CN102466566B - Power device anomalous detection device and detection method thereof - Google Patents

Power device anomalous detection device and detection method thereof Download PDF

Info

Publication number
CN102466566B
CN102466566B CN201110003851.8A CN201110003851A CN102466566B CN 102466566 B CN102466566 B CN 102466566B CN 201110003851 A CN201110003851 A CN 201110003851A CN 102466566 B CN102466566 B CN 102466566B
Authority
CN
China
Prior art keywords
power
equipment
running status
factor
crank pulses
Prior art date
Application number
CN201110003851.8A
Other languages
Chinese (zh)
Other versions
CN102466566A (en
Inventor
钟欣仪
钟欣兰
朱亿龙
曾仕民
Original Assignee
财团法人工业技术研究院
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Priority to TW99137868A priority Critical patent/TWI426242B/en
Priority to TW099137868 priority
Application filed by 财团法人工业技术研究院 filed Critical 财团法人工业技术研究院
Publication of CN102466566A publication Critical patent/CN102466566A/en
Application granted granted Critical
Publication of CN102466566B publication Critical patent/CN102466566B/en

Links

Abstract

The invention provides a power device anomalous detection device and a detection method thereof. The detection device includes a sensing module, a processing module, an optimizing module and a classifying diagnosis module. The sensing module is used for sensing a power device to obtain a plurality of operation signals. The processing module obtains a plurality of character values from the operation signals. The optimizing module establishes a plurality of factor groups by classifying the character values. The classifying diagnosis module connects with the optimizing module and sends a state signal according to a predetermined rule and the established factor groups. The detection method includes steps of using a signal processing method to obtain the character values from the power device; establishing the factor groups according to relevance of the character values; and obtaining the operation state of the power device through a neural network and an experience rule, and judging whether the operation state is normal or not according to the predetermined rule.

Description

Power device anomalous detection device and detection method thereof

Technical field

The present invention relates to a kind of monitoring and diagnosis for power-equipment operation, particularly a kind of power device anomalous detection device and detection method thereof.

Background technology

In general; power-equipment is before fault occurs; usually there will be the decline of performance and the increase of power consumption; but these phenomenons can't have influence on the running of equipment immediately; therefore often can not found by user; this result increases fault Probability and shortens equipment life except meeting, and power-equipment usefulness reduces the increase with power consumption, for industrial competitiveness and environmental protection, has negative impact.

The diagnostic mode of power-equipment (for example motor) is to utilize a sensor to carry out the state of sensing power-equipment operation mostly, then by wired or wireless transmission mode, sensed data is sent to back-end system and is further analyzed.But this kind of pattern need to be by a large amount of data transmission frequency ranges, main cause is in order to ensure the validity of analyzing and diagnosing and stability, as much as possible by complete being sent in back-end system of the received running state data of all sensors.

Therefore, how can be by a kind of method or means, except possessing equally accurate analyzing and diagnosing ability, first the running state data of sensing is processed, to reduce the data volume that need to be sent to back-end system, and then reach effective reduction data transmission frequency range size, promote data transmission stability, shorten diagnosis update time and reduce the effects such as setup cost, be the target that relevant manufacturer makes great efforts for a long time always.

Summary of the invention

In view of above problem, the invention provides a kind of power device anomalous detection device and detection method thereof.By the running state data of sensing is processed in advance, to reach effective reduction data transmission frequency range size, promote data transmission stability, shorten diagnosis update time and to reduce setup cost.

The disclosed power device anomalous detection device according to the present invention, comprises a sensing module, a processing module, an optimal treatment module and a classification diagnosis module.Sensing module in order to sensing one power-equipment to obtain a plurality of CRANK PULSES.Processing module is connected in this sensing module, to receive described CRANK PULSES, and sequentially from this CRANK PULSES respectively, obtains a plurality of eigenwerts.

Optimal treatment module connects this processing module, and to receive described eigenwert, and the described eigenwert of classifying is set up a plurality of factor group.Wherein, optimal treatment module can be utilized factor approach, and eigenwert is sorted out to a plurality of factor group according to relevance.Ge Gai factor group has a variation features value that represents this factor group, and the quantity of described variation features value is less than the quantity of described eigenwert.

Classification diagnosis module connects this optimal treatment module in order to receive described factor group, and sends a status signal according to a preset rules and described factor group.This preset rules can be a tabulation, and tabulation comprises normal project and an anomaly item.Normal condition when normal project is power-equipment running, anomaly item is situation abnormal while representing running, for example, anomaly item can include but not limited to uneven situation, not to mood condition, lubrication circumstances, resonance situation, bearing damage situation, bending shaft situation, loosening situation, unbalance in phase situation, the uneven situation of current potential, harmonic wave frequency multiplication situation and short-circuit conditions.

