Airbag sensors may not top the list of high-tech electronic components in today’s computerized cars. But from passenger-safety and product-liability standpoints, the sensors that tell an airbag when–and when not–to deploy are among the most critical electronic components in an automobile. By deploying correctly in a catastrophic accident, they can prevent serious injuries or fatalities. By unnecessarily deploying in a minor collision, they may cause life-threatening consequences.
Because the results of error are so drastic, the automotive industry is actively developing more sophisticated techniques for airbag-sensor testing. A number of technologies are converging to raise sensor testing to new levels of accuracy, repeatability, affordability and convenience.
Time-Domain Replication
Historically, vibration control systems have attempted to replicate two types of vibration environments:
The transportation environment which a product experiences during shipment to the dealer or end user.
The operational environment which the product confronts over the course of its expected service life.
The object of the transportation tests was to reduce cases of infant mortality–products found to be dead on arrival. Operational tests aimed to prevent premature product failures. For these classical applications, time-domain replication was less important than replication of the frequency-domain components.
Today, however, time-domain replication is a critical element in several automotive test applications. Airbag-sensor testing and ride-quality system design are two prime examples. A brief comparison of these two applications highlights the technical complexity of airbag-sensor testing.
Ride-quality systems increase passenger comfort in many of the newer luxury cars. These dynamic systems monitor wheel inputs and adjust suspension characteristics in real time to minimize the effects of potholes and other road conditions. To test ride-quality systems on a shaker table, engineers replicate various types of road-surface-induced vibration and measure system response in terms of accuracy and speed.
Airbag-sensor testing on a shaker table also looks at the speed and accuracy of response by an active device to vibration inputs. But where potholes and road corrugations cause frequencies in the 5- to 30-Hz range, the catastrophic crashes which trip airbag sensors generate complex waveforms with frequencies typically measuring from 200 to 1,000 Hz.
Because airbag-sensor testing requires simulation of high-speed crashes, shaker tables used for this application must produce displacements ranging from 10 to 20 inches. Such large-stroke displacements are generally not a requirement for testing ride-quality systems.
Quality Assurance
There are three primary reasons why automakers need highly accurate techniques for testing airbag sensors on a shaker table. The first reason is quality assurance. Since there is little or no tolerance for failure in airbag-sensor applications, manufacturers need a high level of quality assurance. This can be achieved by testing every sensor before installation or by screening selected lots of sensors. In either case, shaker-table tests are an attractive technique for quality assurance because they permit simultaneous testing of multiple sensors at a relatively low cost.
Quality-assurance tests verify whether the sensors trip every time the shaker table simulates an impact of the required severity. The threshold is usually a 12 to 18 mph head-on crash.
It is equally important to make sure that they do not trip in response to lower-speed collisions in which airbag deployment might indirectly endanger the occupants of the vehicle. Quality-assurance tests may also verify that the sensors trip in response to front-end collisions, but do not trip in response to side impacts.
Improving Performance
A second reason for testing involves improving sensor performance. In fine-tuning performance, engineers experiment with different methods and parameters of signal processing.
The goal here is to maximize sensor sensitivity to parameters which should trigger bag deployment. Other types of sensitivity analysis make sure that sensors do not respond to random noise or inputs which should not trigger deployment.
This basic process compares the performance of different types of sensors: displacement sensors vs accelerometers or inertia sensors vs piezo-electric types. It generally aims to improve the reliability and lower the cost of airbag sensors. Since a well-designed sensor system may find its way into millions of cars over a period of years, even a small saving per unit can have a major impact.
Product Liability
Product liability issues are also motivating manufacturers to study and adopt advanced testing techniques. Without delving into the fine print, the general import of liability laws is that manufacturers have an obligation to investigate test techniques that can improve product reliability.
If manufacturers have been exposed to a particular test technology and fail to follow up on their knowledge, they increase their exposure to product litigation. Since liability cases involving airbag-sensor failure can be extremely expensive, manufacturers have a clear economic incentive to implement state-of-the-art test techniques.
Test Techniques
One frequently debated issue in airbag-sensor testing is the acceleration vs deceleration argument. Most tests simulate time histories acquired by crashing a car into a wall at speeds up to 30 mph. Common sense argues that the most accurate way to simulate the event is to accelerate the sensor and then put it through severe deceleration, such as impact with the wall.
At the moment of impact, the sensor is actually traveling at a constant velocity. Its acceleration rate is 0. Consequently, it is possible to emulate the acceleration time history by applying only the deceleration transient.
This debate is ongoing. Some companies believe that the sensor has to be moving to properly establish the vibration environment. Other companies, with an equal concern for quality, believe that the deceleration transient alone is sufficient.
Marginal Cases
With the airbag application, all events must occur within a fixed, and very limited, timing sequence. Depending on vehicle speed at the time of the crash, the sensor system must decide whether or not to deploy the airbag during the first 10 to 40 ms after impact.
