a novel technique of crack detection for …aiac-12 twelfth australian international aerospace...

12
AIAC-12 Twelfth Australian International Aerospace Congress Fifth DSTO International Conference on Health & Usage Monitoring – HUMS2007 A Novel Technique of Crack Detection for Helicopter Main Gearbox Planet Carrier Wenyi Wang Air Vehicles Division DSTO Fishermans Bend, Melbourne Australia Jonathan A. Keller US Army RDECOM Aviation Engineering Directorate Redstone Arsenal, Alabama USA Abstract This paper presents a simple but effective technique of detecting helicopter main gearbox planet carrier cracking, which occurred in two US Army UH-60A Blackhawks in 2002 and one US Navy SH-60B Seahawk in 2004. The vibration signal averaged with respect to the rotation of planet carrier contains the gear mesh components between the planet gears and the ring gear, which are modulated by the planet passing events. It is expected that a cracked planet carrier should produce a different modulation pattern to the gear mesh harmonics from an uncracked planet carrier. Therefore, in the proposed technique, our focus is on the extraction of the normalised modulation waveforms. We first normalise the averaged signals to a mean of zero and a standard deviation of one unit, and extract the planet passing components from the normalised signals. We then generate a Condition Index (CI) from them. Using the CI, we can differentiate the cracked planet carrier from the uncracked carrier at 30 percent torque load and above for both the Test-Cell and On-Aircraft data. With further validation, this technique can be readily implemented into any helicopter health monitoring system. 1. Introduction Recently there were three cases of main gearbox planet carrier cracking that occurred in two UH-60A helicopters of the US Army and one SH-60B helicopter of the US Navy. This paper presents a simple but effective technique of detecting cracks in planet carrier plates, with analysis results from two data sets: (1) Test-Cell data generated in a test conducted in the US Navy’s helicopter transmission test facility (HTTF) in Maryland with a healthy gearbox and a gearbox with a cracked carrier; and (2) On-Aircraft data generated in four UH-60A helicopters, one with the cracked carrier and the rest with uncracked planet carriers. The fundamental principle of the technique is based on the fact that a cracked planet carrier makes the load sharing of gear mesh different from that of an uncracked carrier, which may become more significant under higher torque loads. This change of gear mesh load sharing can have an effect on the amplitudes of the planet passing harmonics. In the vibration signal averaged with respect to the rotation of planet carrier, the predominant components are the gear mesh harmonics between the planet gears and the ring gear, modulated by the planet passing event. It is expected that a cracked planet carrier should produce a different modulation pattern to an uncracked planet carrier. Therefore, in the proposed technique, the averaged signals are first normalised to a zero-mean and a unit standard deviation, and the planet passing harmonics (or the modulation waveforms) are then extracted from the normalised signals. A Condition Index (CI) based on the standard deviation of modulation waveform’s 5 th shaft order harmonics (as there are five planet gears) is calculated and used for trending or plotted against the main gearbox load.

Upload: others

Post on 04-Apr-2020

2 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: A Novel Technique of Crack Detection for …AIAC-12 Twelfth Australian International Aerospace Congress Fifth DSTO International Conference on Health & Usage Monitoring – HUMS2007

AIAC-12 Twelfth Australian International Aerospace Congress

Fifth DSTO International Conference on Health & Usage Monitoring – HUMS2007

A Novel Technique of Crack Detection for Helicopter Main

Gearbox Planet Carrier

Wenyi Wang

Air Vehicles Division

DSTO

Fishermans Bend, Melbourne

Australia

Jonathan A. Keller

US Army RDECOM

Aviation Engineering Directorate

Redstone Arsenal, Alabama

USA

Abstract

This paper presents a simple but effective technique of detecting helicopter main gearbox

planet carrier cracking, which occurred in two US Army UH-60A Blackhawks in 2002 and

one US Navy SH-60B Seahawk in 2004. The vibration signal averaged with respect to the

rotation of planet carrier contains the gear mesh components between the planet gears and the

ring gear, which are modulated by the planet passing events. It is expected that a cracked

planet carrier should produce a different modulation pattern to the gear mesh harmonics from

an uncracked planet carrier. Therefore, in the proposed technique, our focus is on the

extraction of the normalised modulation waveforms. We first normalise the averaged signals

to a mean of zero and a standard deviation of one unit, and extract the planet passing

components from the normalised signals. We then generate a Condition Index (CI) from

them. Using the CI, we can differentiate the cracked planet carrier from the uncracked carrier

at 30 percent torque load and above for both the Test-Cell and On-Aircraft data. With further

validation, this technique can be readily implemented into any helicopter health monitoring

system.