Therefore, by above-mentioned power device anomalous detection device, pick-up unit can be arranged on a power-equipment, by optimal treatment module, utilize factor approach that the CRANK PULSES of institute's sensing on power-equipment is simplified, directly by classification diagnosis module, carry out the judgement of power-equipment ruuning situation, the CRANK PULSES of institute's sensing need not be sent to back-end system and carry out direct and instant processing, to reach, shorten diagnosis update time and reduce setup cost.Moreover, even if still need back-end system future, process (for example: by the running status of the whole a plurality of power-equipments of a far-end server remittance), the quantity of the variation features value of summarizing through factor approach is lower than the quantity from the obtained eigenwert of this CRANK PULSES respectively, therefore can reach the effect that reduces data transmission frequency range size and promote data transmission stability.

The disclosed power-equipment method for detecting abnormality according to the present invention, carries out detection and the diagnosis of anomaly parameter by the information of detecting power-equipment running.Power-equipment method for detecting abnormality, first utilizes a signal processing method to obtain a plurality of CRANK PULSES from this power-equipment, and in CRANK PULSES, captures a plurality of eigenwerts.Then, then eigenwert is classified to set up a plurality of factor group according to relevance, Er Gegai factor group has a variation features value.The last factor group that more obtained variation features value is greater than to 1, utilizes neural network and thumb rule to obtain the running status of this power-equipment, and judges that according to a preset rules whether the running status of power-equipment is abnormal.

Wherein, when the running status that judges when neural network and thumb rule is inconsistent, according to described factor group, revise the model of neural network, until the running status result that both judge is consistent.

CRANK PULSES can be vibration signal, temperature signal, magnetic flux signal, current signal or the voltage signal of power-equipment.Processing module system by the CRANK PULSES of institute's sensing by a time domain conversion process or a multi-scale entropy (Multiscale Entropy, MSE) computing to obtain eigenwert, eigenwert can be frequency multiplication peak value or the characteristic frequency value of vibration signal.Time domain conversion process can adopt a discrete Fourier conversion process (Discrete Fourier Transform, DFT), a fast fourier transform is processed (Fast Fourier Transform, FFT), a discrete cosine transform is processed (Discrete Cosine Transformation, DCT), a discrete hartley conversion process (Discrete Hartley Transform, DHT), a wavelet conversion is processed (Wavelet Transform, WT) or power-frequency processing (Power Spectrum).

In order to judge that the whether abnormal preset rules of running status of power-equipment can be a tabulation, tabulation comprises normal project and an anomaly item.Normal condition when normal project is power-equipment running, anomaly item is situation abnormal while representing running, for example, anomaly item can include but not limited to uneven situation, not to mood condition, lubrication circumstances, resonance situation, bearing damage situation, bending shaft situation, loosening situation, unbalance in phase situation, the uneven situation of current potential, harmonic wave frequency multiplication situation and short-circuit conditions.

In this, when obtaining by neural network after the running status of power-equipment, can judge that according to above-mentioned preset rules this power-equipment may be that abnormal conditions partly occur for which.Neural network can adopt a back propagation neural network (Back Propagation Network, BPN), a Hopfield Network (Hopfield Neural Network, HNN), neural network (Radial Basis Function Network at the bottom of a radial basis, RBFN), a fuzzy nerve network (Fuzzy Neural Network, FNN) or one function link neural network (Functional Link Neural Network, FLNN).Thumb rule is a characteristic frequency spectrum, a critical threshold, a trajectory diagram, an envelope, a frequency analysis or its combination.

Therefore, by power-equipment method for detecting abnormality of the present invention, can pass through the CRANK PULSES of sensing one power-equipment, and simplify quantity and the size from the obtained eigenwert of CRANK PULSES by factor approach, in the less microprocessor of resource consumption, directly carry out the running status that calculation process judges power-equipment, sensed CRANK PULSES need not be sent to back-end system, back-end system only needs to receive judged result, can reach and reduce data transmission frequency range size, promote data transmission stability, shorten the effect of diagnosis update time and reduction setup cost.

About feature of the present invention, enforcement and effect, hereby coordinate graphicly as embodiment, to be described in detail as follows.

Accompanying drawing explanation

Fig. 1 is the schematic diagram of power device anomalous detection device of the present invention.

Fig. 2 A is the schematic diagram of power device anomalous detection device one embodiment of the present invention.

Fig. 2 B is the schematic diagram of another embodiment of power device anomalous detection device of the present invention.

Fig. 2 C is the schematic diagram of the another embodiment of power device anomalous detection device of the present invention.

Fig. 3 is the flow chart of steps of power-equipment method for detecting abnormality of the present invention.

Fig. 4 A is an embodiment process flow diagram of step S310 in Fig. 3.

Fig. 4 B is another embodiment process flow diagram of step S310 in Fig. 3.

Fig. 4 C is the flow chart of steps of step S320 in Fig. 3.

Fig. 5 is the process flow diagram of an embodiment of step S340 in Fig. 3.

Main element symbol description

100 power device anomalous detection devices

110 sensing modules

120 processing modules

130 optimal treatment modules

140 classification diagnosis modules

150 alarming devices

160 transport modules

170 memory modules

200 power-equipments.

