Die Casting Traceability: Shot-Level Part Marking, Data Linking and RFP for Thai Plants

Die casting plants in Thailand that are looking at die casting traceability tend to be in a very similar situation. The parts carry a date pin and a cavity number cast in from the die, and records are kept by lot: “which machine, which day, which die.” The die casting machine has a function that monitors injection conditions, and the waveform of every shot is even stored, yet that data stays inside the machine or on a USB stick and is never tied to the parts. When a customer complains about porosity (gas or shrinkage voids in the casting), the suspect range becomes “everything that machine made that day,” and stock in the warehouse and at the customer is sorted all at once. This is a familiar scene at Japanese-affiliated aluminum die casting plants in Chonburi, Rayong, Ayutthaya and Prachinburi.

This article is a practical guide for plant managers, quality assurance managers, production engineers and IT staff at die casting plants that make automotive, motorcycle and EV parts (housings, cases, brackets, knuckles and so on). It sets out what to decide when strengthening traceability: “what to mark, where, by which method, what to link, how to compare options, what it costs, and how to proceed.”

First, here is the argument of this article in brief.

Die casting quality changes from shot to shot. Yet if records are kept by lot, the suspect range in a complaint spreads to the whole lot. The value of tracking at shot level is not “more records” but “a narrower containment scope.”

There are 3 typical reasons why die casting traceability does not work.

• (1) The mark becomes unreadable in downstream processes (heat treatment, shot blasting, machining)
• (2) The correspondence between shot numbers and parts drifts during extraction, cooling and trimming (warm-up shots, scrapped shots, parts placed by hand)
• (3) X-ray, leak and machining results are scattered across separate systems and are not stacked on the same ID

So, before comparing products, we recommend deciding the following 4 points first.

• (a) At what level you need to track (customer requirements, whether the part is safety-critical or pressure-tight, and how IATF 16949 approaches traceability)
• (b) Where to mark (right after the machine exit as a rule, with a depth and cell size that survive downstream processes)
• (c) At which moment to bind the shot number to the part ID (how to detect drift, and how to stop when it drifts)
• (d) Where to stack downstream inspection results

Note that this article does not recommend any specific product or rank products. It is not legal or standards-compliance advice either. IATF 16949, Thailand’s Personal Data Protection Act (PDPA) and incentives from the BOI (Thailand Board of Investment) are introduced as “points to check,” and final judgments should be confirmed with your customers, certification bodies and specialists. All machine counts, part quantities, unit prices and costs in the calculations are hypothetical figures set for this article.

Why Thai Die Casting Plants Should Revisit Traceability Now

Vehicle production swings widely from month to month

Thai vehicle production has swung widely from month to month in 2026. According to a Reuters article (republished on Capital.com) citing the Federation of Thai Industries (FTI), vehicle production in August 2026 was 124,646 units, up 10.93% year on year. Domestic sales rose 25.59%, while exports fell 2.04%. At the same time, the FTI forecasts full-year 2026 production at −3.33%, citing lower exports due to fighting in the Middle East (its earlier forecast was +3%).

According to monthly articles on the Thai industry news site mreport, production was 120,391 units in June (−7.55%) and 117,383 units in July (+6.12%), and BEV (battery electric vehicle) production in August was 13,126 units, up 77.38% year on year.

According to an August 1, 2026 article in The Nation, the FTI cut its 2026 production target from 1.5 million to 1.45 million units and its export target from 950,000 to 900,000 units (the domestic target of 550,000 units was left unchanged). Production in January to June was 717,212 units, down 1.04% year on year. The reasons given were disrupted shipping routes due to the Middle East conflict, US trade barriers, competition with Chinese EV makers and stricter emission rules in importing countries. The same article also says that much of the growth in EV sales came from imported vehicles rather than locally built ones, and that weak sales of pickups, which use more than 90% locally sourced parts, are weakening the supply chain.

The parts industry is expected to grow on EV parts

In “Industry Outlook 2026-2028: Auto Parts Industry” (May 2026), Krungsri Research, part of Bank of Ayudhya, expects Thailand’s auto parts production index to grow 1.5-2.5% a year in 2026-2028. The main reasons are expanding production capacity for EV parts and the compensatory production and local parts use required of BEV makers under the EV 3.5 scheme. The same report puts 2025 auto parts exports at USD 24 billion, up 6.9%. Keep in mind that this is a forecast by a bank-affiliated research house.

Investment and policy are moving too

According to a BOI press release (No.117/2569, July 2026), investment applications in the first half of 2026 totaled 1,299 projects worth USD 43.6 billion (about 1.47 trillion baht, +37% year on year). The automotive sector accounted for 122 projects worth USD 759.2 million (about 25.7 billion baht), and the “Smart and Sustainable Industry” measure, which covers machinery upgrades, adoption of digital technology and automation, received 132 applications worth USD 507.6 million (about 17.2 billion baht). Applications from Japan numbered 123 projects worth USD 970.1 million (about 32.8 billion baht). All of these are application figures, not approved amounts or actual investment.

The BOI has also issued Announcement No.5/2569 (January 15, 2026), a measure to promote Thai-foreign joint ventures in auto parts manufacturing. If conditions are met, such as a Thai entity holding at least 20% of registered capital, corporate income tax (CIT) exemption is extended by 3 more years (up to 8 years in total). The application period runs from the first business day of 2026 to the last business day of 2027 (based on the BOI’s unofficial English translation; please confirm the detailed conditions with the BOI and specialists).

In the die casting industry, the Southeast Asian die casting exhibition and conference “SEA DIECASTING 2026” was held in Chonburi on May 20-22, 2026. According to a Foundry-Planet article, it drew more than 1,000 visitors from about 20 countries and regions and more than 60 exhibitors, and the conference covered topics such as giga casting, EV manufacturing, lightweight structures, automation and simulation.

EV parts have strict requirements for pressure tightness and porosity

What these figures suggest is that Thai die casting plants are in an environment where “volume swings from month to month, but the share of EV parts keeps rising.” The leak tester maker Cincinnati Test Systems explains that cast housings for EVs (battery enclosures, drive units, power electronics cases) require leak testing. A 2016 report from the Mie Prefecture Industrial Research Institute in Japan also explains in its introduction that casting porosity causes cosmetic defects that appear on machined surfaces and pressure-tightness failures. The more a part needs to be pressure-tight, the heavier the burden of controlling and recording porosity.

