A.1.1 Introduction
For equipment integrated in hardware to be launched on ARIANE or on STS, a first evaluation of test levels based on scrutiny of environmental test specifications and also of NASA Standards was performed (see references. A.1.4.1) in order to construct an “envelope”.
To improve the specifications for sinusoidal and random testing at unit level, a second evaluation was based on a statistical analysis using test data obtained from sine, random and acoustic testing at system level (see references A.1.4.2 and A.1.4.3).
For the latter purpose, a database was first prepared using ITS-DIVA software, for which seven major European satellite projects (METEOSAT, MARECS, ISPM, ECS, GIOTTO, TELECOM 1 and SPOT) provided approximately 1500 response curves and associated information.
For the random and acoustic test results, the random response spectrum (RRS) concept was used extensively. A statistical evaluation was then made on the basis of a log-normal distribution for various selections of curves using the parameters available in the database.
The final equipment classification was deduced from these results and the corre-sponding specifications based on a 95 % confidence level (see Figure A--1) were defined.
The acoustic investigation results highlighted the importance of two particular parameters for unit-level random testing:
D the equipment mass, and
D the external panel in bending compared to other results.
In order to have a limited number of classes, the masses were divided in two categories, leading to four basic classes.
A coefficient of four (4) appeared with the power spectral density(PSD) specifica-tions between smaller masses (less than 3 kg) and the higher ones.
Normalized database - Sampling
- Reference axis - Reference environment
Sine Random and acoustic
Selection of
responses Equipment classes
Statistics
Probability level Specification for equip-ment classes
Random Response Spectrum (RRS)
f f
γ
f
γ Wγ
Figure A--1: Data processing arrangement
For unit level sinusoidal testing, statistics for all responses parallel to the excita-tions showed relatively limited levels in the vicinity of first satellite global mode (longitudinal or lateral) due to the notching procedures, and lower levels at the other frequencies. The specifications were given for the same categories of mass as previously, assuming that the effects of the first satellite mode are limited to 30 Hz for the first lateral mode and 60 Hz for the first longitudinal mode.
All of these results were updated using the database supplemented with data from the HIPPARCOS and OLYMPUS projects (about 300 response curves) and improved data processing (see references A.1.4.2 and A.1.4.4), mainly:
a. a statistical analysis specifically adapted to the sinusoidal test data in the vicinity of the satellite first resonant frequency, for which the previous approach was found to be not well adapted;
b. analysis in more detail of the impact of the main parameters governing the equipment responses, particularly the equipment mass for which it was inter-esting to extrapolate in the 20 to 50 kg range for the next generation of satellites.
A detailed analysis of the HIPPARCOS and OLYMPUS acoustic response curves showed that only equipment directly mounted on honeycomb panels and in areas with sufficient equipment density can be specified with reasonable reliability in the present context. Equipment located on small brackets (such as thruster, sen-sor) or isolated from the others on large panels can have higher levels generated by local resonances. The conclusion was to remove all “non-panel” accelerometers from statistical analyses. It was also noted that the test data were clearly valid for
equipment qualification purposes only if the accelerometers were very close to the equipment fixation on the panel. Accelerometers mounted on the equipment itself, between equipment items or between distant fixations tended to produce ques-tionable data.
A study was placed by ESA with INTESPACE, with the objective of improving and analysing the mechanical database (see reference A.1.4.5).
The mechanical database was provided with data related to SOHO sinusoidal and acoustic test results and associated data. A statistical analysis was performed on the whole database (13 specimens from 12 satellite projects). The results globally confirmed the validity of the equipment classes selected for specifications in ran-dom: “external bending” and “other”, and continuous function of mass. In addition, both sinusoidal and random levels results were significantly changed, mainly due to the data from generally heavier satellites.
Evaluation based on the complete database, according to the previous consider-ations led to the results explained in subclauses A.1.2 and A.1.3.
A.1.2 Unit level random testing A.1.2.1
For each of the two classes: “external panels in bending” and “others”, a statistical analysis on acoustic test data was performed to derive specifications in the form of continuous function of mass. Two complementary approaches were used for better reliability:
a. a method using close masses: for a given mass, statistics were performed using curves corresponding to masses close to the considered value;
b. a method using normalization: statistics were performed on all the curves normalized by a function of the equipment mass representing the dependence of the PSD level; this function was selected from various considerations:
1. energy and extrapolation techniques,
2. asymptotic value representing the minimum PSD for large masses, 3. plots showing levels versus mass,
4. results from the first method, and 5. simplicity.
leading to:
PSD(M) = PSD(1 kg) × f(M) where
f(M) = [(1 + m)/(1 + k×m)] × [(M + k×m)/(M + m)]
m = driven mass from the panel
k = PSD(0)/PSD(∞)>1 ratio between extreme values.
