Why NHTSA Should Have Long Ago Thrown Out GM’s Petition

by Kevin Fitzgerald & David Schumann on August 16, 2018

On January 11, 2018 General Motors conducted a test of a Takata SPI-YP GEN 1 passenger inflator from a 2007 Chevrolet Silverado 2500 that had spent its life in Texas’ withering sun and suffocating humidity. The inflator was deployed under laboratory conditions at ambient temperature and reached an internal pressure of over 90 MPa (13,000 psi). It did not explode, but when compared to rupture pressures of the same structure reported by Orbital ATK, it was teetering on the edge and if it were tested at the high end of its required operating temperature, as it should have been, it likely would have crossed the threshold. (See Figure 1)

The SPI-YP GEN 1 single stage passenger inflator was Takata’s first with Paresh Khandhadia’s phase stabilized ammonium nitrate (PSAN) propellant. All the airbags with this inflator are under recall but GM has filed three petitions of inconsequentiality with NHTSA to avoid bringing back airbags with Takata’s GEN 2 version of the same inflator and its dual stage partner, PSPI-L YD. GM argues these inflators are immune to rupturing, 6.8 million of them, or better put $1 billion. They don’t deny the propellant is degrading over time in the presence of heat and moisture, in fact they concede this, but claim their inflators are unique and don’t allow the degraded propellant to ramp to rupture pressure. This is utter nonsense and here’s why.

SPI-YP GEN 1 VS. GEN 2 Design Differences

Let’s continue with the SPI-YP discussion, starting with the major GEN 1 versus GEN 2 design differences. The PSPI-L YD will be addressed later. First it is imperative to understand the changes were made solely to address performance variability which surfaced during USCAR drop test evaluations (3-foot free fall onto a steel plate). The propellant wafers were being damaged by the drop shock, adding to the inflator’s already insufferable variation. Pushed into production regardless, GM wanted the issue fixed and to do this Takata added a shock absorbing system at each end of the propellant wafer stack. The GEN 1 design only had a wave spring opposite the igniter end to serve this function, but the spring reached solid height upon impact, transferring the shock directly to the wafers. Ceramic cushions stopped this but were shredded by vibration. The combination of the two, placing a cushion inside the spring, worked and one of these subassemblies was positioned at each end of the wafer stack. The nominal 2004 tablet load was dropped by 1.25 grams to accommodate the reduction in free volume caused by the additional components, a simple function of the ideal gas law.

There was an additional change made, not highlighted by GM, that improved manufacturability of the propellant tube designed to protect the 2004 tablets from the filter’s rough interior during vibration. The right circular cylinder, sealed on both ends with adhesive backed tape, was replaced with two stamped metal parts and a single piece of tape. This modification also served to initially throttle the tablet load to improve its contribution to the ignition event. Nothing was done to address propellant degradation and neither version contains desiccant. Figures 2 and 3 highlight the design differences

Why the Design Differences Don’t Matter

Now that we understand the changes to the SPI-YP GEN 2 we can delve into why they don’t serve to prevent further runaways. We’ll start with the shock absorbers. The following statement is lifted directly from GM’s Update on GMT900 Takata passenger airbag testing, dated February 12, 2018. “While originally made for drop test performance, GM believes the GEN 2 design changes also improve the long-term durability of the SPI YP inflator.” This is the extent of GM’s argument and it is not supported by any data. Engineers don’t say, ‘We believe’ when it comes to life and death matters. Drop tests are to simulate an operator fumbling an inflator during the installation process, nothing else. There are millions of Takata inflators installed in vehicles with a single wave spring and no cushions that attenuate road vibration effectively. This baseless statement merits no further discussion.

What does that leave GM with to underpin their case? A 1.25 gram reduction in the nominal 2004 tablet load (GEN 1 and GEN 2 tolerances overlap!). The function of the 3100 booster material is to amplify the output of the initiator and begin pressurization of the inflator’s internal free volume. The propellant wafers need both a spark and high pressure to begin combustion. The 2004 tablets serve to further amplify the booster’s output in the process and account for only 6% of the total main chamber load. When the additional spring and cushions were added to the GEN 2 it reduced the internal free volume requiring pressurization. The ideal gas law, PV=nRT, explains why the tablet load was lowered.

When the equation is restated as P=(nRT)/V it is clear as volume is reduced pressure rises. In order to maintain identical restraint performance between the two versions, mass flow had to be matched. The GEN 2 tablet load (n) was accordingly decreased and throttled to provide the same initial pressurization (P) as GEN 1. It’s unclear if GM understands this basic science but since they ‘believe’ the reduced tablet load prevents GEN 2 runaways, they must be educated. They stake their claim on post accelerated aging data of GEN2 inflators.