Embodiment

Please refer to shown in Fig. 1, Fig. 1 is the schematic diagram of power device anomalous detection device of the present invention.Power device anomalous detection device 100 can be an embedded system wafer or a personal digital assistant device (Personal Digital Assistant, the data processing equipment such as PDA), power device anomalous detection device 100 is arranged at a power-equipment 200 and comprises a sensing module 110, a processing module 120, an optimal treatment module 130 and a classification diagnosis module 140.

Sensing module 110 in order to sensing power-equipment 100 to obtain a plurality of CRANK PULSES, take motor as example, sensing module 110 is except by the vibration signal data of extent of vibration sensing power-equipment 100, the CRANK PULSES while also obtaining the motor operations such as temperature, magnetic flux, electric current, rotating speed, voltage.

The CRANK PULSES that processing module 120 senses according to sensing module 110, utilizes a signal processing method in CRANK PULSES, to capture a plurality of eigenwerts.With vibration signal, processing module 120 utilizes fast fourier transform disposal route from time domain, to convert CRANK PULSES to frequency domain, and by rotating speed, converse the fundamental frequency of vibration signal, sequentially from the vibration signal of frequency domain, capture 0.5 frequency multiplication, 1 frequency multiplication to 12 frequency multiplication again, the frequency-doubled signal capturing is the eigenwert of corresponding vibration signal.

Optimal treatment module 130 connects these processing modules 120, and to receive described eigenwert, and the described eigenwert of classifying is set up a plurality of factor group.Wherein, optimal treatment module 130 can be utilized factor approach, and eigenwert is sorted out to a plurality of factor group according to relevance.Ge Gai factor group has a variation features value that represents this factor group, and the quantity of described variation features value is less than the quantity of described eigenwert.

Classification diagnosis module 140 connects these optimal treatment modules 130 in order to receive described factor group, and sends a status signal according to a preset rules and described factor group, to judge the ruuning situation of power-equipment.This preset rules can be a tabulation, and tabulation comprises normal project and an anomaly item.Normal condition when normal project is power-equipment running, anomaly item is situation abnormal while representing running, for example, anomaly item can include but not limited to uneven situation, not to mood condition, lubrication circumstances, resonance situation, bearing damage situation, bending shaft situation, loosening situation, unbalance in phase situation, the uneven situation of current potential, harmonic wave frequency multiplication situation and short-circuit conditions.

Please refer to shown in Fig. 2 A, Fig. 2 A is the schematic diagram of power device anomalous detection device one embodiment of the present invention.Power device anomalous detection device more comprises an alarming device 150, and this alarming device 150 is in order to receive this status signal, and when this status signal be that this is when abnormal, in order to notify user.Alarming device 150 can be but is not limited to a vibration module, a light emitting module, a display module, a sound module or its combination, abnormal to notify the operation of user's power-equipment 200 to occur by modes such as vibration warning, light warning, information warning or sound warnings.

Please refer to shown in Fig. 2 B, Fig. 2 B is the schematic diagram of another embodiment of power device anomalous detection device of the present invention.Power device anomalous detection device also comprises a transport module 160, and transport module 160 connects this classification diagnosis module 140, in order to receive this status signal, and by wired or wireless transmission mode, this status signal is sent to alarming device 150.

Please refer to shown in Fig. 2 C, Fig. 2 C is the schematic diagram of the another embodiment of power device anomalous detection device of the present invention.Power device anomalous detection device also comprises a memory module 170, memory module 170 is in order to store the described CRANK PULSES of this power-equipment 200, when user has the CRANK PULSES that need to read power-equipment, further analyze, then obtain required CRANK PULSES by access memory module 170.Memory module 170 can be for storage card being set to store the CRANK PULSES of sensing, storage card can be a compact flash (Compact Flash, CF) storage card, a micro hard disk (Micro Drive, MD) storage card, a secure digital (Secure Digital, SD) storage card, a miniature secure digital (Micro SD) storage card, a multimedia (Multi Media Card, MMC) storage card, rectangular (Memory Stick, a MS) storage card or one miniature rectangular (Micro MS) storage card.

Therefore, by above-mentioned power device anomalous detection device, pick-up unit can be arranged on a power-equipment, by optimal treatment module, utilize factor approach that the CRANK PULSES of institute's sensing on ultromotivity equipment is simplified, directly by classification diagnosis module, carry out the judgement of power-equipment ruuning situation, the CRANK PULSES of institute's sensing need not be sent to back-end system and carry out direct and instant processing, to reach, shorten diagnosis update time and reduce setup cost.Moreover, even if still need back-end system future, process (for example: by the running status of the whole a plurality of power-equipments of a far-end server remittance), the quantity of the variation features value of summarizing through factor approach is lower than the quantity from the obtained eigenwert of this CRANK PULSES respectively, therefore can reach the effect that reduces data transmission frequency range size and promote data transmission stability.