1. Introduction

Recently there were three cases of main gearbox planet carrier cracking that occurred in two

UH-60A helicopters of the US Army and one SH-60B helicopter of the US Navy. This paper

presents a simple but effective technique of detecting cracks in planet carrier plates, with

analysis results from two data sets: (1) Test-Cell data generated in a test conducted in the US

Navy’s helicopter transmission test facility (HTTF) in Maryland with a healthy gearbox and a

gearbox with a cracked carrier; and (2) On-Aircraft data generated in four UH-60A

helicopters, one with the cracked carrier and the rest with uncracked planet carriers.

The fundamental principle of the technique is based on the fact that a cracked planet carrier

makes the load sharing of gear mesh different from that of an uncracked carrier, which may

become more significant under higher torque loads. This change of gear mesh load sharing

can have an effect on the amplitudes of the planet passing harmonics.

In the vibration signal averaged with respect to the rotation of planet carrier, the predominant

components are the gear mesh harmonics between the planet gears and the ring gear,

modulated by the planet passing event. It is expected that a cracked planet carrier should

produce a different modulation pattern to an uncracked planet carrier. Therefore, in the

proposed technique, the averaged signals are first normalised to a zero-mean and a unit

standard deviation, and the planet passing harmonics (or the modulation waveforms) are then

extracted from the normalised signals. A Condition Index (CI) based on the standard

deviation of modulation waveform’s 5th

shaft order harmonics (as there are five planet gears)

is calculated and used for trending or plotted against the main gearbox load.

Page 2: A Novel Technique of Crack Detection for …AIAC-12 Twelfth Australian International Aerospace Congress Fifth DSTO International Conference on Health & Usage Monitoring – HUMS2007

2

By comparing the cracked and uncracked conditions using the Test-Cell data, we find that the

CI from the gearbox with the cracked planet carrier progressively decreases as the torque load

is increased from 30 percent to 100 percent rated torque. However, we could not separate the

cracked and uncracked conditions at the 20 percent torque load for both Test-Cell and On-

Aircraft data.

The results presented in the paper show that the proposed technique can be effectively used to

detect a cracked planet carrier provided there is a significant torque load (e.g., 30 percent)

applied to the gearbox. With further validation, this technique can be readily implemented

into helicopter health monitoring systems that have synchronous averaging capability.

2. The UH-60A Test Data and Analyses by Others

There were two UH-60A test data sets used in this paper. The first was the Test-Cell data

generated in a test conducted in the US Navy’s helicopter transmission test facility (HTTF) in

Maryland with a healthy gearbox and a gearbox that had an 8.25cm (3.25”) planet carrier

crack. The test was conducted under a range of torque loads from 20 percent to 100 percent.

The second set was the On-Aircraft data generated by four UH-60A helicopters, one with the

same cracked carrier used in the Test-Cell and the rest with uncracked planet carriers, at 20

percent and 30 percent torque loading only, as 30 percent torque was the maximum safe load

while the helicopters remained on ground.

Figure 1 shows the schematic of the other UH-60A carrier crack, which was 25cm (10”) long.

Figure 2 shows the top view of the Blackhawk’s main transmission gearbox and the sensor

locations.

Fig. 1. The US Army’s UH-60A planet carrier with a 25cm long crack.

Page 3: A Novel Technique of Crack Detection for …AIAC-12 Twelfth Australian International Aerospace Congress Fifth DSTO International Conference on Health & Usage Monitoring – HUMS2007

3

Fig.2. Blackhawk’s main transmission gearbox and the sensor locations

The main module of the UH-60A main transmission is a two-stage gearbox. The first is a

bevel gear stage with a 17-tooth pinion and an 81-tooth bevel gear, and the second is an

epicyclic stage with a 62-tooth sun gear, five 83-tooth planet gears, a 228-tooth fixed ring

gear and a planet carrier plate. In addition, there is an auxiliary oil pump drive gear with 152

teeth, which rotates with the carrier plate.