With volumes swinging, a single complaint handled with only lot-level records means heavy sorting work and a large impact on the customer. When you go after new orders, customers may also ask what system you have for individual part traceability. Revisiting traceability is preparation for both.

What Is Die Casting Traceability? Three Levels: Lot, Shot and Individual Part

The 3 levels

Die casting traceability means being able to trace, for any single part, “when it was cast, on which machine, in which die and cavity, under which conditions, and which inspections it passed.” It is easier to organize your thinking if you split it into the following 3 levels, depending on how finely you can trace.

In die casting, 1 shot produces several parts (2-cavity, 4-cavity dies and so on), so “shot level” and “individual part level” are not the same. Combining the shot number with the cavity number lets you identify a single part within 1 shot.

What in-die marking can and cannot tell you

In-die marking (casting characters or pins engraved in the die onto the part) is widely used in die casting. According to an article in the January 2018 issue of Die Casting Engineer (NADCA), written by the CTO of the laser marking maker Laserax and co-authors, in-die marking can show the alloy, the plant, the customer’s logo, the production date and shift, and the die and cavity used. However, it is said that time of day and serial numbers are usually not included, and that in-die marking is not used for putting a unique code on each part, for linking to the process conditions of an individual part, or for 2D codes.

In other words, a date pin and a cavity number tell you only “that cavity on that day.” Even if hundreds of shots were cast that day, you cannot tell which shot a part came from. If you want to link parts to shot-level injection conditions, you need to mark a unique code on each part.

The same article says that more and more cases require 100% individual identification not only for high-reliability parts but also for ordinary die castings (note that this is the view of a company that sells laser marking). A 2016 review paper on identification technologies for die casting parts, written by Laserax and others, also states that high-reliability parts such as structural parts usually require individual identification of all parts produced.

The level you need depends on customer requirements and risk

The core traceability requirement for automotive parts is clause 8.5.2.1 (Identification and traceability) of IATF 16949:2016. Because the standard itself is sold for a fee, here we present a summary based on publicly available secondary sources (a summary quoted on a quality forum and an explanatory website).

• For all automotive products, analyze internal, customer and regulatory traceability requirements, and develop a traceability plan based on the levels of risk and failure
• Through that plan, make it possible to identify and segregate nonconforming and suspect products, and meet customer and regulatory response time requirements
• Retain records in a format (electronic, paper, archive) that allows the response time requirements to be met
• Identify individual products by serial number where specified by the customer or by regulation

The key point here is that 100% serialization of individual parts is not a blanket requirement of IATF 16949. Individual serials become necessary when the customer or regulations require them. The level at which you track is decided by customer-specific requirements and by risks such as part safety and pressure tightness. In a stakeholder communiqué in July 2026, IATF gave the planned publication of IATF 16949 2nd Edition as mid-2027 (provisional), but the content of any changes to the traceability clause has not been published.

As an assessment specific to casting, there is AIAG’s CQI-27 (Special Process: Casting System Assessment). According to AIAG’s public information, the 2nd Edition was published in March 2018, and the announcement of the stakeholder review version listed “aluminum high pressure die casting,” “magnesium high pressure die casting” and “zinc high pressure die casting” among the process tables. However, AIAG’s public pages do not describe specific traceability requirements, so it cannot be said that “CQI-27 requires linking shot data to individual parts.” CQI-9 for heat treatment is covered in our article on heat treatment process record systems. AIAG also released a Traceability Guideline in January 2019, announcing that it includes good and bad examples and forms for self-assessment and fire drills (mock exercises).

How to read customer audits and requirements is covered in detail in our article on automotive parts traceability and IATF 16949 and our article on designing the IATF 16949 traceability evidence chain. This article focuses on how to implement those requirements in a casting cell.

What to Record as Shot Data: Shot Data Traceability in HPDC

Quality changes from shot to shot

Die casting is a casting method in which molten metal is injected into the die cavity at high speed and solidifies. Mazda Technical Review (No.31, 2013) describes injecting the molten metal “on the order of 1/100 of a second.” According to the same review, when the injection speed is low and the filling time is long, the metal cools in areas far from the gate, and filling defects such as cold shuts and misruns become more likely. The casting pressure applied right after filling is completed is mainly intended to crush porosity, but it does not transmit well to areas where solidification has already begun, so porosity tends to remain there.

The same review states that in cylinder block production, the degree of vacuum in the cavity, casting pressure, die temperature and other values are monitored in real time, and each shot is checked to confirm that it is within ideal casting conditions, which suppresses quality variation (this is a case of ultra-high-speed, high-vacuum die casting in which the cavity is evacuated to a gauge pressure of −90kPa or lower, and it is not a standard value for ordinary plants). The fact that a car maker says it “checks every shot” implies, conversely, that conditions can vary from shot to shot.

What published research shows about conditions and defects

Some published research has quantified the relationship between conditions and defects. Each of these is an individual study, not an industry average.

A research report from the Mie Prefecture Industrial Research Institute (No.40, 2016) cut 40 test pieces from 16 mass-produced die cast products (all ADC12) from 4 companies in the prefecture and measured the porosity volume fraction (void volume fraction) by X-ray CT. In test pieces cast at about 40MPa, the void volume fraction was scattered between about 0-4%, but the higher the pressure, the smaller the fraction and the narrower the range, and at about 70MPa it was about 2.5% or less (all of these are approximate figures that the report states based on its graphs). The report also found no correlation between total gas content and void volume fraction, and concluded that “reducing total gas content alone cannot reduce the amount of porosity.” It also showed that slow cooling rates greatly increase porosity.

A paper by the Polish researchers Brzeziński et al. (Materials, 2024) studied an automotive safety system part (2-cavity, alloy AlSi9Cu3(Fe), robotic cell with an 850-ton locking force machine). In this plant, the biscuit thickness (the thickness of excess metal left in the shot sleeve) was recorded automatically for every shot, giving 4,995 data points over 7 days. The nominal biscuit thickness was 25mm with a tolerance of ±10mm, and a relationship was observed in which porosity assessed by X-ray increased as biscuit thickness decreased. The rejection rate due to porosity on machined surfaces fell from an average of 1.74% in the first period to 0.92% and then 0.79%, a 55% reduction from the initial value. The paper attributes this to process stabilization and to bringing biscuit thickness closer to nominal. Note that the roughly 18% improvement in the overall X-ray assessment came from a separate measure, sealing the die slider area to keep leaking metal inside the die, and is a factor separate from biscuit thickness.