Please refer to shown in Fig. 3, Fig. 3 is the flow chart of steps of power-equipment method for detecting abnormality of the present invention.Power-equipment method for detecting abnormality is applied to a power-equipment, comprising:

Step S300: utilize a signal processing method to obtain a plurality of CRANK PULSES from this power-equipment;

Step S310: obtain corresponding to a plurality of eigenwerts of this CRANK PULSES respectively from described CRANK PULSES;

Step S320: described eigenwert is divided into groups, and to set up a plurality of factor group, Ge Gai factor group has a variation features value;

Step S330: according to described factor group, utilize a neural network to judge one first running status of this power-equipment running;

Step S340: according to described eigenwert, utilize a thumb rule to judge one second running status of this propulsion system running;

Step S350: relatively whether this first running status is identical with this second running status;

Step S360: when this first running status is not identical with this second running status, revise this neural network according to described factor group, until this first running status is identical with this second running status;

Step S370: when this first running status is identical with this second running status, judge that according to a preset rules whether this first running status is abnormal;

Step S380: this first running status, for abnormal, sends an abnormal signal if judge; And

Step S390: this first running status, for normal, sends a normal signal if judge.

Please refer to shown in Fig. 4 A, Fig. 4 A is an embodiment process flow diagram of step S310 in Fig. 3.Obtaining corresponding to a plurality of eigenwerts of this CRANK PULSES respectively from described CRANK PULSES described in step S310, CRANK PULSES involving vibrations signal, temperature signal, magnetic flux signal, current signal or voltage signal.Step S310 comprises:

Step S311: this power-equipment of sensing, to obtain described CRANK PULSES;

Step S312: utilize a time domain conversion process, a time domain data of this CRANK PULSES is converted to a frequency domain data; And

Step S313: this frequency domain data captures a plurality of eigenwerts certainly.

Wherein, time domain conversion process can be a discrete Fourier conversion process, a fast fourier transform processing, a discrete cosine transform processing, a discrete hartley conversion process, a wavelet conversion is processed or a power-frequency is processed.

With the vibration signal of motor, when vibration signal, through fast fourier transform, to process, fundamental frequency (harmonic wave) can calculate by following formula:

The first fundamental frequency position=((data length of 1 * rotating speed * CRANK PULSES)/60) * frequency spectrum acquisition frequency;

The second fundamental frequency position=((data length of 2 * rotating speed * CRANK PULSES)/60) * frequency spectrum acquisition frequency; By that analogy.

For example: suppose the motor that sensing one per minute 1800 turns, obtain the CRANK PULSES of one 16 kilobit tuples, get the frequency domain that frequency is 12 KHz, the position of the first fundamental frequency can be 40.

Accordingly, when step S311 obtains one while representing the CRANK PULSES of motor vibrations signal, by step, S312 is converted to frequency domain data by CRANK PULSES from time domain data, sequentially from the CRANK PULSES of frequency domain, capture 0.5 frequency multiplication again, the value of 1 frequency multiplication to 12 frequency multiplication position, the frequency-doubled signal capturing is the eigenwert of corresponding vibration signal, the quantity of these eigenwerts is 24, and be defined as respectively 0.5x according to the size of frequency multiplication, 1x, 1.5x, 2x, 2.5x, 3x, 3.5x, 4x, 4.5x, 5x, 5.5x, 6x, 6.5x, 7x, 7.5x, 8x, 8.5x, 9x, 9.5x, 10x, 10.5x, 11x, 11.5x and 12x.

Please refer to shown in Fig. 4 B, Fig. 4 B is another embodiment process flow diagram of step S310 in Fig. 3.Compare the embodiment of the step S310 shown in Fig. 4 A, the step S310 embodiment shown in Fig. 4 B obtains the eigenwert of CRANK PULSES by multi-scale entropy (Multiscale Entropy, MSE) computing, and this embodiment comprises:

Step S314: this power-equipment of sensing, to obtain described CRANK PULSES; And

Step S315: will remove described CRANK PULSES after noise by a multi-scale entropy (Multiscale Entropy, MSE) computing, to obtain described eigenwert that should CRANK PULSES.

Between step S314 and S315, also can comprise a step S316, step S316: utilize wavelet conversion to carry out noise processed to described CRANK PULSES.Main cause is may have noise through the obtained CRANK PULSES of sensing, by wavelet conversion, is processed and can be reached the effect that suppresses noise.

Please refer to shown in Fig. 3 and Fig. 4 C, Fig. 4 C is the flow chart of steps of step S320 in Fig. 3.Wait to obtain described a plurality of eigenwerts (step S310) corresponding to CRANK PULSES, power-equipment method for detecting abnormality is simplified these eigenwerts by factor approach again, as described in step S320 by as described in eigenwert divide into groups, to set up a plurality of factor group, Ge Gai factor group has a variation features value, and step S320 comprises:

Step S321: utilize factor approach to hive off to described eigenwert, to set up described factor group;

Step S322: sequentially calculate eigenwert described in Ge Gai factor group, to obtain a variation features value; And

Step S323: retain the variation features value that described variation features value is greater than 1.

By above-mentioned step S321, to step S323, and the eigenwert of the obtained corresponding vibration signal of above-mentioned steps S310 of arranging in pairs or groups is parameter, can obtain the result of lower list one.