SPI-YP Orbital ATK Aging Study Discussion

Orbital ATK conducted the long term aging studies as part of an independent investigation into the airbag failures and found that Takata’s inflators are adversely affected by three factors – all of which must be present to cause a rupture.

1. The presence of PSAN propellant without desiccant.
2. Long term exposure to repeated high temperature cycling in the presence of moisture.
3. An inflator that does not adequately prevent moisture intrusion under conditions of high humidity.

While we agree with the majority of this assessment, we take exception to the first factor. No data had been presented with desiccated inflators. In fact, field returns of Takata’s PSDI-5 dual stage driver inflator with calcium sulfate desiccant have already exhibited negatively trending propellant densities, convincing NHTSA to recall all of them. The jury is still out on Takata’s second generation desiccant, Zeolite, but the hard cold facts are that all inflators leak and all desiccants saturate. In the case of Takata’s PSAN inflators, desiccant will only serve to prolong the inevitable. Orbital’s study does not address desiccated inflators and truthfully it wouldn’t have mattered because the experiment is so flawed. That may sound bold but hear us out.

There is so much superfluous information in GM’s presentation of Orbital’s data that it would require a tome to refute point by point. This was clearly done to overwhelm an understaffed and adrift NHTSA that doesn’t have the technical competency or will to sort through the obfuscation, especially in today’s climate of aggressive deregulation. We will stick to what matters. All of the predicted aging models presented by GM claiming their inflators will be safe for the next 30 years were built with data from the aging study. If that work is flawed, so are the predictive models. Dismantle the study and the rest falls apart.

Orbital’s results open with an overview of the post aging deployment data in an easy to read table (Figure 4). Data is also included for PSPI which will be used later.

Across the top of the chart are the number of temperature cycles the inflators were exposed to and directly below that, the maximum temperature of the cycle. Low medium and high refer to the moisture levels added to the inflator and correspond to a nominal build (only assembly moisture), 0.15% and 0.30% of the total propellant load respectively. So, if we look at the last column, six independent inflator sets endured 1960 cycles, one from ambient to 50C, one from ambient to 60C and the last from ambient to 70C and each of these groups had three moisture level subsets. The cycles were conducted absent the presence of external moisture. When we first glanced at the chart, we assumed the 50C, 60C, and 70C were deployment temperatures which concerned us since the USCAR requirement is 85C. Investigating further, we found they were the cycling temperature ceilings, concerning us even more. USACR specifies temperature cycling between -40C to 85C. If the point of the study was to be certain the inflators would never harm anyone, why weren’t they stressed to the temperature extremes, and now equally important, what was the deployment temperature? That could not be found anywhere in the report. Details of the cycle ramp rates and how the moisture was added were also missing. Did the experiment introduce the moisture all at once or was it metered over time like real world? We found ourselves immediately questioning the experiment. The table may have been lit up green, but what we saw were red, flashing warning lights.

To discredit the aging study, we will analyze the post aging tank deployment data and wafer physical characteristics (diameter and density). To measure performance, inflators are tested in a 60L closed volume, or tank, recording pressure vs time. Figures 5 and 6 present this data for the SPI-YP GEN 2 after 1680 and 1960 cycles respectively while Figures 7 and 8 provide the wafer diameter and density as the cycles progress to 1960.

They say a picture is worth a thousand words. These are worth a billion dollars. Recall the SPI-YP is a single stage inflator. It has an output range of 353 kPa to 446 kPa at ambient temperature and at high temperature the range is between 407 kPa and 504 kPa. Since both GM and Orbital are silent on the deployment temperature, it can be deduced from the 1680 cycle nominal build data (no added moisture) they were deployed at ambient. If we are to believe the study, as more moisture is added, the wafers grow in diameter, their density plummets beyond dangerous levels, yet the inflator’s output suppresses. While the swelling and corresponding density drop are typical of PSAN wafers subjected to these conditions, output suppression is not. It is typical of the 3110 booster material and that is what is occurring.

Why are we confident of this? Observe the 1680 and 1960 cycle, high moisture deployments, particularly the 60C and 70C sets. Time to fist gas has almost doubled and they are only reaching half of their required output. This suggests if NHTSA were to allow GM to leave these inflators in the field indefinitely they would just stop working altogether. How many reports have there been of GMT-900 cushions not even exiting the cover? None, but Takata has reams of data that shows as the 3110 booster saturates it becomes inert. Figure 9 presents the moisture pickup in the igniter closure 3110 and main chamber 2004 propellants as the cycles progress.