Since the data collected within that period is very limited, algorithms take on extreme importance. It does not require a particularly sophisticated algorithm to distinguish between crashes at either end of the spectrum; that is, very high-speed or very low-speed collisions.
The difficulty comes in the marginal cases. It is almost always more dangerous to deploy a bag when it is not needed than not to deploy a bag in a marginal crash. Because life and limb ride on the decision, test engineers devote considerable effort to fine-tuning their algorithms.
Enabling Technologies
Recent developments in hardware and software have combined to give test engineers the ability to measure and adjust airbag-sensor performance with unprecedented accuracy. On the software side, one major advance involves the development of programs for morphing digitized waveforms.
Automakers have extensive libraries of collision data. But incompatibilities in sampling rate and bandwidth have historically prevented them from translating this data into waveforms that they can input to the control-system environment and run on a shaker table.
Software for digital waveform processing now makes it possible to morph virtually any waveform to fit a given control system environment by means of filtering and resampling techniques. As a result, automakers can now use accumulated data as well as computer-simulated time histories for sensor testing. This reduces the need for time-consuming and expensive crash tests. Resulting test data can be graphically correlated with response information from acceleration as well as sensor trip/timing data (Figure 1).
Control Systems/Shaker Tables
The new software is designed to run on more powerful control systems. These systems feature the expanded CPU horsepower necessary to process digital waveforms. They also offer an expanded dynamic range.
The newer 16-bit systems can control the waveform events accurately enough to maintain consistency from pulse to pulse. More capable control systems are also incorporating the functions of the array of peripheral equipment, such as oscilloscopes and digital triggers, formerly required for sensor testing. Integrated control simplifies the task and shortens the time required for test setup, execution and report generation.
Shaker tables have also progressed to handle the high frequencies and large displacements that typify airbag-sensor testing. Historically, electrodynamic shaker tables, capable of handling high frequencies, have been limited to strokes in the 1- to 2-inch range.
Longer strokes have been available on hydraulic shakers. But hydraulic systems offer limited frequency ranges, typically 50 to 100 Hz, because they are designed primarily to simulate seismic events.
Now there are long-stroke electrodynamic shakers. Capable of handling displacements up to 24 inches and frequencies up to 1,000 Hz, these new systems eliminate many of the previous barriers to vibration testing airbag sensors.
Long-stroke applications present special safety considerations. When dealing with a 24-inch stroke excursion, engineers may be hampered by a crowded lab environment. To prevent accidents, integrated software now makes it possible to begin the test with the shaker head automatically positioned in the center of the expected stroke so that it has ample clearance on either side.
Practical Applications
Spurred by these technical advances, the role of shaker tables in airbag-sensor testing is rapidly expanding. Presently, the sensor tests most likely to be performed on a shaker table involve either repeatability or parametric analysis.
The goal of repeatability tests is to verify that time-domain waveforms adequately simulate cases for airbag deployment. The typical methodology involves mounting a sensor on a table, running the waveform through the system and verifying that the sensor trips at the required point within a few milliseconds. Once repeatability is firmly established, these tests can be used in quality-assurance programs.
Parametric analysis is a more demanding task. The goal is not merely to see if the sensor trips, but to understand the internal dynamics of sensor response to vibration signals.
Does the sensor look for slope change, rate change or profile fit? Precisely when does the sensor trip and how long does it stay open? The answers come from software that tracks sensor output characteristics and enables engineers to precisely control performance factors.
Conclusion
Historical methods for airbag-sensor testing have relied on relatively primitive techniques. Equivalent energy tests with easy-to-generate waveforms have been the typical starting point. Often such tests did not use shaker tables at all, but relied on drop tables and standard MIL-Spec techniques.
Recent advances in hardware/software have established shaker-table tests as a more accurate, convenient and economical alternative. Test personnel adapt readily to the new equipment and methodologies because Windows-based programs with more intuitive interfaces help take the mystery out of signal processing, waveform replications and control software operations.
The complete cost for a shaker, an advanced control system and software runs around $300,000. This is not a trivial amount, but it is usually less than the total cost of running one car into a wall to gather collision data. In contrast to this one-time expense, the cost of an advanced control system can be leveraged over many years and thousands of tests. By morphing previously gathered data to fit test parameters, a sophisticated system enables automakers to leverage their libraries of measurements and avoid the expense of future crash testing.
About the Authors
Bill Grafton is a Senior Development Engineer at Spectral Dynamics. His career includes many years of developing and testing vibration-related control systems for automotive and off-road applications. Mr. Grafton received a master’s degree in engineering from UC/Berkeley. Spectral Dynamics, 1983 Concourse Dr., San Jose, CA 95131, (408) 474-1700.
Copyright 1996 Nelson Publishing Inc.
January 1996