Table one

The table one of take is the variation features value that how to obtain each factor group as example carrys out description of step S322.First, sample co-variation matrix number S calculates by 24 frequency multiplication eigenwerts (0.5x, 1x, 1.5x, 2x, 2.5x, 3x, 3.5x, 4x, 4.5x, 5x, 5.5x, 6x, 6.5x, 7x, 7.5x, 8x, 8.5x, 9x, 9.5x, 10x, 10.5x, 11x, 11.5x and 12x) in system, and co-variation matrix number is expressed as follows:

Wherein var represents variance, and cov represents co-variation heteromerism.

Then, then from co-variation matrix number, calculate 24 variation features value λ 1..., λ 24, be respectively equation

Solution.So can solve λ 1=7.431, λ 2=3.257, λ 3=1.258, λ 4=1.206, λ 5=1.124 ..., λ 24=0.029, then according to result of calculation, variation features value is greater than to 1 as the principle of choosing factor group number, take table one as example, choose altogether Wu Ge factor group as the input variable of neural network.

And by factor analysis to eigenwert hive off (step S321), owing to having its load value under each frequency multiplication eigenwert Wu Ge factor group, by being chosen in maximum load value under its a certain factor group, come the factor group under representation feature value.The frequency multiplication eigenwert of 1X of take is example, load in factor one is-0.260, load in factor two is 0.899, load in factor three is-0.038, load in factor four is-0.092, load in factor five is-0.015, wherein maximum in the middle of the load value Shi Wuge factor group under factor two with 1X again, that is 1X is under the jurisdiction of in the factor group of factor two.Therefore, by identical mode, remaining 23 eigenwert can be categorized in respectively in the middle of Wu Ge factor group according to peak load value, as listed in table one, therefore factor has comprised 1.5X, 4X, 4.5X, 5X, 5.5X, 6X, 6.5X, 7X, 8.5X, 12X once; Factor two comprises 1X, 2X, 2.5X, 3X, 3.5X, 7.5X, 8X, 9X; Factor three comprises 10X, 10.5X; Factor four comprises 11X, 11.5X; Factor five comprises 0.5X, 9.5X.

As described in step S330, by utilizing five factor combinations to bring neural network into, obtain the first running status respectively.About the setting up this and know for those skilled in the art of neural network, in this, do not tired out and stated.Neural network can adopt a back propagation neural network (Back Propagation Network, BPN), a Hopfield Network (Hopfield Neural Network, HNN), neural network (Radial Basis Function Network at the bottom of a radial basis, RBFN), a fuzzy nerve network (Fuzzy Neural Network, FNN) or one function link neural network (Functional Link Neural Network, FLNN).Thumb rule is a characteristic frequency spectrum, a critical threshold, a trajectory diagram, an envelope, a frequency analysis or its combination.

In like manner, described in step S340, eigenwert is utilized to the second running status of thumb rule judgement propulsion system.Thumb rule is the rule of deriving according to basic theory, take vibration signal as example, thumb rule is the vibration characteristics rule of deriving in mechanical vibration basis, modal the combination of each variation features value (that is characteristic frequency spectrum) collocation critical threshold in frequency spectrum is calculated, also can use the methods such as trajectory diagram, envelope further variation features value to be carried out to decoupling zero and constituent analysis, common frequency analysis also can assistance obtain side frequency data.

In this, the method that the thumb rule system that step S340 adopts utilizes eigenwert and threshold to set, can judge that when eigenwert surpasses threshold setting value the operation generation of power-equipment is abnormal.

For instance, suppose that the peak value of the frequency multiplication eigenwert of 1X and the frequency multiplication eigenwert of 2X is with 5 millimeters of (millimeter/second per second, mm/s) be threshold setting value, the peak value of the eigenwert of 3X take 2 millimeters per second be threshold setting value, as shown in Figure 5, Fig. 5 is the logical flow chart of an embodiment of the second running status that in Fig. 3, step S340 utilizes thumb rule judgement propulsion system.If it is normal that the frequency multiplication eigenwert of 1X is less than 5 judgement power-equipments, otherwise, in the frequency multiplication eigenwert of 1X, be greater than 5, and the frequency multiplication eigenwert of 2X and 3X is less than 5 and 2 simultaneously, judge that unbalanced situation occurs the running status of this power-equipment.

Suppose that by obtained the first running status of neural network second running status obtained with thumb rule when identical (step S340), preset rules can be a tabulation, tabulation comprises normal project and an anomaly item.Normal condition when normal project is power-equipment running, anomaly item is situation abnormal while representing running, for example, anomaly item can include but not limited to uneven situation, not to mood condition, lubrication circumstances, resonance situation, bearing damage situation, bending shaft situation, loosening situation, unbalance in phase situation, the uneven situation of current potential, harmonic wave frequency multiplication situation and short-circuit conditions.

Step S380 judges according to this preset rules whether the first running status is the ruuning situation of noting down in anomaly item, uses and judges whether power-equipment abnormal state occurs.Otherwise, if the first running status is the ruuning situation of noting down in normal project, judge that the running status of this power-equipment is normal.

