CN103940608A - Method for improving wind turbine generator gearbox failure level judgment precision - Google Patents

Abstract

The invention provides a method for improving wind turbine generator gearbox failure level judgment precision. The method comprises the steps that A, a failure type is determined according to the vibration signal characteristic value of a gearbox bearing; B, the distribution condition of all currently-monitored parameters corresponding to all monitored objects of a wind turbine generator gearbox in a failure level matrix of the failure type is determined, the failure level matrix comprises set failure levels, and each failure level comprises the value ranges of the parameters corresponding to the monitored objects of the wind turbine generator gearbox under the failure level; C, the occurrence probability of each failure level is calculated according to the distribution condition, and the current failure level is determined according to the occurrence probability. Therefore, schematic diagram convergence analysis can be carried out based on different weights of multiple kinds of failure signals, the blindness of locating the failure through a single failure parameter is avoided, the false alarm rate is lowered, and the wind turbine generator gearbox failure level judgment precision is improved.

Description

A kind of method that improves gearbox of wind turbine fault level and judge precision

Technical field

The present invention relates to wind-powered electricity generation unit monitoring technique field, particularly a kind of method that improves gearbox of wind turbine fault level and judge precision.

Background technology

As time goes on wind-powered electricity generation unit mechanical drive fault all can cause the sign of each side factor in succession, there is before this abnormal of vibration signal characteristics in for example appearance of a bearing fault, along with the rising of bearing temperature appears in the development of fault immediately, finally there is the damage completely of bearing, the appearance of gear distress is also first occur vibration signal characteristics abnormal, there is immediately the abnormal of gear case inner fluid, finally occur fractureing of gear tooth or even scrapping of whole gear case.

Traditional fault level determination methods only selects vibration amplitude as unique basis for estimation, by algorithm, fault level is judged, has obviously ignored the impact of other parameters on fault level, thereby causes inaccurate that fault level is judged.

Summary of the invention

The application provides a kind of method that improves gearbox of wind turbine fault level and judge precision, different weights based on various faults signal are carried out schematic diagram convergence analysis, avoid the blindness of single failure parameter location fault, reduce rate of false alarm, improved the precision of gearbox of wind turbine fault level judgement.

Described raising gearbox of wind turbine fault level judges that the method for precision comprises step:

A, determine fault type according to box bearing position vibration signal characteristics value;

B, the distribution situation of parameters in the fault level matrix of this fault type of determining current monitoring corresponding to each monitoring target of gearbox of wind turbine;

Described fault level matrix comprises each fault level of setting, and each fault level comprises parameters span corresponding to each monitoring target of gearbox of wind turbine under this fault level;

C, calculate according to described distribution situation the probability that each fault level occurs, and determine current described fault level accordingly.

By upper, the different weights that just can realize based on various faults signal are carried out schematic diagram convergence analysis, have avoided the blindness of single failure parameter location fault, have reduced rate of false alarm, improve the precision of gearbox of wind turbine fault level judgement.

Optionally, described in steps A, determine that fault type comprises: determining fault type according to box bearing position envelope signal corresponding to vibration signal is bearing inner race, bearing outer ring, bearing roller or retainer.

By upper, realize the judgement for fault type.

Optionally, fault level matrix arranges according to statistics in advance described in step B, and the parameters scope corresponding to each monitoring target of described gearbox of wind turbine comprises:

Fault characteristic frequency vibration amplitude [the Vi_x at box bearing position, Vi_y], wear particle rate of growth [IN_Ri_x in gear case fluid, IN_Ri_y], the parameters scope of gear case oil temperature [T_Oi_x, T_Oi_y], bearing temperature [T_Bi_x, T_Bi_y];

Described fault level matrix is:

By upper, realize the division to the fault level under different faults type.

Optionally, parameter corresponding to different monitoring targets is provided with different weights, and described weight comprises static state and/or changeable weight;

While calculating the probability of each fault level generation according to described distribution situation described in step C, also in conjunction with described weight calculation.