The parameters this paper lists as commonly controlled are plunger speed (first and second stages), the switchover point to high speed, and intensification (casting pressure).

Candidate items for HPDC quality data

Using the sources above as a starting point, the candidate items to keep for each shot can be organized as follows. Which items to store, and at what granularity, has to be decided by each plant for each part; there are no general values.

Data volume varies greatly depending on whether you store the full injection waveform for every shot or only representative values calculated from the waveform (maximum speed, peak pressure and so on). A realistic approach is to start by storing representative values and judgment results for every shot and waveforms only for a fixed period, then widen the scope as needed. How to collect data from equipment is covered in our article on PLC data collection, and how to spot changes in conditions early with control charts is covered in our article on SPC anomaly detection AI.

The condition of the equipment also affects quality. The machine maker UBE Machinery explains that when aluminum buildup in the shot sleeve increases, the plunger tip galls during injection, and waves on the metal surface make casting quality unstable (this is the maker’s explanation). Keeping equipment condition data on the same timeline as the shots helps when you later look for causes.

3 Reasons Die Casting Traceability Fails to Take Hold

Even if you install a system that collects shot data, it is of no use for containment unless you can look up the data from the part. The reasons it does not work come down to roughly the following three.

Reason (1): The mark becomes unreadable in downstream processes

After casting, die cast parts go through trimming, heat treatment, shot blasting, machining, washing, conversion coating, painting and so on. The review paper mentioned earlier lists the difficulties of marking die castings: short cycle times, surfaces that are not flat, parts that are hot at the time of marking and still carry release agent and dirt, and the need to survive post-processing such as heat treatment, shot blasting, washing and conversion coating.

For example, laser marking is said to be vulnerable to shot blasting as is, because the black marked area rises above the base surface. Inkjet is also said to come off easily in machining, shot blasting, washing and conversion coating. A code that read fine at the machine exit can no longer be read before shipment, and then you cannot stack the downstream results that matter most on the same ID.

Reason (2): The correspondence between shot numbers and parts drifts

The other typical problem is that the correspondence between shot numbers and parts drifts. In a die casting cell, things like the following happen every day.

• Warm-up shots to heat the die are counted on the same counter as production shots
• Shots whose conditions deviated, or parts dropped during extraction, are scrapped, and the numbers skip by that much
• The extraction robot stops, and an operator places parts by hand on the cooling conveyor or a rack
• The order of parts changes during cooling or trimming
• A part taken out of the line is later put back into the process (re-entry)

If the marking station operates on the assumption that “the next part to arrive gets the next shot number,” a single manual placement or scrapped part shifts the correspondence of every part after it by one. And no one notices that it has drifted. The data looks complete, but the content is wrong, which is the most dangerous state of all.

Reason (3): Inspection results are scattered across separate systems

The third is that results from X-ray, leak testing, machining measurement and so on are kept separately in each device or in Excel and are not stacked on the same part ID. If you collect each department’s data after a complaint arrives and match it by hand, it takes days, and the matching itself can go wrong.

The fact that unreadable records widen the suspect range also shows up in recall reports, regardless of part type. In a Jaguar Land Rover recall report (26V613, rear subframe of the Range Rover Sport SV) filed with the US NHTSA (National Highway Traffic Safety Administration) in September 2026, after the defect was found at the supplier, thousands of production images normally used for dimensional analysis were reviewed by hand, and cracks were confirmed on a small number of vehicles. On the other hand, because of a system storage error, production images could not be retrieved for a larger number of vehicles. Vehicles with confirmed cracks get the part replaced, and vehicles that could not be checked from images are inspected and have the part replaced if cracks are found. The number of vehicles covered was 873 in the initial report (September 24) and 1,073 in the amended report (September 29) (no accidents or injuries have been reported in the US). Even if records exist, if they cannot be retrieved, the vehicle falls on the “could not be confirmed” side and becomes subject to inspection.

Where to Mark: Process Flow and Part Marking Position for Die Castings

Apply the ID right after the machine exit as a rule

The figure above shows a typical process flow for die cast parts, where the ID is applied, and where the results of each process are stacked. Heat treatment is not used for every alloy or part.

To bind shots to parts reliably, the rule is to apply the ID as soon as the part leaves the machine, before it mixes with other parts. If you mark after trimming or cooling, there are more chances for the order to change in between, and the drift described in Reason (2) becomes more likely.

Configurations that mark right after extraction do appear in published material. Laserax explains that its shot-blast-resistant marking can identify die castings “right at the exit of the casting machine” (this is the maker’s explanation). A 2017 paper by Laserax and others describes a configuration in which, while laser marking takes place in a rotary-table enclosure, the robot carries out quenching, trimming, placing the part on the outfeed conveyor, picking up the next part and verifying the mark. As an older example, a 2008 trade magazine article describes flexible layouts by the machine maker Bühler, such as marking after the extraction robot places the part on a cooling rack, or handing it over to the trim press (this is a 2008 article, not a description of current products). It can be said that such configurations exist, but not every plant marks at this position.

Marking “just the number” is enough

As a rule, marking only a unique ID (serial number) on the part is enough. The shot number, injection conditions and inspection results are held on the database side, linked to the ID. If you try to mark information beyond the ID on the part, the code gets bigger and marking time gets longer. Whether to add human-readable characters, and which surface to mark (a surface that is not machined away and that can still be read after assembly), should be decided together with the customer, based on their drawings.

Stack every result on one ID

As the lower row of the figure shows, you stack shot data, the heat treatment batch, the X-ray judgment, machining measurements, the leak test result and shipment packing on one ID, in sequence. This requires a device that reads the code at the entrance of each process. Decide here, too, how to handle parts whose code cannot be read (stop, allow manual entry, or route them to a separate flow).

How to design the part numbering scheme itself is covered in our article on serial number management systems, and the cross-industry approach to linking equipment history to products is covered in our article on production equipment traceability.

Comparing Part Marking Methods for Die Castings: DataMatrix Laser Marking and Alternatives

4 methods

Methods for identifying die cast parts can be broadly divided into 4. Based on the comparisons in the review paper mentioned earlier (2016) and the Die Casting Engineer article (2018), they can be organized as follows.