The data sets were first analysed by Keller and Grabill [1] using several standard diagnostic

parameters and their modifications. They found that only two of the parameters were capable

of detecting the cracked condition using the Test-Cell data, and no parameter was able to

detect the cracked condition using the On-Aircraft data. Later analysis conducted by Blunt

and Keller [2], through detecting changes to the planet passing modulation of the gear mesh

vibration and changes to the meshing behaviour of the individual planet gears, showed an

improvement on the existing techniques and a reliable detection of cracked condition using

the Test-Cell data.

The data were also widely disseminated and analysed by many researchers. Many existing

and new techniques [3-8] were applied with various degrees of success. Among the new

methods, an impressive approach by Dong et al [3] employed an Advanced-Hidden Markov

model (AHMM) to extract condition indices for the cracked and uncracked conditions. Their

analysis results have shown that the AHMM can be effectively used to identify the cracked

carrier using both Test-Cell and On-Aircraft data. Dong and He [4] further investigated the

Hidden Semi-Markov Model (HSMM), which was applied to the detection of UH-60A planet

carrier crack with excellent results. However, it is believed that the increased computational

complexity is a major drawback of the HSMM. McInerny et al [6] employed a method based

on the spectral energy ratio between the non planet passing residuals and the planet passing

GEARS Input

pinion

Bevel

gear

Sun

gear

No. of

planets

Planet

gears

Ring

gear

Oil pump

drive gear

Tooth No. 17 81 62 5 83 228 152

Page 4: A Novel Technique of Crack Detection for …AIAC-12 Twelfth Australian International Aerospace Congress Fifth DSTO International Conference on Health & Usage Monitoring – HUMS2007

4

components. This method can produce good results if the frequency band is appropriately

chosen. The data used by McInerny et al were derived from sensor #6, i.e., the Port-Ring

accelerometer.

3. Analytical procedure

In this paper, we propose a simple and yet effective approach to detecting the planet carrier

cracking in a helicopter main gearbox. Firstly, if we look at any planet-carrier signal average

in time domain, we will find five peaks (or humps) associated with the passing of the five

planet gears in the UH-60A main gearbox, and the higher frequency gear mesh vibration

between the planet gears and the ring gear. To further illustrate this, we can extract the low

frequency planet passing component from the averaged signals by a simple envelope analysis.

The enveloped signal shown in Figure 3 represents the modulation waveform of the gear

mesh vibration from an uncracked planet carrier at two different loads. As can be seen, these

modulation waveforms are very dependent on load, i.e., the five peaks are more prominent at

higher load. They should also be highly dependent on the stiffness of the carrier plate (see

Figures 4 and 5). We need to point out that the envelope signals shown in Figures 3, 4 and 5

only contain the low frequency content of modulation waveforms. They are used to illustrate

the effects of loading and cracking on modulations, and will not be used to extract the

condition index for crack detection as they lack the high frequency content.

Fig. 3. Original planet carrier modulation to gear mesh vibration

When the carrier plate develops a crack, its torsional stiffness will change, which will in turn

change the modulation of the gear mesh vibration. One might expect that a condition index

extracted from the modulation waveform of the original vibration signal averages should

reflect the changes of the weakened carrier plate. However, it is very hard to control the test

conditions such that the acquired vibration signals have identical basis for comparison. This is

Page 5: A Novel Technique of Crack Detection for …AIAC-12 Twelfth Australian International Aerospace Congress Fifth DSTO International Conference on Health & Usage Monitoring – HUMS2007

5

especially important for the On-Aircraft data that were acquired from four different

helicopters. In the proposed technique, we first normalise the averaged signals to a mean of

zero and a standard deviation of one unit, to make the comparison basis identical under

different test conditions, i.e., different loading and different aircraft. We need to point out that

the key to differentiating between a cracked and an uncracked carrier plate is the relative

changes in the modulation waveforms, rather than the absolute changes.

Figure 4 is the normalised version of Figure 3. For the uncracked planet carrier, the difference

between the two modulation waveforms is very obvious but their mean values are very

similar. Figure 5 shows the modulation waveforms from a cracked planet carrier with

loadings at 20 percent and 100 percent. Comparing Figure 4 with Figure 5, we find that the

modulation waveforms associated with the cracked carrier differ significantly from those with

the uncracked carrier. With the cracked carrier, the five peaked modulation pattern tends to

disappear, but the mean value does not change significantly at different loads due to the

normalisation.

