</mi>
<mn>1</mn>
</msub>
<mo>,</mo>
<msub>
<mi>W</mi>
<mn>2</mn>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mo>-</mo>
<mi>H</mi>
<mrow>
<mo>(</mo>
<mi>y</mi>
<mo>)</mo>
</mrow>
<mo>+</mo>
<munderover>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</munderover>
<mi>H</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>y</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mo>=</mo>
<mi>H</mi>
<mrow>
<mo>(</mo>
<mi>s</mi>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mi>log</mi>
<mo>|</mo>
<msub>
<mi>W</mi>
<mn>1</mn>
</msub>
<mo>|</mo>
<mo>+</mo>
<msubsup>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</msubsup>
<mi>H</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>y</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
</mtr>
</mtable>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mi>b</mi>
<mo>)</mo>
</mrow>
<mo>;</mo>
</mrow>
(5) it is nonlinear activation function that, which defines phasor function f (y), and calculates phasor function according to sigmoid functions:
<mrow>
<mi>f</mi>
<mrow>
<mo>(</mo>
<mi>y</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<msup>
<mrow>
<mo>&lsqb;</mo>
<mo>-</mo>
<mfrac>
<mrow>
<msubsup>
<mi>P</mi>
<mrow>
<mi>y</mi>
<mn>1</mn>
</mrow>
<mo>&prime;</mo>
</msubsup>
<mrow>
<mo>(</mo>
<msub>
<mi>y</mi>
<mn>1</mn>
</msub>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<msub>
<mi>P</mi>
<mrow>
<mi>y</mi>
<mn>1</mn>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<msub>
<mi>y</mi>
<mn>1</mn>
</msub>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
<mo>,</mo>
<mo>-</mo>
<mfrac>
<mrow>
<msubsup>
<mi>P</mi>
<mrow>
<mi>y</mi>
<mn>2</mn>
</mrow>
<mo>&prime;</mo>
</msubsup>
<mrow>
<mo>(</mo>
<msub>
<mi>y</mi>
<mn>2</mn>
</msub>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<msub>
<mi>P</mi>
<mrow>
<mi>y</mi>
<mn>2</mn>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<msub>
<mi>y</mi>
<mn>2</mn>
</msub>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
<mo>,</mo>
<mn>...</mn>
<mo>,</mo>
<mo>-</mo>
<mfrac>
<mrow>
<msubsup>
<mi>P</mi>
<mrow>
<mi>y</mi>
<mi>n</mi>
</mrow>
<mo>&prime;</mo>
</msubsup>
<mrow>
<mo>(</mo>
<msub>
<mi>y</mi>
<mi>n</mi>
</msub>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<msub>
<mi>P</mi>
<mrow>
<mi>y</mi>
<mi>n</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<msub>
<mi>y</mi>
<mi>n</mi>
</msub>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
<mo>&rsqb;</mo>
</mrow>
<mi>T</mi>
</msup>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mi>c</mi>
<mo>)</mo>
</mrow>
<mo>;</mo>
</mrow>
(6) utilizes the gradient descent algorithm of Minimum mutual information, respectively to W in formula (b)1And W2Derivation:
<mrow>
<mfrac>
<mrow>
<mi>d</mi>
<mi>I</mi>
<mrow>
<mo>(</mo>
<mi>y</mi>
<mo>,</mo>
<msub>
<mi>W</mi>
<mn>1</mn>
</msub>
<mo>,</mo>
<msub>
<mi>W</mi>
<mn>2</mn>
</msub>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<msub>
<mi>dW</mi>
<mn>1</mn>
</msub>
</mrow>
</mfrac>
<mo>=</mo>
<mi>&eta;</mi>
<mrow>
<mo>(</mo>
<mi>I</mi>
<mo>-</mo>
<mi>f</mi>
<mo>(</mo>
<mi>y</mi>
<mo>)</mo>
<msup>
<mi>y</mi>
<mi>T</mi>
</msup>
<mo>)</mo>
</mrow>
<msub>
<mi>W</mi>
<mn>1</mn>
</msub>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mi>d</mi>
<mo>)</mo>
</mrow>
<mo>,</mo>
</mrow>
<mrow>
<mfrac>
<mrow>
<mi>d</mi>
<mi>I</mi>
<mrow>
<mo>(</mo>
<mi>y</mi>
<mo>,</mo>
<msub>
<mi>W</mi>
<mn>1</mn>
</msub>
<mo>,</mo>
<msub>
<mi>W</mi>
<mn>2</mn>
</msub>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<msub>
<mi>dW</mi>
<mn>2</mn>
</msub>
</mrow>
</mfrac>
<mo>=</mo>
<mi>&eta;</mi>
<mrow>
<mo>(</mo>
<mi>I</mi>
<mo>-</mo>
<mi>f</mi>
<mo>(</mo>
<mi>y</mi>
<mo>)</mo>
<msup>
<mi>y</mi>
<mi>T</mi>
</msup>