Inkjet is said to come off easily in machining, shot blasting, washing and conversion coating. In practice, one option is to combine in-die marking (lot and cavity identification) with a unique code applied by laser or dot peen. How to choose the marking equipment itself is covered in our article on industrial laser marking machine selection, and how to choose between code types (barcode, QR, DataMatrix, direct marking, RFID) is covered in our article on traceability barcode selection.

Does marking time fit within the cycle time?

The first thing to check with laser marking is whether marking time fits within the machine’s cycle time. A 2017 test by Laserax with Mercury Marine and others reported the following values. These are values under a single company’s test conditions and will vary with alloy, shape, equipment and marking content.

In the same test, high-contrast marking on aluminum with a 2D head required the focal position to be within ±3mm of the surface, and contrast dropped sharply outside that range. If part position is not stable, a 3D head that corrects with a distance sensor is used (the added delay was reported as under 50ms).

Deep engraving that survives shot blasting

For parts that go through shot blasting, “deep engraving” is used: the laser engraves down over multiple passes, then blackens the mark, so that the surrounding surface protects it. In the same 2017 test (Mercury Marine parts in 362 alloy), the results were as follows.

• For cell size (1 module of the code), 0.4mm could not be read after blasting, and 0.6mm was optimal
• For the number of engraving passes, 3 passes disappeared, 6 passes could be read but used a lot of error correction, 9 passes were good, and 12 passes gave maximum contrast, approaching the pre-blast average (0.61)
• A code of about 13mm took 37 seconds with a 100W laser
• Surface marking without engraving took about 2 seconds for 20mm with a 1.0mm cell, but contrast after blasting was low

37 seconds can be longer than the cycle time of many die castings. This is why the paper presents a rotary-table enclosure configuration that hides marking time behind other work. In a separate document, Laserax shows marking in under 6 seconds at a depth of 0.15mm for time-constrained cases (this is the maker’s explanation; the test in that document used A380 alloy, a 10×10 matrix and 90 seconds of shot blasting).

For parts that go through electrocoating (E-coat), cells of 0.75mm or smaller could not be read after coating, and the best conditions were a 1.25mm cell, 20mm and about 2 seconds of marking. It was reported that electrocoating after shot blasting actually increased contrast. Regarding heat treatment, the Die Casting Engineer article says that chromate treatment and T6/T7 heat treatment had little effect on the readability of laser marks, provided the technique and parameters were chosen appropriately.

In other words, marking conditions are determined by “which downstream processes the part goes through.” In your RFP, have vendors do trial marking on your own parts under your own downstream conditions (blasting media and time, heat treatment conditions, painting), and compare them on the read results after the parts have gone through those processes.

Read the code right after marking and measure its quality

The 2017 paper recommends, for “full traceability,” reading the code with a camera right after marking and checking it against the data that was supposed to be marked. The reading step also lets you confirm whether code quality meets the customer’s and your own criteria.

The quality of codes marked directly on parts (DPM) is measured using the approach of ISO/IEC 29158 (a quality guideline for direct part marking based on AIM DPM). According to Cognex documentation, the main evaluation parameters are cell contrast (CC), cell modulation (CMOD), distributed damage grade (DDG), minimum reflectance (MR) and decode, and in addition to the default 4-direction 45° lighting, 30° and 90° coaxial diffuse lighting can also be used. DataMatrix has error correction and can sometimes be read even if part of it is missing, but it is important not to rely on that and to set conditions that maintain a given grade even after downstream processes. How to choose a verifier is covered in our article on barcode verifiers.

The Moment You Link the Shot Number to the Part ID

Do not assume “the next part is the next shot”

As we saw in Reason (2), the correspondence between shot numbers and part IDs drifts if you do nothing about it. The most important thing in linking is to not assume that “the next part to arrive is the next shot,” and instead build a mechanism that can confirm the correspondence is correct at the moment you bind them.

Here are some example approaches.

• At the time of marking, receive the shot number and judgment for that shot from the machine. Rather than the marking station keeping its own counter, it marks only after receiving information from the machine (or the cell controller) such as “the part just extracted is shot No. XX, cavity X, judged good”
• Do not issue IDs for warm-up shots or scrapped shots. Shots judged “warm-up” or “scrap” on the machine side are excluded from marking, and only their records are kept
• Confirm the shot-ID pair by reading right after marking. Read the marked code, and mark it as “linked” only when it matches the data that was sent
• Fix where each cavity’s part is placed. For multi-cavity dies, decide where the robot places the part from each cavity, and match the marking position to the cavity number

How to detect drift

Even so, drift will happen. Decide at the design stage how drift will be detected.

• Count check: At fixed intervals, compare the machine’s number of good shots × number of cavities with the number of parts marked and read. Raise an alert if there is a difference
• Time check: Raise an alert if the difference between the shot time and the marking time falls outside what is expected (the time from extraction to marking). This requires the clocks of the machine, the marker and the data collection server to be aligned. How to align them is covered in our article on factory time synchronization
• Recording manual placement: When an operator places a part by hand or returns it to the process, have them scan that part’s code at the terminal and confirm manually on screen which shot the part came from. Parts that cannot be confirmed are treated as “shot unknown”

Stop when it drifts

Also decide what happens when drift is detected. The principle is that it is better to stop and confirm than to keep recording while drifted. Data after a drift cannot be trusted, even if it looks complete. If the checks do not match, stop marking, put the parts from that period into the “shot unknown” category, and fall back to lot-level handling (managed within the same scope as before). Write into the work standard the criteria for stopping the cell and who checks and restarts it.

Link die information with the same key

In addition to shots and IDs, link the die ID and cavity number and the die maintenance history (repairs, part replacement, cleaning) on the same timeline. As we will see later, there are cases where defects started after die maintenance. Managing die shot counts and maintenance is covered in our article on mold management systems.

Stacking X-ray, Leak and Machining Results on One ID: Die Casting X-ray Inspection Records

Keep X-ray inspection results on the ID

For X-ray inspection of aluminum and magnesium die castings, there are the reference radiographs of ASTM E505 (currently E505-22, December 2022). According to the summary on AFNOR’s sales page, E505-22 is a set of reference radiographs showing discontinuities (defects) in aluminum and magnesium die castings by category and severity, with 3 categories each for aluminum and magnesium and 4 severity levels for each, covering wall thicknesses from 1/8 inch (3.2mm) to 5/8 inch (15.9mm) (the standard itself is sold for a fee, and this is an introduction based on the summary). The study by Brzeziński et al. mentioned earlier also used ASTM E505 and internal criteria for X-ray porosity assessment.