Therefore, by power-equipment method for detecting abnormality of the present invention, can pass through the CRANK PULSES of sensing one power-equipment, and simplify quantity and the size from the obtained eigenwert of CRANK PULSES by factor approach, in the less microprocessor of resource consumption, directly carry out the running status that calculation process judges power-equipment, sensed CRANK PULSES need not be sent to back-end system, back-end system only needs to receive judged result, can reach and reduce data transmission frequency range size, promote data transmission stability, shorten the effect of diagnosis update time and reduction setup cost.

Claims (19)

1. a power device anomalous detection device, this device comprises:
One processing module, in order to receive from a plurality of CRANK PULSES of this power-equipment sensing and sequentially to capture a plurality of eigenwerts from this CRANK PULSES respectively;
One optimal treatment module, connect this processing module, this optimal treatment module receives described eigenwert, and the described eigenwert of classifying is to set up a plurality of factor group, wherein, Ge Gai factor group has a variation features value that represents this factor group, and the quantity of described variation features value is less than the quantity of described eigenwert; And
One classification diagnosis module, connects this optimal treatment module, in order to receive this factor group under described variation features value, and sends a status signal according to a preset rules and described factor group; Wherein this classification diagnosis module sends a status signal according to a preset rules with described factor group and comprises particularly: according to described factor group, utilize a neural network to judge one first running status that this power-equipment turns round; According to described eigenwert, utilize a thumb rule to judge one second running status of this propulsion system running; Relatively whether this first running status is identical with this second running status; When this first running status is not identical with this second running status, according to described factor group, revise this neural network, until this first running status is identical with this second running status; When this first running status is identical with this second running status, according to this preset rules, judge that whether this first running status is abnormal; If judge, this first running status, for abnormal, sends an abnormal signal; And if judge that this first running status, for normal, sends a normal signal.
2. power device anomalous detection device as claimed in claim 1, also comprises an alarming device, in order to receive this status signal, and when this status signal is while being abnormal, notifies this power-equipment operation to occur abnormal.
3. power device anomalous detection device as claimed in claim 2, also comprise a transport module, this transport module connects this classification diagnosis module, in order to receive this status signal and by wired or wireless transmission mode, this status signal to be sent to this alarming device.
4. power device anomalous detection device as claimed in claim 1, also comprises a memory module, the described CRANK PULSES of this this power-equipment of storage module stores.
5. a power-equipment method for detecting abnormality, is applied to a power-equipment, and this power-equipment method for detecting abnormality comprises:
Utilize a signal processing method to obtain a plurality of CRANK PULSES from this power-equipment;
From described CRANK PULSES, obtain corresponding to a plurality of eigenwerts of this CRANK PULSES respectively;
Described eigenwert is divided into groups, and to set up a plurality of factor group, Ge Gai factor group has a variation features value;
According to described factor group, utilize a neural network to judge one first running status of this power-equipment running;
According to described eigenwert, utilize a thumb rule to judge one second running status of this propulsion system running;
Relatively whether this first running status is identical with this second running status;
When this first running status is not identical with this second running status, according to described factor group, revise this neural network, until this first running status is identical with this second running status;
When this first running status is identical with this second running status, according to a preset rules, judge that whether this first running status is abnormal;
If judge, this first running status, for abnormal, sends an abnormal signal; And
If judge, this first running status, for normal, sends a normal signal.
6. power-equipment method for detecting abnormality as claimed in claim 5, wherein, this CRANK PULSES is a vibration signal.
7. power-equipment method for detecting abnormality as claimed in claim 5, wherein, this CRANK PULSES is a temperature signal.
8. power-equipment method for detecting abnormality as claimed in claim 5, wherein, this CRANK PULSES is a magnetic flux signal.
9. power-equipment method for detecting abnormality as claimed in claim 5, wherein, this CRANK PULSES is a current signal.
10. power-equipment method for detecting abnormality as claimed in claim 5, wherein, this CRANK PULSES is a voltage signal.
11. power-equipment method for detecting abnormality as claimed in claim 5, wherein, this CRANK PULSES is a tach signal.
12. power-equipment method for detecting abnormality as claimed in claim 5, wherein, this utilizes this signal processing method from this power-equipment, to obtain the step of described CRANK PULSES, comprising:
This power-equipment of sensing, to obtain described CRANK PULSES;
Utilize a time domain conversion process, a time domain data of this CRANK PULSES is converted to a frequency domain data; And
From this frequency domain data, capture described eigenwert.
13. power-equipment method for detecting abnormality as claimed in claim 12, wherein, this time domain conversion process is a discrete Fourier conversion process, a fast fourier transform processing, a discrete cosine transform processing, a discrete hartley conversion process, a wavelet conversion is processed or a power-frequency is processed.
14. power-equipment method for detecting abnormality as claimed in claim 5, wherein, this utilizes this signal processing method from this power-equipment, to obtain the step of described CRANK PULSES, comprising:
This power-equipment of sensing, to obtain described CRANK PULSES; And
The described CRANK PULSES of removing after noise is passed through to a multi-scale entropy computing, to obtain the described eigenwert of corresponding described CRANK PULSES.
15. power-equipment method for detecting abnormality as claimed in claim 14, wherein, in this power-equipment of this sensing, to obtain the step of described CRANK PULSES and this, will remove this CRANK PULSES after noise by this multi-scale entropy computing, to obtain between the step of described eigenwert that should CRANK PULSES, also comprise:
Utilize wavelet conversion to carry out noise processed to described CRANK PULSES.
16. power-equipment method for detecting abnormality as claimed in claim 5, wherein, this divides into groups described eigenwert, to set up the step of described factor group, comprises:
Utilize factor approach to hive off to described eigenwert, to set up described factor group;
Sequentially calculate eigenwert described in Ge Gai factor group, to obtain this variation features value; And
Retain the variation features value that described variation features value is greater than 1.
17. power-equipment method for detecting abnormality as claimed in claim 5, wherein, this neural network is neural network at the bottom of a back propagation neural network, a Hopfield Network, a radial basis, a fuzzy nerve network or a function link neural network.
18. power-equipment method for detecting abnormality as claimed in claim 5, wherein, this preset rules is a tabulation, this tabulation comprises normal project and an anomaly item, and this anomaly item comprises uneven situation, not to mood condition, lubrication circumstances, resonance situation, bearing damage situation, bending shaft situation, loosening situation, unbalance in phase situation, the uneven situation of current potential, harmonic wave frequency multiplication situation and short-circuit conditions.
19. power-equipment method for detecting abnormality as claimed in claim 5, wherein, this thumb rule is a characteristic frequency spectrum, a threshold setting value, a trajectory diagram, an envelope or a frequency analysis.
CN201110003851.8A 2010-11-03 2011-01-04 Power device anomalous detection device and detection method thereof CN102466566B (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
TW99137868A TWI426242B (en) 2010-11-03 2010-11-03 Diagnosing device and an associated method for a motor device
TW099137868 2010-11-03

