Takata タカタ Airbag Failure Modes, Effects, Causes, and Root Causes - Humidity and Moisture

The overall Takata タカタ airbag recall effort baffles me. I have only been able to pick out bits and pieces of the specific defects from the news. Different articles seem to focus on different problems. It is confusing. I want to make sense of it all and develop actionable PFMEA items.
As I noted in April 2013, Takata stated that the airbag problem is caused by:

1) excess humidity was allowed into the propellant wafers.  I am not sure if this occurred during manufacture or in the warehouse.
2) less dense propellant wafers.  I am not sure if the low density is only caused by poor machine capability, but now Takata states that there was an auto reject AR that was able to detect poor compaction, and therefore low density, but the AR could be turned off by the operator.
The result of both problems is that the when the airbag is triggered the propellant can burn too fast.  If the burn is too fast, the pressure gets too high too fast.  And if the pressure is too high then pieces of the airbag assembly can blow off.  Here is Takata's letter to NHTSA.

It appears that the 1st item, HUMIDITY, could be treated at least two ways.

1st. In manufacturing the HUMIDITY could be built into the canister.

Failure mode: 

1. Moisture level within the propellant wafer is above allowed level
2. Moisture level within the canister is above allowed level

Failure effects: 

Humidity can lead to moisture in the propellant and over time moisture can deteriorate the propellant and lead to quicker combustion, excess pressure during combustion, housing rupture and debris generation, and injury to a vehicle occupant

Potential Causes: 

1. Propellant compound humidity too high
2. Propellant wafer exposed /absorbs humidity (prior to sealing in canister)

3. Humidity/Moisture sealed inside canister with propellant wafer

2nd. In the field the humidity could be a result of environmental isolation failures.  I am picking these items from a NY Times article describing some of the first efforts by Takata to determine what happened in the first incident in 2004. In my mind the split or ruptured housings could be a stand alone cause of the housings breaking apart and generating debris, or the splits/ruptures could allow humidity into the propellant and causes rapid combustion, excess pressure, and the debris generation.

Two of the airbag inflaters Takata had retrieved from the junkyards showed cracks and the start of “rapid disassembly” during the tests, Takata’s preferred term for explosion, according to the two people. They said Takata engineers at the time theorized that a problem with the welding of the inflater’s canister, intended to hold the airbag’s explosives, made its structure vulnerable to splitting and rupturing. The two people said engineers designed prototypes for possible fixes, including a second canister to strengthen the unit.

As described, the weld and canister may have been designed wrong, but I am going to conjecture at the welding process and possible impacts. I do not know what type of weld is used.

Failure mode:
1. Wrong steel tube used in canisters. (Reuters)
2. Part that is not welded completely (Reuters)
3. Housing damage from weld operation. (damage needs better definition)
Failure effects: 

1.Wrong steel can allow incomplete welding or housing to split/rupture over time in the field, allowing humidity into the propellant. Moisture in the propellant leads to rapid combustion, excess pressure, and the debris generation, and injury to a vehicle occupant.

2. A missing or incomplete weld can allow humidity into the propellant. Moisture in the propellant leads to rapid combustion, excess pressure, and the debris generation. (or incomplete welds could cause other effects such as misdirected inflator gas and bad airbag deployment.  Or maybe a missing weld could cause the housing to fragment without the humidity effect)
Potential causes:

1. Wrong steel: Similar steel used in the plant, poor material handling process and confirmation

2. Incomplete welding: cycle interrupt, start up piece, PM piece, etc...  lots of causes

3. Without knowing the welding used...  It is hard for me to say what damage the welding could do to cause the housing to split/rupture in the field (assuming the split/rupture was not there from the beginning)

Possible controls (prevention and detection)