In the same study, X-ray inspection was not done on every part but on representative samples over a wide period that took process variation into account, and the authors state that X-raying every part is quantitatively impossible. Precisely because not every part is inspected, it is meaningful to keep which parts were X-rayed and which shot conditions they came from. If you link the sample judgments to shot conditions, you can narrow the range of “other parts cast under the same conditions.”

Systems that automate inline X-ray inspection have also been published. Fraunhofer IIS explains that its ISAR detects voids and porosity by automatic defect recognition, sends inspection results to production in real time, connects to existing production data acquisition systems, and is used for trend monitoring and process optimization. Fraunhofer IIS also explains that its RoboCT is automated 2D X-ray and CT inspection for large, complex die castings (cross members, battery trays, giga castings and so on), and that the related product Cast Control links die casting process data with inspection results to identify systematic causes of defects (these are product descriptions by a research institute). On the other hand, the Mie Prefecture Industrial Research Institute wrote in its 2016 report that determining porosity causes by X-ray CT had not become widespread and was difficult for small and medium-sized companies to use (a statement from 10 years ago).

Keep leak test results on the ID

For housings and cases that must be pressure-tight, also stack leak test results on the same ID. Cincinnati Test Systems explains that the most common method for leak testing EV cast housings is differential pressure decay.

Within the research for this article, we could not find a primary source that specifically shows how to link leak test results to part IDs. In general, the code is read at the tester entrance, and the test conditions (test pressure, time, program number), measured value, judgment, tester number and date and time are written to that ID. Also decide how many retests are allowed for rejected parts and how retest results are kept (kept as history rather than overwritten). How to choose leak test equipment is covered in our article on industrial leak test machine selection.

Keep machining and heat treatment results on the ID

If porosity appears on a machined surface, being able to look up from the part ID which shot it was cast in lets you investigate the relationship between injection conditions and porosity after the fact. As mentioned at the start, casting porosity causes cosmetic defects on machined surfaces and pressure-tightness failures (introduction of the Mie Prefecture Industrial Research Institute report). The study by Brzeziński et al. uses “rejection due to porosity on machined surfaces” as its indicator because some porosity only comes to the surface when the part is machined. For parts that are heat treated, link the heat treatment batch number and furnace records to the IDs of the parts in that batch.

Stacking inspection results on one ID in this way lets you trace the causes of defects back to shot conditions. How to hold quality data is covered in our article on quality data management systems, and how to go about identifying causes is covered in our article on defect root cause identification systems.

Containment Scope: Lot-Level vs Shot-Level Traceability

How much does the suspect range change?

The value of shot-level tracking shows most clearly in containment when a complaint arrives. Here we compare using the same hypothetical plant as in the later calculation. All quantities and amounts below are hypothetical figures set for this article.

• Assumed plant: aluminum die casting, 8 machines, 240,000 parts a month
• Assuming 25 operating days a month, each machine makes 240,000 ÷ 8 ÷ 25 = 1,200 parts a day
• Current state (lot level): records are “machine × day.” A complaint can be traced only to “these 2 days on this machine,” so the suspect range is 1 machine × 2 days = 2,400 parts
• After introduction (shot level): the shot can be identified from the ID of the complained part, and the range can be narrowed to the shot whose conditions deviated and the shots before and after it. The suspect range is 120 parts (for example, 60 shots in total before and after with a 2-cavity die)

The left side of the figure above compares the number of parts in the suspect range, and the right side compares the cost per complaint.

The difference is 246,000 − 54,800 = 191,200 THB per complaint. The customer-side cost is assumed to fall because we assume that a narrower suspect range shortens sorting and stoppage at the customer; how much it actually falls depends on your agreement with the customer.

An example of narrowing the range: die maintenance and lots

This is a lot-level example rather than a shot-level one, but there is a case where a defect triggered by die maintenance was handled by narrowing the scope to a single lot. In a recall report Hyundai filed with NHTSA (26V069, February 2026), the company said that front steering knuckles on the 2026 Kona (gasoline models) from a single production lot may have been improperly cast. The cause was contamination of the die cavity after scheduled maintenance (die repair in October 2025). The supplier repaired the die on October 16, 2025; the recall covered 4,555 vehicles, the estimated defect rate was 6%, and the production period was October 31 to December 22, 2025. The remedy is for dealers to inspect the knuckle lot information and replace the part if necessary (no accidents or injuries have been reported).

What this case shows is that if die maintenance history and part lots are linked, the scope can be narrowed to a specific lot. Note that what was used for the inspection here was lot information, not individual serials. Going to shot level or individual part level can narrow the range further, but that requires the marking and linking mechanisms described in the previous sections.

Verify the containment mechanism with drills

Installing a system does not tell you whether you can really narrow the suspect range. Use a mock recall (fire drill) to check how many minutes it takes to trace from the ID of a complained part to the shot, the conditions, parts made under the same conditions, and where they were shipped. How to run one is covered in our article on mock recall testing. For mechanisms that stop parts from shots with deviated conditions before shipment, see also our article on defect outflow prevention.

Die Casting Quality Systems for HPDC Quality Data: Product Types and How to Combine Them

4 types

A mechanism that links shot data and part IDs is not necessarily complete within a single product. Divided by role, it is a combination of the following 4 types.

Examples of published functions

Here are examples of functions published by machine makers and others. Each is an introduction within the scope of what is written on the companies’ official pages or in trade magazine articles, and is not a recommendation or comparison. Effect figures are not covered, because no basis is given for them.

• Bühler “SmartCMS”: According to the maker, it collects process values from each peripheral device in the cell, can assign them to a unique ID for each part, and can always reference the part’s position in the production flow. An article on the “Digital Cell” concept announced in 2019 explains that SmartCMS provides time-synchronized process traceability for all parts
• UBE Machinery: The maker’s material introduces ICT tools such as camera monitoring of aluminum buildup in the shot sleeve and visual inspection with a 3D camera (automatically judging flow marks and similar defects and ejecting NG parts within the cycle), and says that all ICT data is managed linked to shot numbers and centrally managed in coordination with core systems
• Shibaura Machine “machiNetCloud”: A 2022 trade magazine article describes remote access to real-time machine data, automatic recording of downtime, and tracking of operating status, KPIs and history (the article does not mention shot data or part traceability)
• Visi-Trak: The product names “True-Trak” and “Sure-Trak” can be confirmed on the official site (details of their functions were not checked for this article)

None of the above had prices published on their official pages.