<mo>-</mo>
<mi>D</mi>
<mo>(</mo>
<mi>u</mi>
<mo>)</mo>
<msubsup>
<mi>W</mi>
<mn>1</mn>
<mi>T</mi>
</msubsup>
<mi>f</mi>
<mo>(</mo>
<mi>y</mi>
<mo>)</mo>
<msup>
<mi>y</mi>
<mi>T</mi>
</msup>
<mo>)</mo>
</mrow>
<msub>
<mi>W</mi>
<mn>2</mn>
</msub>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mi>e</mi>
<mo>)</mo>
</mrow>
<mo>,</mo>
</mrow>
Wherein D (u)=diag [P 'y1(y1) ..., P 'yn(yn)], diag [] is that vector is turned into diagonal matrix, diagonal element
For the respective components of vector, η is Learning Step;
(7) finally brings formula (c) in formula (d) and (e) into, solves separation matrix W1And W2:
W1(k+1)=W1(k)+η (I-f (y) yT)W1(k) (f),
W2(k+1)=W2(k)+η (I-f (y) yT- W1(k)T f(y)yT)W2(k) (g),
Wherein η is adaptive learning step-length;
(8) solves source signal y (t) using separation matrix separation collection signal X (t);
(9) draws;
(10) carries out signature analysis according to source signal y (t) is isolated;
(11) terminates.
4. the healthy efficiency detection method of Air Conditioning Facilities device cluster according to claim 1, it is characterised in that:The step
Vibration spectrum analyzer in 1 uses CSI2310 vibration spectrum analyzers.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710866822.1A CN107607342B (en) | 2017-09-22 | 2017-09-22 | Healthy energy efficiency detection method for air conditioner room equipment group |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710866822.1A CN107607342B (en) | 2017-09-22 | 2017-09-22 | Healthy energy efficiency detection method for air conditioner room equipment group |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107607342A true CN107607342A (en) | 2018-01-19 |
CN107607342B CN107607342B (en) | 2020-05-26 |
Family
ID=61061807
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710866822.1A Active CN107607342B (en) | 2017-09-22 | 2017-09-22 | Healthy energy efficiency detection method for air conditioner room equipment group |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107607342B (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110146317A (en) * | 2019-04-29 | 2019-08-20 | 北京和欣运达科技有限公司 | A kind of diagnostic method and device of building electromechanical equipment health status |
CN112432317A (en) * | 2020-11-16 | 2021-03-02 | 东南大学 | Sensor optimal arrangement method for classroom and ventilation monitoring system thereof |
CN113884236A (en) * | 2021-08-24 | 2022-01-04 | 西安电子科技大学 | Multi-sensor fusion dynamic balance analysis method, system, equipment and medium |
CN116400627A (en) * | 2023-04-13 | 2023-07-07 | 深圳市丰源升科技有限公司 | Industrial remote control system and method based on 5G |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2003078423A (en) * | 2001-09-03 | 2003-03-14 | Clarion Co Ltd | Processor for separating blind signal |
CN1914683A (en) * | 2004-02-26 | 2007-02-14 | 南承铉 | Methods and apparatus for blind separation of multichannel convolutive mixtures in the frequency domain |
CN104155134A (en) * | 2014-08-06 | 2014-11-19 | 北京信息科技大学 | Judgment method of applicability of high-order cumulant feature extraction method |
CN104390780A (en) * | 2014-11-25 | 2015-03-04 | 沈阳化工大学 | Gear case fault diagnosis method based on blind source separation |
-
2017
- 2017-09-22 CN CN201710866822.1A patent/CN107607342B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2003078423A (en) * | 2001-09-03 | 2003-03-14 | Clarion Co Ltd | Processor for separating blind signal |
CN1914683A (en) * | 2004-02-26 | 2007-02-14 | 南承铉 | Methods and apparatus for blind separation of multichannel convolutive mixtures in the frequency domain |
CN104155134A (en) * | 2014-08-06 | 2014-11-19 | 北京信息科技大学 | Judgment method of applicability of high-order cumulant feature extraction method |
