Why recalls need to be narrower
Automotive recalls have become part of the operating landscape — not exceptional, not rare. The Takata airbag inflator recall reached an estimated 100 million vehicles worldwide. GM’s ignition switch recalls covered up to 30 million. Recall cost scales with batch size and consumer exposure: direct costs run into billions, legal and regulatory fines compound, and brand-trust erosion hits revenue long after the campaign closes.
Recalls typically sweep a whole batch or series rather than individual units. The reason is usually missing traceability: the process conditions under which a specific part was built aren’t linked to the quality record of that part.
Most plants still lean on sample-based quality control — pieces pulled off the line, tested in a lab with proprietary tools, results filed in a separate quality-data system (sometimes still on paper). The test data sits disconnected from the machine conditions at the moment the part was produced.
Joining the two changes the economics. When each SKU carries the machine conditions that produced it, root cause runs retrospectively on the bad units — and a recall targets only the parts with elevated defect probability, not the whole batch.
Condition monitoring meets quality control
The condition of a production machine directly impacts the quality of a work piece. Traceability based on an integrated approach towards process data and quality data is a prerequisite for both real-time in-line non-contact quality control and automated retrospective root-cause analysis. But traceability of machine data imposes a lot of technical requirements on data processing and data storage – for instance, relational database technologies are not able to process high-frequency data streams generated from machine sensors. Big Data is the key technology to make these data streams accessible and make them actionable. Pairing of machine condition data (originating from sensors to measure vibrations, pressures, speed changes, temperatures, torques etc.) and quality data of SKUs (provided by RFIDs) delivers new insights for quality control at a cost structure that was not feasible with prior tooling.
Quality fingerprinting, step by step
DATATRONiQ delivers prebuilt functionalities to establish a data driven quality fingerprint for each manufacturing step of a work piece. By tracking machine health conditions for each work piece that is being processed we are creating a unique data quality universe for real-time quality control and retrospective root cause analysis. Along the production line, each work piece inherits a quality index derived from the health condition of the underlying machine, takes it to the next manufacturing step, synchronizes the quality index, takes it to the next manufacturing step etc., and keeps it until the end-of-line inspection. Having DATATRONiQ’s real-time traceability established, work pieces can be dropped from the production line if quality index falls below a work piece-specific threshold. As a consequence, poor quality of work pieces can be automatically identified and alerted before they are examined in end-of-line inspections. For root-cause analysis, DATATRONiQ works on the fingerprint history directly. Given a set of SKUs — end-of-line rejects or consumer returns — DATATRONiQ triggers a root-cause analysis against the quality fingerprint history of each SKU. The output is a set of rules describing the failure conditions, which feeds back into future production runs.
Fewer rejects, narrower recalls
With DATATRONiQ, manufacturers catch defects at the point they are produced, not at the end-of-line or after consumer returns. The traceability of process data and quality data reduces quality risks during the manufacturing process and leads to a reduced number of rejects. With quality checks automatically carried out 24/7 on each produced work piece (and no longer on samples) DATATRONiQ provides a comprehensive data driven non-contact in-line quality control solution, that is not only affordable and easy to deploy but one that also pays off after a very short time period. In case of looming recalls, manufacturers can now quickly respond with automated root-cause analysis based on quality fingerprinting to better understand and identify the affected products and components, in order to significantly limit the number and corresponding costs of recalls.