Process Control Prevention

• Compound Moisture Content: dehumidifier, start up process, time
• Propellant Wafer Moisture Content: wafer forming work cell is sealed and humidity controlled(?). At this point I am thinking the whole workshop needs to be humidity controlled.  There might be a wafer cure or dry process.
• Moisture Content within the canister: Assuming the wafer has the correct moisture content, you would need to control the sealed canister assembly and weld work cells or have a humidity controlled workshop.
• Canister Steel Tube: I will assume the correct steel was specified and that the plant process allowed the wrong tube to be selected from stock and used in production. Preventions here would include, ideally, not have two similar tubes with two types of materials.  Either commonize or change the tube designs so they cannot be interchanged. Short term preventions could include bar code tube stock and bar code read before usage on the line.  Simple manual confirmation won't suffice.
• Incomplete Welds: Welding is a robust process in many ways, but has many ways the process can fail. Preventions range from weld tip maintenance, parameter controls, part preparations...  there are too many to mention and would need to be detailed for the machine and weld type.
• Welding Damage to Canister/Housing: Parameter control.  Repair control. This item needs more information.

Process Control Detection

• Compound Humidity: humidity sensor checks  or 100%
• Wafer Moisture: Wafer moisture audit.
• Canister Moisture: Audit?
• Correct Steel: I do not know what the difference is between the two steels at the Takata Monclova plant are.  There are some tests that can be done on the finished part.  The best plan would be to prevent the chance of mixed steel tubes.
• Incomplete Weld: Usually for welds there is a pull requirement.  But, that is just an audit. Also, cross sectioning helps confirm the parameters. But again, it is an audit and usually just shows what the normal weld looks like.  For the purpose of incomplete weld leading to moisture in the propellant you need a good leak test. (A leak test  would pass a weak weld that might crack later in the field)
• Housing Damage: I am not sure about the details here.

Looking at the range of failure modes and causes and the possible controls it seems to me the process could have been far more robust in terms of controlling moisture getting into the propellant.  I would love to see the PFMEA from 2000 and how it evolved over the years.

The Reuters article also reported:

Before June, the prior recalls were linked to problems in the way that the explosive propellant packed into Takata's air bag inflators had been handled between 2000 and 2002, not issues with the inflator now under review by NHTSA. Between June and August, Honda and General Motors (GM.N) recalled another 96,000 vehicles for a separate defect after determining Takata workers at the Monclova plant had put the wrong part into some driver's side inflators.
That defect came to light after GM was sued by a Georgia woman who said a Takata air bag in her Chevy Cruze hit her with such force in a minor accident in October 2013 that it left her blind in one eye. 

In April 2011, Apud told other Takata supervisors that chewing gum had been found in an inflator, one of what he called several "grave problems" in inflator production at the Monclova plant.

The Takata saga leaves me frustrated. Looking in from the outside it seems the problems could have been prevented, and then once they were missed they could have been robustly fixed.  Missed opportunity after missed opportunity.

PFMEA for Receiving Inspection

One of the rules of PFMEA is that an operation should consider the incoming component as good. A way to handle potential defective incoming material is to use the "receiving inspection" operation for the purpose of screening for failure modes and defects that can affect subsequent operations. The receiving inspection operation is already included in most PFMEAs. Sadly, the opportunity is almost always squandered.

Typically the entries are:

Potential Failure Mode (PFM): 

• Part damaged in shipping,
• Damaged packaging
• Dirty
• Defective

Be Specific About Failure Modes of Concern

Failure modes such as "dirty" and "defective" are too general to be useful. Sometimes you do find more specific PFMs:

• Part length too long or too short.
• Out of round

This is more specific, but focused on specific defects and not on potential failure modes.  More can be done to use the PFMEA beneficially to improve receiving inspection..

Include Failure Modes of High Risk to Your Operation

Why not use receiving inspection to actively question incoming material for ways it may affect subsequent operations? Try to use your Process Flow Diagram and PFMEA development to populate the receiving inspection operation with specific defects that can be a concern to your plant.

For example, consider an incoming die casting with sealing surface that your plant assembles with a seal. In terms of PFMEA you might include:

• PFM: Sealing surface does not seal
• Potential Effects of Failures (PEF): Leak fail at leak test
• Potential Cause or Mechanism of Failure (PCMF):
• Seal groove too rough
• Seal groove depth too deep or too shallow
• Sharp edge or missing chamfer

Using the PFMEA to focus in on areas of concern, like leak test failures, could help you populate receiving inspection with thing to spot check. In this case you might want to have receiving inspection spot check the sealing surface for roughness, grove depth, and sharp edges.
Another source of ideas are past supplier defect report (or 8Ds, or PRR, etc).