CN104390780A (en) * | 2014-11-25 | 2015-03-04 | 沈阳化工大学 | Gear case fault diagnosis method based on blind source separation |
Non-Patent Citations (1)
Title |
---|
周勃: "基于盲分离的空调机组故障振声诊断研究", 《中国博士学位论文全文数据库 工程科技Ⅱ辑(月刊)》 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110146317A (en) * | 2019-04-29 | 2019-08-20 | 北京和欣运达科技有限公司 | A kind of diagnostic method and device of building electromechanical equipment health status |
CN112432317A (en) * | 2020-11-16 | 2021-03-02 | 东南大学 | Sensor optimal arrangement method for classroom and ventilation monitoring system thereof |
CN113884236A (en) * | 2021-08-24 | 2022-01-04 | 西安电子科技大学 | Multi-sensor fusion dynamic balance analysis method, system, equipment and medium |
CN113884236B (en) * | 2021-08-24 | 2022-06-21 | 西安电子科技大学 | Multi-sensor fusion dynamic balance analysis method, system, equipment and medium |
CN116400627A (en) * | 2023-04-13 | 2023-07-07 | 深圳市丰源升科技有限公司 | Industrial remote control system and method based on 5G |
Also Published As
Publication number | Publication date |
---|---|
CN107607342B (en) | 2020-05-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104614179B (en) | A kind of gearbox of wind turbine state monitoring method | |
CN107607342A (en) | The healthy efficiency detection method of Air Conditioning Facilities device cluster | |
CN102944416B (en) | Multi-sensor signal fusion technology-based fault diagnosis method for wind turbine blades | |
Xu et al. | Predicting fan blade icing by using particle swarm optimization and support vector machine algorithm | |
CN109858140B (en) | Fault diagnosis method for water chilling unit based on information entropy discrete Bayesian network | |
CN111459700A (en) | Method and apparatus for diagnosing device failure, diagnostic device, and storage medium | |
US8751423B2 (en) | Turbine performance diagnostic system and methods | |
CN105003453A (en) | Online monitoring and fault diagnosis system of mine fan | |
CN111946559B (en) | Method for detecting structures of wind turbine foundation and tower | |
CN110311709B (en) | Fault judgment method for electricity consumption information acquisition system | |
CN103631681A (en) | Method for online restoring abnormal data of wind power plant | |
CN105759201A (en) | High voltage circuit breaker self-diagnosis method based on abnormal sample identification | |
GB2476246A (en) | Diagnosing an operation mode of a machine | |
CN204113701U (en) | A kind of mine fan on-line monitoring and fault diagnosis system | |
KR20110116378A (en) | Data collecting method for detection and on-time warning system of industrial process | |
CN103822786A (en) | Wind turbine generator set mechanical equipment state diagnosis method based on multivariate statistical analysis | |
CN104343709A (en) | Draught fan failure detection apparatus and method | |
CN103970124A (en) | On-line detection method for industrial control loop multi-period oscillation | |
CN104005975B (en) | The diagnostic method of a kind of axial fan stall and surge | |
CN101738998A (en) | System and method for monitoring industrial process based on local discriminatory analysis | |
CN106441843A (en) | Rotating machinery fault waveform recognition method | |
CN104569814B (en) | A kind of DC traction motor health status real-time analysis method based on approximate entropy | |
CN106126949A (en) | A kind of steam turbine generator running status appraisal procedure | |
CN102778632A (en) | Double normalization recognition method for directly forecasting and recognizing transformer winding fault type | |
US11339763B2 (en) | Method for windmill farm monitoring |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |