Essais & Simulations n°122

dossier automobile

dossier automobile process will yield the widest possible margin between UUT capabilities and the environment in which it will operate, thus increasing the product’s reliability, reducing the number of field returns and realizing long-term savings. HALT isused to supplement Qualification testing – not replace it. HALT test is a “qualitative” test, not a “quantitative” test as an “Accelerated test”. An “Accelerated test” must apply astress load which can precipitate afailure as in the field. HASS [1] The principle ofthe Highly Accelerated Environmental Stress Screening (HA- ESS) is to submit equipments coming out of production to levels of constraint far above the specified values while staying below the destructive limits which could have been revealed by a campaign of Highly Accelerated Tests The most evident objective of screening is to discover and precipitate any weaknesses in the product; the consequences are to improve -the operational reliability of the population of similar (same definition) items by precipitating latent failures and removing the weak items -the manufacturing processes and eliminate residual design defects >HALT and HASS main objectives are to: • Reveal more quickly the latent defects • Detect more quickly the process faults production • Identify the defects that may not have been revealed by a «conventional ESS” operation • Reinforce the product maturity and robustness 3. NewMethodoloyproposed This methodology proposes acriteria to decide if aprecipitated failure during HALT isrelevant (or not) to achange of Design. Step #1: calculation of the absolute FDS of the HALT repetitive shock stimuli Figure 1. Synopsis of the methodology After the observation of failure during HALT test, a new measurement of the component must be done at lower stress. With a Rainflow counting, a Basquin’s law can be used to assess the Fatigue Damage Spectrum (FDS HALT) Step #2: calculation of the absolute FDS of the life profile sequence to which is expected to be applied to the product Based on Life Profile specification, like aPower Spectrum Density (PSD in g 2 /Hz) representative of the vehicle stress, aFatigue Damage Spectrum is identified (FDS Lp). Step #3: comparison of the 2above FDS, in the range of frequencies concerned by the precipitated defect. Obtain the ratio of damage between HALT test (DHALT) and damage of Life Profile (DLp) focus in the critical frequency bandwidth. Step #4: application of criteria to conclude on the relevance of the failure to be corrected. Compare the ratio in the step #3 with the criteria: Variability of the Environment (ie: uncertainties of the measurement) called Env multiply by Variability of the Product (ie: various strength from the process) called Prod. This scalar is with the exponent of Basquin’s b coefficient. If the ratio in step #3 is higher of the criteria, the failure seen in HALT test is not arisk for the vehicle (no design change). In the opposite conclusion, the risk for the vehicle seems to be relevant and adesign change is necessary. 4. Concretecase study Figure 2. Hammers in HALT bench Essais & Simulations • SEptEmbrE 2015 • pAGE 46

dossier automobile An electronic equipment has been submitted to aHALT vibration stress test. The corresponding stimulation is the result of repetitive Shock Vibration, a vibration originating from a repeated shock impulse excitation, typically created from pneumatic hammers impacting avibration table to which the Unit Under Test is attached. Several points were measured on PCB card and components by accelerometers sensors and by a3Dvibrometer laser. During aprevious HALT test, a failure appeared on the capacitor at 45 gRMS. Figure 3. HALT test with sensors measurement The question is the same: “Do we have to change the Design of the product regarding the vehicle risk?” a. HALT procedure The standard HALT procedure [2] recommends that Vibration step stress begins at achamber set point of 1to 10 Grms, as measured over a2Hz to 2000Hz or greater bandwidth (5 Grms recommended) and increases in 1to 10 Grms increments (5 Grms recommended) upon completion of the dwell period and subsequent functional test The dwell time at each level of vibration is aminimum of ten minutes. Functional testing is performed at the conclusion of the 10 minutes dwell period, thus the total dwell at each level of vibration of time it takes to run one cycle of functional testing on the sample. Note that the functional test may be, and is recommended to be performed throughout the step, however it must minimally be performed at the conclusion of the 10 minutes dwell. The vibration step stress is continued until the operational limit of the sample The vibration step stress is continued until the operational limit is determined or the chamber maximum is achieved. b. Particularities of this repetitive shock vibration The particularities of this repetitive Shock Vibration stimulation are: 1. Mechanical excitation on a given point of the testing table (receiving the UUT) is the result of synchronized impact on all the pneumatic hammers. So, the relation of the effect (vibration on agiven point) and the cause (the impact at the different points of impact) is not biunivocal: the effect is the result of the nhammers .Therefore, if we compute the Coherence function between aresponse measured on the UUT and apoint on the table representing the input, it will be very far from the ideal value of 1. The consequence of that is the difficulty to characterize atransfer function. AFatigue Damage Spectrum (FDS) will be more appropriate to characterize the range of frequencies where there is adynamic work between an input and an output. 2. The repetitive shock stimuli, represented by the FDS is very similar in the 3axis; the supplier ofthe HALT machine claims that the rotations around each of the 3axis are also very significant and are participating to generate a thorough UUT stimulation. 3. When increasing the RMS value of overall acceleration, all the frequencies are concerned by the growth and the FDS is uniformly translated. 4. The distribution of the instantaneous acceleration values is not following a Normal law (also called Gaussian Law). One shape parameter of the distribution is given by the Kurtosis, that is ameasure of the flatness of the probability distribution of areal-valued random variable about its mean. Figure 4. Kurtosis [3] For aGaussian distribution, the Kurtosis should be lower than 3. Above a kurtosis of 3, we consider ahigh transient phenomenon (important shock in the signal). Figure 5. Kurtosis values on different measurement points. The figure 5 shows the evolution of the Kurtosis (not centered) along the 20 seconds of measured signal for 3 different points of measurement: it is very high for certain points ,approaching several tenths. This observation confirms asignal highly shocked (far from Gaussian signal). c. Fatigue Damage Approach As explained before, the nature of the signal from HALT test does not allow us to use asimple Fourier Transform. We propose to calculate the absolute Fatigue Damage Spectrum corresponding to this excitation. The French Standard NFX-50144 [4] explain in details the formulation of the Fatigue Damage Spectrum. For example, the graph hereafter present aFDS graph described by Bonato and Delaux’s paper [5]: Figure 6. Basquin’s law representation The FDS curve can be interpreted only in the critical frequency bandwidth. In our study case, we use the bandwidth 250-400 Hz as asupposed band closed to the natural frequencies of the capacitor. Essais & Simulations • SEptEmbrE 2015 • pAGE 47