Moves toward communication standards

Different machine makers mean different ways of getting data out. On this point, an international working group led by VDMA and CEMAFON has been set up and is working on standardizing the interface for die casting cells based on OPC UA. A June 2023 trade magazine article said that the first modules, “machine identification,” “machine status” and “alarms and warnings,” were scheduled for release in summer 2023. Within the research for this article, we have not confirmed which module will handle shot data such as injection waveforms. In your RFP, ask each vendor specifically what information it can provide over OPC UA.

How to think about the combination

One possible division of roles is to build “shot to part ID” in the machine maker’s cell management, hold “part ID to downstream results to shipment” in the MES or quality system, and leave long-term waveform storage to a historian. Decide first which system is the master (where part IDs are issued, and which side is treated as correct when data does not match). How to choose a historian is covered in our article on industrial historian selection.

Model Calculation of Costs and Benefits (Hypothetical Figures)

Here we estimate the benefits and costs of introducing shot-level traceability. All machine counts, quantities, numbers of cases, unit prices and costs below are hypothetical figures set for this article. For actual decisions, replace them with your own figures.

Model assumptions

• Aluminum die casting, 8 die casting machines, 240,000 parts a month
• Current state: in-die marking (date, cavity) and lot (machine × day) management
• Shot conditions are stored in the machine but are not linked to parts
• Currency: Thai baht (THB)

Scenario A: Introduce on all 8 machines

Benefit (1): Containment

• As in the previous section, the difference per complaint is 191,200 THB
• Assuming 6 complaints a year, 6 × 191,200 = 1,147,200 THB a year

Benefit (2): Lower defect rate through automatic segregation of parts with deviated conditions and condition improvements

• Assume the defect rate falls by 0.3 percentage points
• 240,000 parts × 0.3% = 720 parts a month. Assuming a loss of 60 THB per part, 720 × 60 = 43,200 THB a month
• 43,200 × 12 = 518,400 THB a year

The 0.3 percentage points is an assumption. In the study by Brzeziński et al. mentioned earlier, the rejection rate fell from 1.74% to 0.79%, but that is a research case from a single plant and a single part, and it is not used directly in this calculation. Treat it only as a reference for comparison.

Benefit (3): Paperwork for paper records and transcription

• 80 hours a month × 200 THB × 12 months = 192,000 THB a year

Costs are set as follows.

• Initial cost: 5,000,000 THB (8 marking and reading cells, data collection, linking software, integration with existing systems, training)
• Annual cost: 400,000 THB (maintenance, licenses, consumables)

The calculation is as follows.

• Annual net benefit: 1,857,600 − 400,000 = 1,457,600 THB a year
• Payback period: 5,000,000 ÷ 1,457,600 ≈ 3.43 → about 3.4 years

Sensitivity A: 2 complaints a year

• Benefit (1): 2 × 191,200 = 382,400 THB
• Total: 382,400 + 518,400 + 192,000 = 1,092,800 THB
• Annual net benefit: 1,092,800 − 400,000 = 692,800 THB a year
• Payback period: 5,000,000 ÷ 692,800 ≈ 7.22 → about 7.2 years

Scenario B: Start with the 2 machines that run safety-critical and pressure-tight parts

Assume that 4 of the 6 complaints a year involve parts from the 2 machines that run safety-critical and pressure-tight parts.

• Benefit (1): 4 × 191,200 = 764,800 THB
• Benefit (2): 518,400 × 2/8 = 129,600 THB
• Benefit (3): 192,000 × 2/8 = 48,000 THB
• Total: 764,800 + 129,600 + 48,000 = 942,400 THB a year
• Initial cost: 1,900,000 THB, annual cost: 150,000 THB
• Annual net benefit: 942,400 − 150,000 = 792,400 THB a year
• Payback period: 1,900,000 ÷ 792,400 ≈ 2.40 → about 2.4 years

Sensitivity B: 1 complaint a year on these 2 machines

• Benefit (1): 1 × 191,200 = 191,200 THB
• Total: 191,200 + 129,600 + 48,000 = 368,800 THB
• Annual net benefit: 368,800 − 150,000 = 218,800 THB a year
• Payback period: 1,900,000 ÷ 218,800 ≈ 8.68 → about 8.7 years

The figure above lines up the initial cost, annual net benefit and payback period for the 4 cases.

How to read the calculation

In this model, what determines the payback period is not the number of machines but the number of complaints (containment events) and the width of the suspect range. Even with all 8 machines, 2 complaints a year means about 7.2 years, while just 2 machines pay back in about 2.4 years if they are the cell running the parts that draw many complaints. Conversely, the same 2 machines take about 8.7 years at 1 complaint a year.

So starting with the cell that runs the parts that draw complaints pays back faster. On the other hand, in a plant with almost no complaints, it is hard to decide on cost recovery alone, and the decision becomes one of responding to customer requirements (a condition for winning orders).

What this calculation does not include

• Recall costs (because no reliable published general figure for cost per vehicle could be found)
• Orders won by having an individual part traceability system
• Tax incentives from the BOI and others
• The effect on output if marking lengthens the cycle time (this needs particular checking when deep engraving is required)

How to build cost estimates and think about TCO is covered in our article on traceability system costs.

What to Check in an RFP and FAT/SAT for Die Casting Traceability

Test with your own parts and downstream processes, not catalog features

In the RFP (request for proposal), rather than asking yes/no questions about whether features exist, we recommend asking vendors to actually run marking, reading and linking with your own parts under your own downstream conditions. Set the criteria for FAT (factory acceptance test before shipment) and SAT (site acceptance test) on the same items.

Acceptance tests should run “abnormal” cases

In acceptance tests, check not only that things run normally, but also deliberately cause abnormal situations and see whether the system stops and whether records are kept. Warm-up shots, scrap, manual placement, communication loss and unreadable codes will always happen in mass production. The most important acceptance criterion for shot-level traceability is that the system does not silently shift the correspondence when these occur.

How to Introduce Die Casting Traceability: 90 Days, Starting with 1 Cell for Critical Parts

Rather than introducing it on every machine at once, we recommend testing it on 1 cell that runs critical parts, reviewing the benefits and issues, and then expanding. The following is a guide.