Publications (2)

Publication Number Publication Date
CN102466566A CN102466566A (en) 2012-05-23
CN102466566B true CN102466566B (en) 2014-08-13

Family

ID=46070470

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201110003851.8A CN102466566B (en) 2010-11-03 2011-01-04 Power device anomalous detection device and detection method thereof

Country Status (2)

Country Link
CN (1) CN102466566B (en)
TW (1) TWI426242B (en)

Families Citing this family (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102707232B (en) * 2012-06-01 2015-10-07 深圳市海亿达能源科技股份有限公司 Motor apparatus state on_line monitoring device and monitoring method thereof
CN102768318A (en) * 2012-08-03 2012-11-07 深圳市海亿达能源科技股份有限公司 Online energy consumption management device for energy using equipment and management control method thereof
TWI443356B (en) * 2012-09-14 2014-07-01 Chunghwa Telecom Co Ltd Detection system and method of abnormal operating states in an electrical appliance
TWI467138B (en) * 2012-10-26 2015-01-01 Ancad Inc A vibration wave analysing device and method, the application thereof and the system using the same
CN104008294B (en) * 2014-05-30 2017-05-03 东南大学 System and method for detecting abnormality of bearing
US10336472B2 (en) * 2014-10-15 2019-07-02 The Boeing Company Motor health management apparatus and method
CN105974859A (en) * 2015-03-13 2016-09-28 青岛孚迪尔电气自动化有限公司 Wireless monitoring system based on vibration sensor
CN104833884A (en) * 2015-05-18 2015-08-12 国家电网公司 Fault detection method of voltage class equipment
CN106646205A (en) * 2015-10-30 2017-05-10 国网山西省电力公司电力科学研究院 Random big-disturbance signal removing algorithm for analyzing circuit breaker fault through sound and vibration combination
CN107121268A (en) * 2016-02-24 2017-09-01 王智中 Smart machine detection method
CN107991097A (en) * 2017-11-16 2018-05-04 西北工业大学 A kind of Method for Bearing Fault Diagnosis based on multiple dimensioned symbolic dynamics entropy
CN108871438A (en) * 2018-06-22 2018-11-23 武汉众犇慧通科技有限公司 A kind of motor monitoring, diagnosing method based on three shaft vibrations
CN108983097A (en) * 2018-07-27 2018-12-11 北京天诚同创电气有限公司 motor resonance detection system and detection method

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP0209862B1 (en) * 1985-07-19 1990-09-26 Hitachi, Ltd. Rotating machine diagnosis apparatus with acoustic emission technique
CN1112238A (en) * 1993-11-09 1995-11-22 新苏舍柴油机有限公司 Method for the recognition of disturbances in the operation of diesel engines
CN1758042A (en) * 2005-02-02 2006-04-12 沈阳黎明航空发动机(集团)有限责任公司 Engine bearing failure testing and diagnosing method and failure detecting instrument
CN101038239A (en) * 2007-02-02 2007-09-19 浙江大学 Device for detecting engine condition based on pure vibration signal and method thereof
CN101135601A (en) * 2007-10-18 2008-03-05 北京英华达电力电子工程科技有限公司 Rotating machinery vibrating failure diagnosis device and method
CN101354315A (en) * 2008-09-05 2009-01-28 爱立迈科(宁波)计测仪器有限公司 Device and method for tracking and detecting engine state based on vibration signal
CN101382468A (en) * 2008-09-05 2009-03-11 华南理工大学 Automobile engine failure diagnosis system and method based on sparse representation
CN101858778A (en) * 2010-05-28 2010-10-13 浙江大学 Vibration monitoring-based wind generator set automatic fault diagnosis method
KR20100112734A (en) * 2009-04-10 2010-10-20 한국전기연구원 On-site complex abnormal diagnosis method of induction motor