A simplistic explanation of this could be that a local crack in the carrier plate is weakening

the local gear mesh stiffness between the planet gear adjacent to the crack and the ring gear,

and the weakening is redistributing the load of the affected planet to the other planets. The

flattening of the modulation peaks could be a consequence of this. As the crack propagates, it

could be expected that more planet gears will become affected and the modulation peaks will

be flattened more.

Fig. 4. Normalised planet carrier modulation to gear mesh vibration from an uncracked planet carrier

Based on the above observations, our focus is placed on the difference in the modulation

patterns. Because the five-peaked modulation pattern disappears for the cracked planet carrier,

the harmonics of the 5th

shaft order should diminish as well, which will be more obvious

Page 6: A Novel Technique of Crack Detection for …AIAC-12 Twelfth Australian International Aerospace Congress Fifth DSTO International Conference on Health & Usage Monitoring – HUMS2007

6

under heavy load. Hence, if we can extract a condition index from the 5th

shaft order

harmonics, e.g., the standard deviation of all 5th

shaft order harmonics, we should see a

declining trend of the CI values against load. This forms the basis for the proposed detection

technique. The procedure is outlined in Figure 6. The most important step in the procedure is

the normalisation which allows the relative changes to be tracked.

Fig. 5. Normalised planet carrier modulation to gear mesh vibration from a cracked planet carrier

Fig. 6. A diagram of the proposed planet carrier crack detection technique for helicopter main

transmission gearboxes

Raw

vibration

signal

Reference

tacho

signal

Synchronous

averaging with

respect to the

rotation of planet

carrier

Normalisation to

zero-mean & unit-

standard deviation

Fourier

transform

Extract planet

passing

harmonics

Inverse

Fourier

transform

Calculate

standard

deviations

Condition

Index (CI)

Page 7: A Novel Technique of Crack Detection for …AIAC-12 Twelfth Australian International Aerospace Congress Fifth DSTO International Conference on Health & Usage Monitoring – HUMS2007

7

4. Analysis results

In this section, we first present a spectral analysis of the epicyclic gearbox vibration. Then,

the analysis results for both the Test-Cell and On-Aircraft data are given. In this paper, we

focus on the analysis of the data acquired by sensor #6 (Port-Ring).

Spectral Analysis

Based on the information given in Section 2 about the UH-60A main gearbox, we expect to

see all the gear mesh harmonics plus bearing vibration and background noise in the spectrum

of any raw signal. Because the fault is on the carrier plate, we need to isolate the vibration

associated with the rotation of the carrier plate. We can achieve this with synchronous

averaging, using the tachometer signal on the output shaft as a rotational reference. In the

spectrum of such an averaged signal, one might expect to see frequency components at the

harmonics of the 228th

shaft order and the 152nd

order, accompanied by modulation sidebands

at 5 shaft orders. In fact, as shown in Figure 7, it is the harmonics of the planet passing

frequency (i.e., multiples of the 5th

shaft order) around the harmonics of the 228th

shaft order

that dominate the spectrum. The harmonics of the 228th

shaft order gear mesh frequency are

actually suppressed (not shown as the dominant components). This phenomenon was

thoroughly explained by McFadden and Smith [9]. The harmonics of the oil pump drive gear

(152nd

shaft order) are also noticeable in Figure 7 as this gear rotates together with the planet

carrier and the vibration induced by this gear is not averaged out.

Fig.7. Spectrum of an averaged signal from Test-Cell accelerometer #6.

Analysis of Test-Cell Data

First, let us take a look at the original signal average to see whether it can lead us to separate

the uncracked and cracked planet carriers. Figure 8 shows a plot of the standard deviation of

Page 8: A Novel Technique of Crack Detection for …AIAC-12 Twelfth Australian International Aerospace Congress Fifth DSTO International Conference on Health & Usage Monitoring – HUMS2007

8

the original Port-Ring signal averages versus the load. Obviously, the curves associated with

the uncracked and cracked carriers are not separated, which means that the overall variations

in planet carrier signal average are not indicative to the carrier fault. However, based on the

discussions given in Section 3, the vibration induced by the passage of the planet carrier

should be indicative to the carrier cracking.