The moisture levels in the 3110 material are much greater than 2004 which is typical of Takata field returns. It is the better desiccant of the two. Almost all the high moisture inflators failed to light properly, indicative of a saturated booster that failed to pressurize the main chamber adequately to bootstrap combustion. The inflators cut out, never stressing the swollen, low density wafers and suggesting the green, “nominal” indicator in the tank testing summary means nothing more than they didn’t blow up! This phenomenon may be a function of how the moisture was introduced, whether all at once or metered, but it is not real world and could have been easily answered by deploying the inflators at elevated temperature. Orbital’s own data should have tipped them off to this.

The following paragraphs are lifted directly from Orbital’s Takata Inflator Root Cause Summary Report, dated September 2016:

Moisture transport inside an inflator follows expected behavior of the hydroscopic, desiccating and deliquescing behaviors that would be expected for these materials based on general principles of chemistry. That is, sodium bentonite, which is a component of both the 2004 and 3110 propellants, can function as a drying agent and PSAN can absorb significant amounts of moisture if available. There are not dramatic cliffs but rather gradual changes as a function of temperature, as shown in the experiments conducted by Fraunhofer ITC and confirmed by our efforts. However, these cumulative changes over the temperature range to which the inflator is exposed appear to be significant.

In each case, the change in the amount of water that is available is dramatic when transitioning from 30°C to 70°C. This suggests transport is much more likely at the higher end of the range. The 3110 propellant shows a modest increase in propensity to give up moisture compared to 2004. This variability gives rise to the “x-graph” when suitably drawn although the effect is not as dramatic as suggested in some depictions.

Figure 10 presents a graphical representation of the headspace moisture above 2004 and 3110 propellants with moisture added as a function of temperature. In layman’s terms, headspace is the air or empty space left above the contents in a sealed container. More simply put, this is a representation of how each propellant liberates its bound up moisture as temperature increases.

And there it is. The 3110 booster shows a modest increase in propensity to give up its moisture, especially at the higher end of the inflator’s required operating temperature. Conditioning the inflator to 85C before deployment, would have liberated a portion of the 3110’s moisture, drying and reactivating it. If Orbital had tested the aged inflators at elevated temperatures we contend the effect of the booster would have been gravely different and the swollen, low density wafers would have ramped to rupture. We do not agree the inflator’s output suppresses over time as moisture is added and vehemently object the phenomenon being assigned to 1.25 grams less of 2004 propellant. Again, unlike the SPI-YP GEN 1 inflator, the GEN 2 2004 tablets are housed in their own pressure vessel that is initially throttled and they have less initial free volume to help pressurize. It simply does not compute. The booster was disarmed.

Another way to demonstrate the futility of the study is to examine the results of the nominal build aged inflators (no added moisture) where the fire wasn’t doused by water. Recall what may have seemed like an innocuous statement made earlier. All the aging was performed in the absence of external moisture, including the inflators with no moisture added. Why cycle a dry inflator from ambient to 70C? We don’t understand this but are thankful for the data as it builds an even stronger case to get these inflators off the road and now. Figure 11 plots the SPI-YP GEN 1 near field rupture against the 1960 cycle results.

By Orbital’s own admission a nominal build inflator exhibited the same characteristic curve shape and time shift as the near rupture. It was on its way and simply from moderate temperature cycling. Remember the anomalous inflator was returned from Texas where it experienced temperature swings in a humid environment. The aged nominal inflators suggest the moisture sealed in the device during assembly and being transported back and forth between the propellants is enough on its own to become a problem over time. In fact, after 1960 cycles the nominal build wafer densities were as bad as those where moisture was intentionally added (Figure 8). Full stop. Based on the GEN 1 near rupture and this data point, GM’s petitions of inconsequentiality should have never been submitted and should be rejected immediately.

GEN 1 vs. GEN 2 Wafer Stack Spring Load Comparison

We will close the SPI-YP discussion with a comparison of the GEN 1 vs. GEN 2 spring working load on the wafer stack. GM states that the addition of the second spring increased this load by 30%. It is not clear if this is being sated as a benefit, but in this application, it is a detriment. Figure 12 shows the wafer O.D. growth vs. years in the field for both GEN 1 and GEN 2. A limited informational recall must have been ordered to obtain the GEN 2 data.