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
TWI386653B (en) * 2008-11-28 2013-02-21 Ind Tech Res Inst Method for diagnosing energy consumption of a power plant
CN101799359B (en) * 2010-01-27 2011-05-25 北京信息科技大学 Failure monitoring and predicting method and system of power equipment

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP0209862B1 (en) * 1985-07-19 1990-09-26 Hitachi, Ltd. Rotating machine diagnosis apparatus with acoustic emission technique
CN1112238A (en) * 1993-11-09 1995-11-22 新苏舍柴油机有限公司 Method for the recognition of disturbances in the operation of diesel engines
CN1758042A (en) * 2005-02-02 2006-04-12 沈阳黎明航空发动机(集团)有限责任公司 Engine bearing failure testing and diagnosing method and failure detecting instrument
CN101038239A (en) * 2007-02-02 2007-09-19 浙江大学 Device for detecting engine condition based on pure vibration signal and method thereof
CN101135601A (en) * 2007-10-18 2008-03-05 北京英华达电力电子工程科技有限公司 Rotating machinery vibrating failure diagnosis device and method
CN101354315A (en) * 2008-09-05 2009-01-28 爱立迈科(宁波)计测仪器有限公司 Device and method for tracking and detecting engine state based on vibration signal
CN101382468A (en) * 2008-09-05 2009-03-11 华南理工大学 Automobile engine failure diagnosis system and method based on sparse representation
KR20100112734A (en) * 2009-04-10 2010-10-20 한국전기연구원 On-site complex abnormal diagnosis method of induction motor
CN101858778A (en) * 2010-05-28 2010-10-13 浙江大学 Vibration monitoring-based wind generator set automatic fault diagnosis method

Also Published As

Publication number Publication date
TW201219756A (en) 2012-05-16
TWI426242B (en) 2014-02-11
CN102466566A (en) 2012-05-23

Similar Documents

Publication Publication Date Title
US10338555B2 (en) Methods and systems for the industrial internet of things
Li et al. A fault diagnosis scheme for planetary gearboxes using modified multi-scale symbolic dynamic entropy and mRMR feature selection
Gangsar et al. Comparative investigation of vibration and current monitoring for prediction of mechanical and electrical faults in induction motor based on multiclass-support vector machine algorithms
Wang et al. Prognostics and health management: A review of vibration based bearing and gear health indicators
Sohaib et al. A hybrid feature model and deep-learning-based bearing fault diagnosis
Li et al. Fault diagnosis of rotating machinery with a novel statistical feature extraction and evaluation method
Liu et al. Adaptive feature extraction using sparse coding for machinery fault diagnosis
Li et al. Deep learning-based remaining useful life estimation of bearings using multi-scale feature extraction
US10330647B2 (en) Automatic mechanical systems diagnosis
Pandya et al. Fault diagnosis of rolling element bearing with intrinsic mode function of acoustic emission data using APF-KNN
Yang et al. A fault diagnosis approach for roller bearing based on IMF envelope spectrum and SVM
Yan et al. Permutation entropy: a nonlinear statistical measure for status characterization of rotary machines
Zheng et al. A rolling bearing fault diagnosis method based on multi-scale fuzzy entropy and variable predictive model-based class discrimination
CA2875071C (en) Method and system for testing operational integrity of a drilling rig
CN102102629B (en) On-line data acquisition and analysis device of wind generating set
Figueiredo et al. Home electrical signal disaggregation for non-intrusive load monitoring (NILM) systems
Lin et al. Detection and classification of multiple power-quality disturbances with wavelet multiclass SVM
Zhao et al. Rolling bearing fault diagnosis based on wavelet packet decomposition and multi-scale permutation entropy
CA3072045A1 (en) Methods and systems for detection in an industrial internet of things data collection environment with large data sets
US10598520B2 (en) Method and apparatus for pneumatically conveying particulate material including a user-visible IoT-based classification and predictive maintenance system noting maintenance state as being acceptable, cautionary, or dangerous
Wang et al. A generic probabilistic framework for structural health prognostics and uncertainty management
Chen et al. Remaining life prognostics of rolling bearing based on relative features and multivariable support vector machine
Wu et al. Multi-scale analysis based ball bearing defect diagnostics using Mahalanobis distance and support vector machine
Safizadeh et al. Using multi-sensor data fusion for vibration fault diagnosis of rolling element bearings by accelerometer and load cell
Saidi et al. Application of higher order spectral features and support vector machines for bearing faults classification

Legal Events

Date Code Title Description
PB01 Publication
C06 Publication
SE01 Entry into force of request for substantive examination
C10 Entry into substantive examination
GR01 Patent grant
C14 Grant of patent or utility model