Days 1-30: Decide the tracking level and the target cell

• List each customer’s traceability requirements (whether individual serials are needed, retention period, response time)
• Review past complaints and count which parts and which machines they involved and how many parts were in the suspect range
• Choose 1 cell that runs safety-critical or pressure-tight parts that have had complaints
• Identify that cell’s downstream processes (heat treatment, shot blasting, machining, painting, leak testing) and write down the conditions the mark must survive
• Check which shot data items can be obtained from the machine and how to get them out

Days 31-60: Test marking and linking on 1 cell

• Do trial marking on your own parts, and measure reading and quality grades after downstream processes
• Place marking and reading right after the machine exit, and link parts by receiving the shot number, cavity and judgment
• Run production including warm-up, scrap and manual placement, and test the rules for drift detection and stopping
• Check the count every day (good shots × number of cavities versus the number of linked parts)

Days 61-90: Stack downstream results and verify with a mock recall

• Stack X-ray (sample), machining and leak results on the same ID
• Set up 1 fictitious complaint and run a mock recall that traces from the ID to shot conditions, parts made under the same conditions, and where they were shipped
• Compare with the situation before introduction: how many parts the suspect range was narrowed to, and how many minutes the trace took
• Based on the results, decide whether to expand to other cells and in what order

Points to Check Especially in Thai Plants

PDPA: when operator IDs or biometrics are linked

Shot data itself is not personal data. However, if you link operator IDs to shots or parts, that data can become personal data. According to the unofficial English translation of Thailand’s Personal Data Protection Act (PDPA, B.E. 2562), Section 24 prohibits collecting personal data without the data subject’s consent, while setting out exceptions such as performance of a contract, legitimate interest and compliance with the law. Section 26 treats biometric data such as facial, iris and fingerprint recognition data as sensitive data and, as a rule, requires explicit consent. If operators log in to cell terminals using biometric authentication, the issues in Section 26 come into play.

A guide by the law firm Suwanvara (updated September 2026) advises, on the legal basis for employee data, that security and IT monitoring may rely on legitimate interest, whereas consent in an employment relationship is unlikely to be regarded as freely given, so consent should be limited to processing that is genuinely optional. Please confirm with legal specialists which data to attach operator IDs to, its purpose and retention period, and how to explain it to employees.

Multilingual screens and work standards

Confirmation during manual placement, and stopping and restarting when drift is detected, are operated by operators and team leaders on the floor. Make screens and alarms understandable in Thai, and align work standards in Japanese, Thai and English. It is also safer to make the storage location and labeling of parts in the “shot unknown” category understandable without relying on language.

Retention period and storage

Retention periods differ by customer. Storing waveforms for every shot makes data volume large. Decide early how to split storage, for example keeping representative values and judgments for a long time and waveforms for a fixed period, and where backups go (on site or in the cloud). Also check whether the cell can hold data temporarily in case of power outages or network loss.

Check the latest conditions for BOI incentives

Under the conditions published in the 2023 Investment Promotion Guide, the BOI’s “Smart and Sustainable Industry” measure provides, for machinery upgrades and automation in existing businesses, exemption from import duty on machinery and a 3-year CIT exemption (capped at 50% of the investment). The investment must be at least 1 million baht, excluding land and working capital. For BOI-promoted projects, the condition is that the CIT exemption has ended or was never granted. If linkage with the Thai automation industry is at least 30% of the value of the upgraded machinery, the cap becomes 100%, and investment in software and IT integrated with machinery for control, monitoring and manufacturing support is also counted toward the investment (there are full-amount and half-amount categories). Some activities are not eligible; the guide lists general automobile manufacturing (activity 3.6) and others, but parts manufacturing (activities 3.4 and 3.5) is not on the ineligible list.

However, these are the conditions in the 2023 guide. The details of the 2026 version could not be confirmed in the research for this article, so if you are considering an application, please confirm the latest conditions with the BOI and specialists.

Requirements differ by customer

Japanese, European and Chinese OEMs and Tier 1 suppliers differ in the traceability level, retention period and submission format they require. If the same cell runs parts for multiple customers, decide whether to meet the strictest requirement or to separate by part. Also check in the RFP whether the system can handle changes in customer requirements through settings (code scheme, retention period, output format).

Frequently Asked Questions

What is die casting traceability?

It is the state of being able to trace, for 1 die cast part, when it was cast, on which machine, in which die and cavity, under which injection conditions, and which downstream processes and inspections it went through. Depending on how finely you can trace, it is considered at lot level, shot level or individual part level. The purpose is not to increase records but to narrow the suspect range in a complaint and to be able to trace causes back to conditions.

What is the difference between lot control and shot data traceability?

Lot control records by units such as “machine × day” or “die × shift,” and this is also as far as the date pin and cavity number cast into the part can tell you. Shot-level control marks a unique code on each part and links that code to the shot number, so that you can look up the injection conditions of each shot from the part. In this article’s hypothetical calculation, we assumed that the suspect range narrows from 2,400 parts at lot level to 120 parts at shot level.

Where should part marking for die castings be done?

As a rule, as soon as the part leaves the machine, before it mixes with other parts. If you mark after trimming or cooling, the order can change in between, and the correspondence with shots is more likely to drift. Published material also describes configurations that mark right after extraction while the robot carries out quenching and trimming during marking. Choose a surface that is not machined away and can still be read after assembly, together with the customer and based on their drawings.

Does shot blasting erase DataMatrix laser marking on castings?

Laser marks made only on the surface become hard to read after shot blasting. Deep engraving, in which the mark is engraved down over multiple passes and then blackened, makes it possible to survive, but it takes longer. In a 2017 test by Laserax and others (a single company’s test conditions), a 0.6mm cell was optimal, and a code of about 13mm took 37 seconds at 100W. Do trial marking on your own parts under your own blasting conditions, and check the read results and quality grades after blasting.

What shot data should be recorded?

Candidates include the shot number and timestamp, die ID and cavity number, injection speed (slow and fast) and fast-shot switchover position, casting pressure, biscuit thickness, die temperature, degree of vacuum (for vacuum die casting), cycle time and the shot judgment result. Which items to keep has to be decided by each plant for each part; there are no general values. A realistic approach at first is to store representative values and judgments for every shot and waveforms for a fixed period.

Does IATF 16949 require individual part traceability?