Optionally, described static weight comprises one of at least following:

The static weight of the fault characteristic frequency vibration amplitude at box bearing position is 1;

According to the degree of correlation of wear particle rate of growth, gear case oil temperature, bearing temperature and fault characteristic frequency vibration amplitude in gear case fluid, calculate this static weight of three respectively.

Optionally, described changeable weight comprises one of at least following:

Determine the changeable weight of the fault characteristic frequency vibration amplitude at box bearing position according to wind-powered electricity generation unit realtime power and rated power ratio;

Determine the changeable weight of wear particle rate of growth in gear case fluid according to the ratio of current moved number of days and the total number of days of design and operation;

Determine the changeable weight of gear case oil temperature and bearing temperature according to the ratio of the maximal value difference of the maximal value of environment temperature and the maximal value of current environmental temperature difference and environment temperature and environment temperature.

By upper, by the setting of different weights, or static weight is set, or changeable weight is set, or static weight and changeable weight are set simultaneously, to the division of each parameter weight, be beneficial under different situations different parameters is merged mutually with the analysis to fault, obviously improved accuracy.

Optionally, it is characterized in that, while calculating according to described distribution situation the probability that each fault level occurs described in step C, under each fault level, corresponding described probability adopts following manner to calculate:

P＝(Vi_Factor*W1_V*W2_V+IN_Ri_Factor*W1_IN_Ri*W2_IN_Ri+TBi_Factor*W1_TB*W2_TB+TOi_Factor*W1_TO*W2_TO)/(W1_V*W2_V+W1_IN_Ri*W2_IN_Ri+W1_TB*W2_TB+W1_TO*W2_TO)；

In formula, W1_V, W1_IN_Ri, W1_TB and W1_TO represent respectively the static weight of wear particle rate of growth, gear case oil temperature and bearing temperature in the fault characteristic frequency vibration amplitude, gear case fluid at box bearing position;

In formula, W2_V, W2_IN_Ri, W2_TB and W2_TO represent respectively the changeable weight of wear particle rate of growth, gear case oil temperature and bearing temperature in the fault characteristic frequency vibration amplitude, gear case fluid at box bearing position;

In formula, Vi_Factor, IN_Ri_Factor, TBi_Factor and TOi_Factor represent respectively the calculated factor under current fault level, when the parameter of current monitoring corresponding to monitoring target is under this fault level within the scope of corresponding parameter value, calculated factor value is 1, otherwise value is 0.

By upper, the different weights based on various faults signal are carried out schematic diagram convergence analysis, have avoided the blindness of single failure parameter location fault, have reduced rate of false alarm, improve the precision of gearbox of wind turbine fault level judgement.

Brief description of the drawings

Fig. 1 is the process flow diagram that raising gearbox of wind turbine fault level judges the method for precision;

Fig. 2 is the corresponding fault parameter scope of the different brackets of fault schematic diagram;

Fig. 3 is fault level matrix schematic diagram.

Embodiment

A kind of method that improves gearbox of wind turbine fault level and judge precision involved in the present invention, based on the method for wear particle, gear case oil temperature, four fault parameter convergence analysis of box bearing temperature in box bearing position vibration signal, gear case fluid, realize the precision that improves the judgement of gearbox of wind turbine fault level.

Principle schematic as shown in Figure 1, comprises step:

S10: according to historical data, for different faults type, determine the fault parameter scope of its different faults grade.

The fault of gearbox of wind turbine is common in bearing inner race, bearing outer ring, bearing roller and retainer four parts.Thus, the fault that comes across respectively above-mentioned position is everywhere divided into four kinds of fault types.

Further, in above-mentioned Four types, all comprise level Four fault level, be respectively: non-fault level (F_Level1), minor failure level (F_Level2), generic failure level (F_Level3) and catastrophic failure level (F_Level4).