Include Failure Modes of High Risk to Your Customer

Another type of PFM that can be included in the receiving inspection operation are those related to component features that are not processed, used, or tested inside your plant. You might call these features "customer features" or "pass through features". An example of a customer feature might be a tapped hole on a machined housing that your plant assembles into a compressor.  Your plant does not assemble a bolt into the hole, your customer does.  If that tapped hole is defective for some reason, then your plant is responsible.  You can roll back the pain to your housing supplier eventually. But, everyone loses. You do the PFMEA exercise for these items in your receiving inspection section:

• Potential Failure Mode (PFM): Customer could not drive their bolt
• Potential Effects of Failures (PEF): Rejects at customer, containment, scrap costs
• Potential Cause or Mechanism of Failure (PCMF):
• Tap wrong size
• Tap too shallow
• Debris in tapped hole

Your controls section could include:

• Run a thread gauge for tap size
• Also check tap depth.
• Spot check for machining debris or swarf or turnings, etc

Example:

Take a simple example of an incoming wooden handle for a hammer.

Some typical incorrect Potential Failure Mode (PFMs) might be:

• Wood is dirty
• Wood is scratched
• Wood is wet

Some typical Potential Effects of Failures (PEF):

• Rejected hammer
• Bad handle attachment

A typical PCMF: 

• Bad receiving inspection

These PFMs are pretty much what you would come up with from the top of your head. There is no real additional value in listing them out. But what if we consider how the incoming material could lead to a failed operation? This is a situation where it makes sense to switch from specific defects to general PFMs. We want to use the PFMEA process to probe for new PFMs and PCMFs (in the case of receiving inspection, these will be incoming defects) that could nip us in the butt.


Here are a few examples of how this might work:

• PFM: Wood has some defect that prevents varnish adhesion
• PEF: If varnish does not adhere to handle it will flake off in the field
• PCMF: (need to consider defects that would cause poor adhesion of varnish)
• Wet wood? 
• Oily wood? 
• Rough texture?

Controls (detection) might be:

• Visual inspection for wet or oily wood.  No! typical and useless.
• You could do a humidity test with a sensor. (for water) 
• Maybe do a water bead test. (for oily)

Detection vs Prevention

By now you may be asking, why are we laying all of this work on receiving inspection?

True, what we have laid out are all detection type controls.  You really want to focus on prevention controls.

You can consider that we may be robust during launch with heavy receiving inspection activity.  And, leading up to launch the specific defects should be driven back to purchasing and supplier quality to confirm the suppliers have these defects in their PFMEA failure mode sections with good controls.

Over time the receiving inspection activity can be reduced based on data.

To reduce the burden on receiving inspection some of the defects could potentially built into your own process:

 As a poke yoke to detect incoming defects. Example, a pin to detect a machined hole.

 As potential failure modes in our own plant in operations that might can CAUSE the defect.  Example, damaging the tapped hole in your plant with an alignment pin or fixture. Or with a varnish or coating operation.

Summary:

• Be Specific About Failure Modes of Concern
• Include Failure Modes of High Risk to Your Operation
• Include Failure Modes of High Risk to Your Customer
• Continually Drive for Prevention at Your Supplier.

PFMEA

The PFMEA is one of the most useful yet underutilized tools for analysing a process to find potential problems and assign controls. Do you find that the failure modes for quality problems we find at the customer are usually well known and not documented in the PFMEA or not well controlled? Any comments about effective PFMEA usage?

QualityTrainingPortal has a description of the PFMEA.

What It Is Used For
To methodically examine a process or product design to identify where failures can occur and what the relative risks are for each mode of failure.
When to Use It
Start using this tool during the design stage for any process or product.
On existing products and processes, this tool can be used at any time as part of improvement efforts.
Important Notes
FMEAs are a complex undertaking and should always be conducted by a team.
FMEAs need to be updated whenever changes are made to the process or product.