Without knowing the climatic zones these inflators are from, it’s hard to draw too many conclusions but what is certain is the wafers grow in diameter over time, and this has to be influenced by the load they are constantly under. There are GEN 2 inflators reaching the same wafer O.D. in less time, recording a maximum diameter close to 29.4 mm. If we refer back to Figure 7, the aging study produced wafer diameters as large as 29.9 mm. Why is this important? Orbital’s Root Cause Summary Report attributes the inflator ruptures to erosive burning of the wafers.

The 2004 propellant wafers in passenger inflators grow after repeated temperature cycling with moisture present (transitioning in and out of propellant grains). This effect was observed in both wafers and tablets in primary and secondary chambers of passenger inflators. The growth results in reduced envelope density but very little change in pycnometry density, suggesting that the increased volume is void spaces that are connected. These connected pores, flaws or fissures allow hot gas penetration resulting in increased mass flow when ignited (porous and permeable burning).

When gas is allowed to penetrate into the wafers through fissures and connect with void spaces bad things happen. As the wafers continue to grow in diameter they are forced into the filter, a continuous wrap of pierced sheet metal. Often the pattern of the diamond shaped piercings can be observed carved into the outside diameter of aged wafers. A higher spring force will have a greater propensity for this to happen, creating the fissures necessary for erosive burning. It’s just a bad idea.

PSPI-YD Discussion

GM’s arguments to avoid recalling the dual stage PSPI-YD inflator are even more perplexing than their SPI-YP GEN 2 assertions. This time they base their case on the design differences between it and the PSPI-FD used in their Pontiac Vibe which has the highest propensity of any passenger inflator to rupture and their prediction it is immune from failure again relies on Orbital’s flawed aging study. The same logic used to discredit the SPI-YP claims will be applied here.

PSPI-YD vs PSPI-FD Design Differences

Figure 13 presents the field rupture rate of the PSPI-FD inflator while Figure 14 is a representation of how it differs from GM’s PSPI-YD.

The PSPI-YP’s primary chamber uses predominantly 8.1 gram wafers because the vehicle crash pulse it was designed for required a higher initial mass flow rate than the PSPI-FD’s thicker 10.8 gram wafers could provide. The thinner wafers present a greater initial propellant surface area which generates gas at a faster rate. This requires larger vents to manage the inflator’s internal peak pressure or said otherwise, its structural margin. The combination of a higher gas generation rate and larger vents has the effect of releasing gas into the airbag faster. This was done simply for performance reasons, but GM claims the PSPI-YD’s thinner wafers and increased vent area compared to the PSPI-FD are what prevents it from rupturing. With 200 recorded PSPI-FD field return failures, this argument should be bullet proof. Instead, it is riddled with holes.

PSPI-YD Orbital ATK Aging Study Discussion

How did GM come to this conclusion? Orbital’s aging study. Figures 15 and 16 present the PSPI-FD’s tank data after 1680 and 1960 cycles respectively while Figures 17 and 18 do the same for PSPI-YD. The moisture levels applied to the both inflator’s primary chamber matched the SPI-YP experiment, however the secondary chamber levels were raised to 0.45% and 0.75% based on field return measurements.

After overcoming the horror of the aged PSPI-FD results, the only conclusion that can be drawn is GM better be moving heaven and earth to get every one of them off the road. There are primary and secondary chamber nominal build (no moisture added) ruptures after only 1680 moderate temperature cycles demonstrating once again the moisture sealed in the device during assembly and being transported back and forth between the propellants is enough on its own to become a problem. According to the NHTSA website there are still 246,000 PSPI-FD inflators on our roads and that simply unacceptable. They present a clear and present danger and GM should be focused on remedying this situation rather than trying to avoid additional recalls.

Now let’s analyze the PSPI-YD post aging tank results. Why GM points to its thinner wafers and larger vent area as mechanisms to prevent the same behavior observed with PSPI-FD is simply dumbfounding. Compare the PSPI-YD primary chamber to the SPI-YP and the argument falls to pieces. The SPI-YP inflator has nine 8.1 gram wafers for a total load of 73 grams with a vent area of 42.5 mm2. The PSPI-YD has eight 8.1 gram wafers and two 5 gram wafers for a total load of 75 grams and a vent area of only 36.2 mm2. Is anybody else scratching their heads at this point? The PSPI-YD has more total surface area due to the 5 gram wafers, but 15% less vent area. If we follow GM’s logic, the SPI-YP inflator should be far safer than the PSPI-YD, yet it has already demonstrated a near rupture after 10 years in the field. Furthermore, a nominal build 1960 cycle PSPI-YD demonstrated the same characteristic curve shape and time shift as the SPI-YP near rupture with a tank pressure 25% greater than the rest of the population. Another one well on its way after only moderate temperature cycling in the absence of external moisture. Understanding this, there really is no need to go further, but the aged PSPI-YD wafer densities are so awful they must be discussed. Before that though, it’s important to note the 1960 cycle, high moisture deployments are displaying a similar loss of ignition energy as the aged SPI-YP inflators, evidenced by their time shift in peak combustion pressure. This is again a clear indicator the aged inflators should have been tested at high temperature to gain a thorough understanding of the plummeting wafer densities.