The fault parameter of gearbox of wind turbine comprises: the fault characteristic frequency vibration amplitude (Vi) being detected by vibration transducer; Wear particle rate of growth (IN_Ri) in the gear case fluid being detected by online fluid sensor; The gear case oil temperature (T_Oi) being detected by temperature sensor and bearing temperature (T_Bi).

Historical data when employing is broken down, counts the corresponding fault parameter scope of different brackets of different faults type.As shown in Figure 2, the vibration amplitude when there is bearing inner race fault describes as example.Gearbox of wind turbine occurs, after bearing inner race fault, transferring the historical data that sensor gathers.Fig. 2 horizontal ordinate is the time series of sampled point, and ordinate is vibration amplitude.By visual in figure, when vibration amplitude interval is [Vi_1, Vi_2], time, equipment is non-fault level; When vibration amplitude interval is [Vi_3, Vi_4], time, equipment is minor failure level; When vibration amplitude interval is [Vi_5, Vi_6], time, equipment is generic failure level; In the time that vibration amplitude interval is [Vi_7, Vi_8], equipment is catastrophic failure level.In the present embodiment, Vi_1=0.85, Vi_2=1.2, Vi_3=1.7, Vi_4=2.2, Vi_5=2.4, Vi_6=3.2, Vi_7=3.8, Vi_8=4.3.

By upper, with this type of heap, statistics show that four fault levels of bearing inner race distinguish wear particle rate of growth in corresponding fluid, bearing temperature and gear case oil temperature range respectively.

In like manner, then count respectively under bearing outer ring fault, bearing roller fault and retainer failure condition, level Four fault level is distinguished corresponding each fault parameter scope.

Step S20: for without fault type, set up respectively fault level matrix.

As shown in Figure 3, describe as an example of bearing inner race fault example.

When non-fault level, be F_Level1, fault characteristic frequency vibration amplitude Vi, wear particle rate of growth IN_Ri, oil liquid temperature value T_Oi and bearing temperature value T_Bi between corresponding parameter region are respectively: [Vi_1, Vi_2], [IN_Ri_1, IN_Ri_2], [T_Oi_1, T_Oi_2] and [T_Bi_1, T_Bi_2];

When minor failure level, i.e. F_Level2, four fault parameters corresponding interval are respectively: [Vi_3, Vi_4], [IN_Ri_3, IN_Ri_4], [T_Oi_3, T_Oi_4] and [T_Bi_3, T_Bi_4];

When generic failure level, i.e. F_Level3, four fault parameters corresponding interval are respectively: [Vi_5, Vi_6], [IN_Ri_5, IN_Ri_6], [T_Oi_5, T_Oi_6] and [T_Bi_5, T_Bi_6];

And when catastrophic failure level, i.e. F_Level4, four fault parameters corresponding interval are respectively: [Vi_7, Vi_8], [IN_Ri_7, IN_Ri_8], [T_Oi_7, T_Oi_8] and [T_Bi_7, T_Bi_8].

In like manner, fault level matrix while more respectively setting up bearing outer ring fault, bearing roller fault and retainer fault.Due to the difference that different fault types shows, therefore, although four fault types all adopt parameter above to set up ranking matrix, each interval value in the matrix of each fault type is different.

Step S30: determine fault type, and determine the fault level under this fault type according to each fault parameter that gathers.

Concrete, step S30 comprises:

Step S301: determine fault type, determine bearing inner race, bearing outer ring, bearing roller and retainer four parts any part in this example and break down.

The vibration performance value that Hilbert (Hilbert) transfer pair gathers is resolved, and asks for the envelope signal of analytic signal, and envelope signal is carried out to Fast Fourier Transform (FFT) (FFT), to obtain envelope spectrum figure.By determining fault type to the analysis of envelope spectrogram.For example, in the time that bearing inner race position exists fault, on envelope spectrogram, 5～6 times turn frequent rate place and there will be compared with amplitude; In the time that bearing outer ring position exists fault, on envelope spectrogram, 3～4 times turn frequent rate place and there will be compared with amplitude.The concrete steps of above-mentioned definite fault type are same as the prior art, therefore repeat no more.

