Predicting a likely future failure in vehicle components and systems in advance of the occurrence is important today, and will be even more critical for autonomous vehicles, perhaps far from their home bases. Recent discussion related to prognostics – predictive diagnostics – illustrates the progress being made by SAE’s Industry Technology Consortia (ITC) via Recommended Practice JA 6268.
Adapted from aerospace, where the safety aspects are obvious, the objective of prognostics is to change motor vehicle care and repair from mileage-and-time based intervals for maintenance to one using analysis of real-time data. As much as feasible, diagnosis of failures after they occur would be replaced by predicting a likely future failure in a “sweet spot” of as close to two weeks in advance as possible. The prognostic “prediction” is intended to detect impending failure while the performance still is seemingly within a normal range.
The aim is to derive maximum life from components and systems without subjecting the user to loss of safe use at inconvenient times, compared with the economic disadvantage of service replacements at unnecessarily short intervals. Currently, most automobile diagnosis is by a technician, using service information (manuals and/or on-line), test equipment and shop tools. Any needed software updates are made in the shop, using an internet connection with an SAE J2534 pass-thru device or an OEM tool. Only a few cars are able to use telematics to update software “over-the-air.”
Creating a data stream
Determining when an automotive component, obviously designed for long life, is close to failure means there must be continuously-available data on its “health,” such as provided by OnStar for GM, which introduced automotive prognostics on a range of 2016 Chevrolet models with OnStar 4G connectivity—covering key parts of engine starting and operation including the battery, starter and fuel pump.
Some components “age gracefully,” i.e. deteriorate on a known curve, and lend themselves readily to prognosis. Those with key component lives that presently are less predictable would benefit more from real-time health assessment, perhaps enhanced by machine learning or other assistance.
The health assessment has to be built into components and systems, and the industry is looking to the SAE’s ITC to create the framework. Last year it released JA 6268, an aerospace and automotive Recommended Practice (RP) entitled “Design and Run-Time Information Exchange for Health-Ready Components.” A “health-ready” component is a supplier-delivered part or sub-system that is enhanced, perhaps with additional sensors, to report on its health, and/or provide the integration information so components and/or an entire on-vehicle system can be covered.
The first JA 6268 level (Level 3) is at the component level, and the ITC is seeking to build a registry. Initially it includes the component identification, and the supplier, plus the validation approach (design-time and/or run-time information), format of the health-ready information (math model or math relationship in machine-readable format), name of OEM or integrator, and dates of compliance.
This registry is an ITC spin-off to begin the development of HRCS (Health-Ready Components and Systems). It’s intended to lead to IVHM—Integrated Vehicle Health Management—and eventually (primarily for autonomous vehicles) to Self-Adaptive Health Management, explained Steven Holland, a consultant who helped GM deploy the company’s prognostics initiative while he was employed there.
Holland noted the difficulties both at the diagnostic level and also in prognostics. Rates for NTF (No Trouble Found) or the aero industry’s NFF (No Fault Found) may exceed 50%, even up to greater than 90%, perhaps indicating that “NMF (Not My Fault) is a better name,” he said. Among the reasons for NTF he suggested were:
- The test method doesn’t log all failure modes. This often was identified as the issue when replacement of an electronic module that didn’t log a code or have any other adverse sign, but actually was found to have fixed the problem.
- The test environment is unrepresentative of when the failure occurred (temperature, pressure, humidity, vibration, etc.)
- There are wiring/connection issues not identifiable by the test procedure.
- Other modules, which should be cooperating in the circuit/system, are not performing as expected.
- A key specification for the component was waived as a cost-saving decision.
- Maybe there really is nothing wrong with the component.
The objective of the registry is to enable participants to know the process is consistent with the RP and was verified. A registry of known suppliers also enables sharing of costs and some leveraging of knowledge, Holland observed. Sensors incorporating health assessment would be integrated in an IVHM architecture based on ISO 13374, which includes capabilities for data acquisition, manipulation, measurements of state and health, along with a prognosis if indicated and generation of an advisory if needed.
Failure mode classification also is part of the assessment strategy, led by cost-per-vehicle for repairing a predicted failure. Next in the picture is the severity, led by Most Severe (vehicle non-operational or there is a safety issue), followed by Urgent, Important (including customer inconvenience factor), then Minor Repair and finally Least Severe (goes into routine maintenance).Continue reading »