A.1.2.2
Data evaluation using the method above (A.1.2.1 b.) corroborated by the first method (A.1.2.1 a.) led to m = 1 kg and k = 20, giving the specifications of sub-clause 5.1.11.3.
These specifications apply only to equipment directly mounted on honeycomb panel and in areas with sufficient equipment density (if not, adequate corrections are used to come to this case).
This represents a significant improvement on the qualification of units by random vibration testing and it is very easy to use. It can be seen that, for lateral axes of equipment located on “external panels” and for all axes of “other” equipment:
a. The maximum values for very small masses, less than 0,5 kg, are close to the relatively high value of 1 g2/Hz.
b. For medium masses, between 1 kg and 10 kg, the specifications are decreasing significantly (factor of 4 between 1 kg and 10 kg).
Extrapolation to large masses gives a slow convergence towards the asymptotic PSD value of 0,05 g2/Hz (e.g. .a mass of 30 kg gives 0,08 g2/Hz). However, the approach is valid only if the equipment is rigid. If some very flexible parts are included in the unit, M does not include the corresponding masses and represents the rigid part driven by the fixations. 50 kg seems to be an extreme limit in the present context, giving a minimum value of 0,07 g2/Hz.
Significantly higher values exist in vertical axis for equipment located on “exter-nal panels” (asymptotic value of 0,12 g2/Hz).
A.1.3 Unit level sinusoidal testing
During sinusoidal tests at system level, as opposed to random or acoustic tests, the responses depend mainly on the overall characteristics of the satellite and also on the notching criteria. Consequently, a very large number of parameters are involved in the equipment behaviour and a statistical approach can only give global results, which were treated with caution in the present context.
Concerning the influence of notching on the satellite responses, leading to a seri-ous problem for database normalization, among the possible assumptions, the following one was selected for its logic and its simplicity: “the levels occurring during notching cannot be exceeded”, leading to relatively limited levels in the vicinity of the first satellite global mode. In these frequency bands (up to 30 Hz in lateral and 60 Hz in longitudinal), selecting the resonance levels and performing statistics on these particular data was found to be a better approach than using statistics at each sampled frequency (better log-normality and more efficient analysis).
Data evaluation led to the specification of subclause 5.1.10.3. The 20 g below 60 Hz, independent of the equipment mass, are governed completely by the first satellite longitudinal mode.
In addition, the normalization approach was used to determine the influence of the parameters on the resonance levels. The only case showing no ambiguity was the distance to interface for lateral levels. If specifications had to be derived in that case for equipment having a known location, it can be the following:
Distance z < 1 m 6 g
Distance z ≥ 1 m 7 g × z
In fact, extrapolation above 3 m, and even 2 m, can be pessimistic: the available results show limited levels above 2 m which can be a direct effect of notching. The levels at large distances from interface strongly depend on the design of the payload and on its notching philosophy.
A.1.4 References
A.1.4.1 M. MAAGER “Summary on Derivation of Mechanical Test Level (qualification) for General Environmental Test Spec-ification for Spacecraft Equipment”
MBB/ERNO -- AIV STUDY -- July 26, 1984.
A.1.4.2 A. GIRARD “Mechanical Test Data Evaluation for Environ-mental Test Specification -- Final Report”
INTESPACE N.T. 86.596/EI/ET, April 18, 1986.
A.1.4.3 A. GIRARD, D. MOREAU “Derivation of Satellite Equip-ment Test Specification from Vibration and Acoustic Test Data”
ESA Journal 1986, Vol. 10--3.
A.1.4.4 A. GIRARD, D. MOREAU “Elaboration of a Continuous Function of the Unit Mass for Vibration Testing”
ESA Journal 1987, Vol. 11--4/1988, Vol. 12--1.
A.1.4.5 A. GIRARD, P.E. DUPUIS “Improvement and Analysis of the Mechanical Test Database at INTESPACE -- Final Re-port”
INTESPACE DO 94.186 EI/ED, December 16, 1994.