Step S302: determining on the basis of fault type, determining the fault level under this fault type.

In the present embodiment, the probabilistic algorithm formula that adopts following fault level to occur

P _ij=(Vi_Factor*W1_V*W2_V+IN_Ri_Factor*W1_IN_Ri*W2_IN_Ri+TBi_F actor*W1_TB*W2_TB+TOi_Factor*W1_TO*W2_TO)/(W1_V*W2_V+W1_IN_Ri*W2_IN_Ri+W1_TB*W2_TB+W1_TO*W2_TO), j=1,2,3,4, calculates respectively the probability that different faults grade occurs.

In formula, P _ijrepresent under fault type of the same race the probability that different faults grade occurs, P _i1represent trouble-free probability, P _i2represent the probability of minor failure level, P _i3represent the probability of generic failure level, P _i4represent the probability of catastrophic failure level.

Vi_Factor (vibration amplitude), IN_Ri_Factor (wear particle rate of growth), TBi_Factor (bearing temperature) and TOi_Factor (gear case oil temperature) represent respectively under certain fault level, the calculated factor of Real-time Collection fault parameter, described calculated factor value 1 or 0.Illustrate, in the time calculating the probability of non-fault level generation, the matrix of setting up according to step S20, if Real-time Collection fault parameter is respectively at [Vi_1, Vi_2], [IN_Ri_1, IN_Ri_2], [T_Oi_1, T_Oi_2] and [T_Bi_1, T_Bi_2] in scope, corresponding, each calculated factor value 1, otherwise, if which Real-time Collection fault parameter is in above-mentioned scope, the calculated factor value 0 of which correspondence.

W1_V, W1_IN_Ri, W1_TB and W1_TO represent respectively each image data static weight that statistical computation goes out.

Because fault characteristic frequency vibration amplitude is the Main Basis that judges gearbox of wind turbine fault, therefore, set the weighted value W1_V=1 of fault characteristic frequency vibration amplitude;

Adopt cross correlation algorithm to calculate the weight of other three fault parameters than W1_V, the static weight W1_ of wear particle rate of growth

$\mathrm{IN}\_\mathrm{Ri}=\frac{{\mathrm{\Σ}}_{i=1}^N(V_i-\stackrel{\‾}{V})(\mathrm{IN}\_\mathrm{Ri}-\stackrel{\‾}{\mathrm{IN}\_R})}{\sqrt{{\mathrm{\Σ}}_{i=1}^N{(V_i-\stackrel{\‾}{V})}^2}\sqrt{{\mathrm{\Σ}}_{i=1}^N{(\mathrm{IN}\_\mathrm{Ri}-\stackrel{\‾}{\mathrm{IN}\_R})}^2}},$ In formula represent the mean value of fault characteristic frequency vibration amplitude, represent the mean value of wear particle rate of growth;

The static weight of bearing temperature $W1\_\mathrm{TB}=\frac{{\mathrm{\Σ}}_{i=1}^N(V_i-\stackrel{\‾}{V})(T\_\mathrm{Bi}-\stackrel{\‾}{T\_B})}{\sqrt{{\mathrm{\Σ}}_{i=1}^N{(V_i-\stackrel{\‾}{V})}^2}\sqrt{{\mathrm{\Σ}}_{i=1}^N{(T\_\mathrm{Bi}-\stackrel{\‾}{T\_B})}^2}},$ In formula represent the mean value of bearing temperature;

The static weight of gear case oil temperature

$W1\_\mathrm{TO}=\frac{{\mathrm{\Σ}}_{i=1}^N(V_i-\stackrel{\‾}{V})(T\_\mathrm{Oi}-\stackrel{\‾}{T\_O})}{\sqrt{{\mathrm{\Σ}}_{i=1}^N{(V_i-\stackrel{\‾}{V})}^2}\sqrt{{\mathrm{\Σ}}_{i=1}^N{(T\_\mathrm{Oi}-\stackrel{\‾}{T\_O})}^2}},$ In formula represent the mean value of gear case oil temperature.