Fig.8. Plot of standard deviations of planet carrier signal averages against gearbox torque load using

the data at Test-Cell accelerometer #6.

With the normalised signal averages of the planet carrier, we now separate the planet passing

components from the gear (i.e., planet gears, ring gear and oil pump drive gear) mesh

components. Figures 9 ~ 11 show the plots of the standard deviations of these separated

components versus load. The fundamental information that we can derive from these plots is

that, with the cracked planet carrier, the standard deviation of:

• the planet passing harmonics decreases with increasing load;

• the ring/planet gear mesh harmonics plus residuals increases with increasing load, which

is probably why the unseparated signal average does not show a monotonic trend; and

• the oil pump drive gear mesh harmonics increases with increasing load.

This is in contrast to the results from the normalised signal averages of the uncracked planet

carrier (the dashed lines in the plots), which are relatively stable as the load increases.

Therefore, any or a combination of the above mentioned measures can be used as the

condition index for detecting a planet carrier crack. Due to the dominance of the planet

passing harmonics in the averaged signals, we have employed the standard deviation (shown

in Figure 9) of these extracted components (transformed back into the time domain) as the

condition index in our study. The process of computing the CI is outlined in Figure 6 of this

paper. However, for consistency and possibly fewer false alarms, the results from all three

plots could be employed together.

Page 9: A Novel Technique of Crack Detection for …AIAC-12 Twelfth Australian International Aerospace Congress Fifth DSTO International Conference on Health & Usage Monitoring – HUMS2007

9

Fig.9. Plot of standard deviation of planet passing harmonics versus torque load using the data at

Test-Cell accelerometer #6.

Fig.10. Plot of standard deviation of ring/planet gear mesh harmonics& residuals versus torque load

using the data at Test-Cell accelerometer #6.

Page 10: A Novel Technique of Crack Detection for …AIAC-12 Twelfth Australian International Aerospace Congress Fifth DSTO International Conference on Health & Usage Monitoring – HUMS2007

10

Fig.11. Plot of standard deviation of oil pump drive gear mesh harmonics versus torque load using the

data at Test-Cell accelerometer #6.

We need to emphasise that the above three standard deviations shown in Figures 9 ~ 11 are

relative to the overall standard deviation of one unit, which is independent of the original unit

(i.e., G for acceleration, or Volt, etc.) of the vibration signals. However, the contributions

from the individual harmonic components may not necessarily sum up to one because of the

phase relationships between the individual components.

Analysis of On-Aircraft Data

The Test-Cell data were acquired under controlled conditions, which improved the ability to

separate the modulation features caused by the carrier crack. However, the On-Aircraft data

were obtained from four different UH-60A helicopters (Aircraft A, B, C and X), in which one

(Aircraft X) had the cracked carrier installed. Because of the structural variability among these

aircraft, identification of the faulty gearbox with the cracked carrier in this data set is far more

difficult than using the Test-Cell data. Furthermore, the loads applied to the aircraft on the

ground could not be more than 30 percent rated load due to safety concerns.

Following the procedure shown in Figure 6, we extracted the CI, as defined by the standard

deviation of the planet passing harmonics, and the result is shown in Figure 12. This figure

shows that the gearboxes with cracked and uncracked carriers can be separated at 30 percent

load, and presumably at other higher loads if they were available. However, we also find that

the variability of CI from all the aircraft with healthy gearboxes at 20 percent load is much

larger, and we cannot separate the healthy and faulty gearboxes at this load. Further, referring

back to Figure 9, we may conclude that the 20 percent load is not sufficiently high to allow an

identification of any faulty gearbox with a cracked planet carrier for both Test-Cell and On-

Aircraft conditions.

Page 11: A Novel Technique of Crack Detection for …AIAC-12 Twelfth Australian International Aerospace Congress Fifth DSTO International Conference on Health & Usage Monitoring – HUMS2007

11

Fig.12. Plot of proposed CI’s versus torque load using the data at On-Aircraft accelerometer #6.

5. Comparison and discussion

Among the techniques presented by other researchers, there is a similar method to our

technique that was discussed by McInerny et al [6]. This method employed the spectral

energy ratio between the non planet passing residuals and the planet passing components.