Figures 19 presents predicted internal peak combustion pressure as a function of wafer density for the PSPI-FD, PSPI-YD and the SPI-YP inflators.

These are not empirical data points. They are derived from an uncalibrated ballistic model, yet are used to make the shocking assertion that the wafer rupture density threshold is between 1.61 -1.58 g/cc. The theoretical maximum density of a wafer is 1.70 g/cc and a nearly 10% reduction from that is simply intolerable. No predicted internal pressures are provided for PSPI-YD wafer densities between 1.67 g/cc -1.58 g/cc and it is well understood within Takata, based on empirical data, that anything below 1.65 g/cc is on the highway to the danger zone. These kind of assumptions are extremely disturbing, especially when calculated from a simulation that Orbital itself concedes is supported with only minimal augmented wafer data. The rupture density threshold is 1.65 g/cc.

Now let’s examine the post aging wafer densities. Figure 20 provides the PSPI-FD and PSPI-YD primary chamber densities out to 1960 cycles while Figures 21 does the same for the secondary chambers.

The first thing that stands out is the nominal build primary chamber densities are worse than any with moisture added. Are we now to believe the addition of moisture suppresses density loss. This is ridiculous and once again throws into question the entire study, but let’s set that aside. There is a nominal build PSPI-YD wafer density reported at Orbital’s own softball threshold and data in all three moisture subsets below the real world threshold of 1.65 g/cc. Wafer densities are presented that would undoubtedly lead to a rupture and worse, the experiment suggests no external moisture is necessary to lead to disaster.

There are no words to describe how terrible the PSPI-YD secondary chamber wafer densities are. There is a data point as low as 1.56 g/cc and densities in all three moisture subsets below Orbital’s softball threshold. These are guaranteed ruptures. Why weren’t the deployments conducted at high temperature? It is incomprehensible that a petition to avoid recalling these inflators was submitted.

SPI-DH Field Rupture & Booster Loads

We will close the PSPI-YP discussion with two critical and related points. Firstly, Orbital’s root cause analysis report identifies a rupture of an SPI-DH inflator in Florida. Like the SPI-YP and PSPI- YD primary chamber, the SPI-DH uses 8.1 gram wafers. Why is there no mention of this and what sets this inflator apart from the SPI-YP and PSPI-YD? This leads us to our last point. Why is there no discussion regarding the PSPI-YD booster load? The entire SPI-YP GEN 2 argument hinges on it having a lower 3110 load than the GEN 1. What is the nominal booster load of the PSPI-YD inflator and now, equally as important, the SPI-DH that ruptured in Florida? If their booster loads are higher than the SPI-YP inflator, it is yet another dagger in GM’s arguments, especially considering the PSPI-YD primary chamber has 15% less vent area than SPI-YP. This items are glaring omissions.

Final Remarks

This ends our discussion of the SPI-YP and PSPI-YD inflators. We did our best to analyze the data released to the public, but it is incomplete at best. How the moisture was added and at what temperature the aged inflators were deployed are critical items that were not provided. We ask Orbital ATK and GM to release the full data set for a deeper understanding and if there is an objection to our analysis or incorrect assumptions have been made, let’s have the debate. That’s the least we can do when life or death hangs in the balance.

Below is the list of the GMT-900 vehicles with Takata passenger inflators that remain in recall limbo while NHTSA rules on GM’s petitions. It is noteworthy that the Government of Australia has dismissed GM’s objections and ordered these vehicles remedied. We demand NHTSA do the same!

• 2007-2014 Cadillac Escalade
• 2007-2014 Cadillac Escalade ESV
• 2007-2013 Cadillac Escalade EXT
• 2007-2013 Chevrolet Avalanche
• 2007-2014 Chevrolet Silverado HD
• 2007-2013 Chevrolet Silverado LD
• 2007-2014 Chevrolet Suburban
• 2007-2014 Chevrolet Tahoe
• 2007-2014 GMC Sierra HD
• 2007-2013 GMC Sierra LD
• 2007-2014 GMC Yukon
• 2007-2014 GMC Yukon XL

08/21/2018: 2004 10g wafers updated to 8.1g. 2004 chamber load updated from 5% to 6%