W2_V, W2_IN_Ri, W2_TB and W2_TO represent respectively each image data changeable weight that statistical computation goes out.

Above-mentioned four class fault parameters are affected by varying environment, and for example fault characteristic frequency vibration amplitude is along with the variation of power of fan P; Wear particle rate of growth changes with gear case of blower working time, and bearing temperature and gear case oil temperature are with variation of ambient temperature.Thus, under different environmental parameters, there is the dynamic change weight to fault degree location in four class fault parameters.

The changeable weight W2_V=1-Pi/P_rate of fault characteristic frequency vibration amplitude, in formula, P_rate is blower fan rated power, Pi is blower fan realtime power.Wherein, in the time that the realtime power Pi of blower fan is larger, represent that its impact for fault characteristic frequency vibration amplitude is just stronger, corresponding, in the time determining fault type by fault characteristic frequency vibration amplitude, its precision becomes and can decline, based on this, adopt above-mentioned formula, in the time that the realtime power Pi of blower fan is larger, for reducing its impact for fault characteristic frequency vibration amplitude, therefore make its changeable weight value less.

The changeable weight W2_IN_Ri=1-DAYSi/DAYS_Design of wear particle rate of growth, in formula, DAYS_Design is the total number of days of equipment design and operation, DAYSi is current the moved number of days of equipment.Wherein, when equipment institute, days running is more, and its degree of aging is also just higher, therefore, likely occurs, even if in the time of non-fault, wear particle rate of growth also can rise.Based on this, while setting the changeable weight of wear particle rate of growth, reduce due to the impact of time on it.

The changeable weight W2_TB of bearing temperature is identical with the changeable weight W2_TO of gear case oil temperature, $W2\_\mathrm{TB}=W2\_\mathrm{TO}=\frac{{T\_\mathrm{amb}}_{\max }-{T\_\mathrm{amb}}_i}{{T\_\mathrm{amb}}_{\max }-{T\_\mathrm{amb}}_{\min }},$ T_amb in formula _ifor actual measurement environment temperature, T_amb _maxfor the maximal value of environment temperature, T_amb _minfor the minimum value of environment temperature.Owing to not coexisting season, for the difference that affects in equipment cabin.Due to summer high temperature, affected by it, bearing temperature and gear case oil temperature can increase, and obviously, above-mentioned high temperature is not that fault causes, but environment temperature causes.Thus, adopt above-mentioned formula just can effectively reduce because the caused bearing temperature of rising and the gear case oil temperature of environment temperature rise.

By upper, calculate respectively the probability occurring in different faults grade, calculate the probability that fault level 1-4 occurs separately, the high person of probability is broken down grade.Adopt said method, based on the diagnostic method of metallic particles, bearing temperature and four fault parameter convergence analysis of gear case oil temperature in bear vibration, gear case fluid, can greatly improve the accuracy of gearbox of wind turbine fault level judgement, the blindness of having avoided single failure parameter location fault, has reduced rate of false alarm.

Step S303: fault type and fault level judged result are shown to user, so that user learns type and fault level when prior fault.

The foregoing is only preferred embodiment of the present invention, not in order to limit the present invention, for example, although above-mentioned taking four class fault types as example, using four fault parameters in each fault type is example, this be because, here fault is the fault that most probable occurs, these fault parameters are parameters the closest with these faults, and unrestricted only for four class fault types, four class fault parameters, and four above-mentioned fault levels.Within the spirit and principles in the present invention all in a word, any amendment of doing, be equal to replacement, improvement etc., within all should being included in protection scope of the present invention.