Using this technique they found the most consistent result was the GM-1 metric (given in

Table II of the paper) from the frequency band around the fundamental ring/planet gear mesh

frequency. The consensus between our approach and McInerny’s method is the fact that the

non-planet passing residuals behave differently from the planet passing components when a

carrier crack is present. However, the technique proposed in this paper differs from the

McInerny’s energy ratio method in the following three aspects:

• There is no need for our technique to select frequency bands for best detection.

• Our technique is based on a time domain normalisation process, which might be

somehow comparable to the usage of ‘ratio’ in McInerny’s method.

• The CI in our technique is drawn from time domain instead of the frequency domain.

Overall, in comparison to other methods, the proposed technique has a unique procedure for

processing both the Test-Cell and the On-Aircraft data. It does not need any subjective

selection of frequency band to demodulate vibration signals and it does not rely on knowledge

about system resonances. Moreover, it is not based on any prescribed model, such as

parametric model (e.g., AR or ARMA model) that needs to estimate model parameters or

statistical model (Markov model) that needs to be trained. The normalisation process is the

key to the successful application of the technique, as it sets a unique comparison basis for data

acquired under different loading conditions and from different aircraft.

Page 12: A Novel Technique of Crack Detection for …AIAC-12 Twelfth Australian International Aerospace Congress Fifth DSTO International Conference on Health & Usage Monitoring – HUMS2007

12

6. Concluding remarks

It has been shown in this paper that the proposed technique can be used to separate the

cracked and uncracked helicopter main gearbox planet carriers provided that there is a

significant load (e.g., 30 percent rated load and above) applied to the gearbox. In conclusion,

this technique is conceptually straightforward and yet effective in detecting planet carrier

cracking. It does not involve subjective judgement, and with further validation, it can be

readily implemented into any helicopter health and usage monitoring system (HUMS) that has

synchronous averaging capability.

Acknowledgement

The authors would like to acknowledge David Blunt, Brian Rebbechi and Albert Wong of

DSTO for their support to the work and comments for the paper.

References

1. J.A. Keller and P. Grabill, “Vibration Monitoring Of UH-60A Main Transmission

Planetary Carrier Fault,” Proceedings of American Helicopter Society 59th

Annual

Meeting, Phoenix, Arizona, USA, 6-8 May 2003.

2. D.M. Blunt and J.A. Keller, “Detection of a Fatigue Crack in a UH-60A Planet Gear

Carrier using Vibration Analysis,” Mechanical System and Signal Processing, Vol.

20(8), pp. 2095-2111, 2006.

3. M. Dong, et al, “Equipment Health Diagnosis and prognosis Using Advanced Hidden

Markov Models,” Proceedings of the 58th

Meeting of the Machinery Failure

Prevention Technology (MFPT) Society, Virginia Beach, Virginia, USA, April 2004.

4. M. Dong and D. He, “Hidden semi-Markov models for machinery health diagnosis

and prognosis,” Transactions of NAMRI/SME (the North American Manufacturing

Research Institution – Society of Manufacturing Engineering), Vol. 32, pp. 199- 206,

2004.

5. B.Q. Wu, et al, “An Approach to Fault Diagnosis of Helicopter Planetary Gears,”

2004 IEEE AUTOTESTCON.

6. S.A. McInerny, et al, “Detection of a Cracked-Planet Carrier,” Proceedings of the 10th

International Congress on Sound and Vibration, Stockholm, Sweden, 7-10 July 2003.

7. P. Sparis and G. Vachtsevanos, “A Helicopter Planetary Gear Carrier Plate Crack

Analysis and Feature Extraction based on Ground and Aircraft Tests,” Proceedings of

the 2005 IEEE International Symposium on Intelligent Control, Limassol, Cyprus,

June 27-29, 2005.

8. A. Saxena, et al, “A Methodology for Analyzing Vibration Data from Planetary Gear

Systems using Complex Morlet Wavelets,” Proceedings of 2005 American Control

Conference, Portland, OR, USA, June 8-10, 2005.

9. P.D. McFadden, and J.D. Smith, “An Explanation for the Asymmetry of the

Modulation Sidebands about the Tooth Meshing Frequency in Epicyclic Gear

Vibration,” Proceedings of the Institution of Mechanical Engineers, Part C:

Mechanical Engineering Science, Vol. 199, No. 1, Jan. 1985, pp. 65-70.