Claims (7)

1. improve the method that gearbox of wind turbine fault level judges precision, it is characterized in that, comprise step:

A, determine fault type according to box bearing position vibration signal characteristics value;

B, the distribution situation of parameters in the fault level matrix of this fault type of determining current monitoring corresponding to each monitoring target of gearbox of wind turbine;

Described fault level matrix comprises each fault level of setting, and each fault level comprises parameters span corresponding to each monitoring target of gearbox of wind turbine under this fault level;

C, calculate according to described distribution situation the probability that each fault level occurs, and determine current described fault level accordingly.

2. method according to claim 1, it is characterized in that, determine that fault type comprises described in steps A: determining fault type according to box bearing position envelope signal corresponding to vibration signal is bearing inner race, bearing outer ring, bearing roller or retainer.

3. method according to claim 1, is characterized in that, fault level matrix arranges according to statistics in advance described in step B, and the parameters scope corresponding to each monitoring target of described gearbox of wind turbine comprises:

Fault characteristic frequency vibration amplitude [the Vi_x at box bearing position, Vi_y], wear particle rate of growth [IN_Ri_x in gear case fluid, IN_Ri_y], the parameters scope of gear case oil temperature [T_Oi_x, T_Oi_y], bearing temperature [T_Bi_x, T_Bi_y];

Described fault level matrix is:

4. method according to claim 3, is characterized in that, parameter corresponding to different monitoring targets is provided with different weights, and described weight comprises static state and/or changeable weight;

While calculating the probability of each fault level generation according to described distribution situation described in step C, also in conjunction with described weight calculation.

5. method according to claim 4, is characterized in that, it is one of at least following that described static weight comprises:

The static weight of the fault characteristic frequency vibration amplitude at box bearing position is 1;

According to the degree of correlation of wear particle rate of growth, gear case oil temperature, bearing temperature and fault characteristic frequency vibration amplitude in gear case fluid, calculate this static weight of three respectively.

6. method according to claim 4, is characterized in that, it is one of at least following that described changeable weight comprises:

Determine the changeable weight of the fault characteristic frequency vibration amplitude at box bearing position according to wind-powered electricity generation unit realtime power and rated power ratio;

Determine the changeable weight of wear particle rate of growth in gear case fluid according to the ratio of current moved number of days and the total number of days of design and operation;

Determine the changeable weight of gear case oil temperature and bearing temperature according to the ratio of the maximal value difference of the maximal value of environment temperature and the maximal value of current environmental temperature difference and environment temperature and environment temperature.

7. method according to claim 6, is characterized in that, while calculating according to described distribution situation the probability that each fault level occurs described in step C, under each fault level, corresponding described probability adopts following manner to calculate:

P＝(Vi_Factor*W1_V*W2_V+IN_Ri_Factor*W1_IN_Ri*W2_IN_Ri+TBi_Factor*W1_TB*W2_TB+TOi_Factor*W1_TO*W2_TO)/(W1_V*W2_V+W1_IN_Ri*W2_IN_Ri+W1_TB*W2_TB+W1_TO*W2_TO)；

In formula, W1_V, W1_IN_Ri, W1_TB and W1_TO represent respectively the static weight of wear particle rate of growth, gear case oil temperature and bearing temperature in the fault characteristic frequency vibration amplitude, gear case fluid at box bearing position;

In formula, W2_V, W2_IN_Ri, W2_TB and W2_TO represent respectively the changeable weight of wear particle rate of growth, gear case oil temperature and bearing temperature in the fault characteristic frequency vibration amplitude, gear case fluid at box bearing position;

In formula, Vi_Factor, IN_Ri_Factor, TBi_Factor and TOi_Factor represent respectively the calculated factor under current fault level, when the parameter of current monitoring corresponding to monitoring target is under this fault level within the scope of corresponding parameter value, calculated factor value is 1, otherwise value is 0.