It is not a blanket requirement. According to publicly available summaries, clause 8.5.2.1 of IATF 16949 requires developing a traceability plan based on the levels of risk and failure, and calls for identification of individual products by serial number where specified by the customer or by regulation (because the standard itself is sold for a fee, this is a summary based on secondary sources). Customers may require individual part traceability for safety-critical or pressure-tight parts, so please check customer-specific requirements. Please confirm compliance judgments with your customers and certification body.

How much does die casting traceability cost, and how long does it take?

It varies greatly with the number of machines, the marking method (whether deep engraving is needed), how data is obtained from existing machines and the scope of integration with downstream processes, and there are no published prices, so no single figure can be given. In this article’s hypothetical calculation, we assumed an initial cost of 5,000,000 THB and an annual cost of 400,000 THB for all 8 machines, and an initial cost of 1,900,000 THB and an annual cost of 150,000 THB for 2 machines. As a guide for the timeline, we showed an approach of spending 90 days on 1 cell for critical parts to test marking, linking, consolidation of downstream results and a mock recall, and then expanding to other cells.

Summary

• Die casting quality changes from shot to shot. If records are kept by lot, the suspect range in a complaint spreads to the whole lot. The value of shot-level tracking lies in a narrower containment scope
• A date pin and cavity number in the die tell you only “that cavity on that day.” Linking to shot conditions requires a unique code on each part
• Individual serials are not a blanket requirement of IATF 16949; they become necessary where specified by the customer or by regulation (summary based on secondary sources). Decide the tracking level based on customer requirements and risk
• The typical reasons it fails to take hold are 3: (1) the mark becomes unreadable in downstream processes, (2) the correspondence between shot numbers and parts drifts, and (3) inspection results are scattered across separate systems
• Apply the ID right after the machine exit, and decide through trial marking the conditions (depth, cell size) under which it can still be read after your own downstream processes
• Do not assume “the next part is the next shot”; mark only after receiving the shot number and judgment from the machine. Do not issue IDs for warm-up or scrapped shots, and stop and confirm when drift occurs
• Stack X-ray, leak, machining and heat treatment results on the same ID. Even if X-ray is not done on every part, linking samples to shot conditions lets you narrow the range
• In the hypothetical calculation, the difference per complaint is 191,200 THB. Payback is about 3.4 years for all 8 machines at 6 complaints a year, and about 2.4 years for 2 machines at 4 complaints a year. What determines payback is not the number of machines but the number of complaints and the width of the suspect range
• In acceptance tests, deliberately cause warm-up shots, scrap, manual placement and communication loss, and check whether the system stops and whether records are kept
• In Thailand, check early on the PDPA when operator IDs or biometrics are linked, multilingual screens and work standards, retention periods, the latest BOI conditions and differences in requirements between customers

TOMAS TECH helps factories in Thailand with collecting data from equipment, building part identification and traceability systems, and integrating them with existing systems. You can also consult us before choosing a product, for example at the stage of deciding the level at which to track, or at the stage of testing the link between marking and shot data on 1 cell. Please contact us through this form.

References

• Capital.com (Reuters): Thailand’s August car production rises 10.9% on year (September 28, 2026)
• mreport: Electric Vehicle tag list (monthly articles on Thai vehicle production, July-September 2026)
• The Nation: Thailand cuts 2026 car output target as exports weaken (August 1, 2026)
• Krungsri Research: Industry Outlook 2026-2028: Auto Parts Industry (May 7, 2026)
• BOI: Press Release No.117/2569 (July 23, 2026)
• BOI: Announcement of the Board of Investment No.5/2569, unofficial English translation (January 15, 2026)
• BOI: Measure for Industrial Upgrades towards Smart and Sustainable Industry (Investment Promotion Guide 2023)
• Foundry-Planet: Southeast Asia increasingly moves into focus for the die casting industry (July 3, 2026)
• Mazda Technical Review No.31: High-precision thin-wall die casting and die technology that made SKYACTIV technology possible (2013) (in Japanese)
• Mie Prefecture Industrial Research Institute Research Report No.40: Effects of cooling rate, casting pressure and gas content on porosity formation in aluminum alloy die castings (2016) (in Japanese)
• Brzeziński et al.: Prediction of the Stability of the Casting Process by the HPDC Method on the Basis of Knowledge Obtained by Data Mining Techniques (Materials, vol. 17, no. 23, 5935, 2024)
• UBE Machinery: Productivity improvement system using ICT (2025) (in Japanese)
• Laserax / Viami / Université Laval: Review of technologies for identification of die casting parts (2016)
• Laserax / Mercury Marine / Viami: Important considerations for laser marking an identifier on die casting parts (2017)
• Die Casting Engineer (NADCA): Traceability and Laser Marking of Die Castings (January 2018 issue)
• Laserax: Shotblast Resistant Laser Marking for Die Casters (January 7, 2019)
• Keyence: Laser Marking on Die-Cast Components
• Foundry-Planet: Bühler: complete retraceability (March 19, 2008)
• Elsmar Cove Forum: IATF 8.5.2.1 – Identification and traceability
• 16949store: Identification and Traceability (IATF 16949 explanation)
• IATF: Stakeholder Communiqué SC-2026-005 IATF 16949 2nd Edition Status (July 2026)
• AIAG Blog: AIAG’s New Traceability Guideline Now in Stock (January 17, 2019)
• AIAG Blog: New Casting System Assessment CQI-27: Voice of the Customer (February 15, 2016)
• AIAG Blog: AIAG CQI-27 2nd Edition Now Available for Stakeholder Review and Commenting (December 14, 2017)
• AIAG Blog: 2nd Edition CQI-27 Casting System Assessment is Now Available (March 28, 2018)
• AFNOR: ASTM E505-22 Standard Reference Radiographs for Inspection of Aluminum and Magnesium Die Castings (sales page, December 2022)
• Cognex: AIM DPM (DataMan manual)
• Fraunhofer IIS: ISAR – intelligent system for automatic X-ray inspection
• Fraunhofer IIS: RoboCT
• Cincinnati Test Systems: How to Leak Test EV Cast Housings (May 16, 2024)
• Bühler: SmartCMS cell management system (viewed October 2026)
• Foundry-Planet: A bold new vision for the die casting industry: Bühler launches the Digital Cell (March 6, 2019)
• Foundry Management & Technology: Real-Time IIoT for Diecasting (April 25, 2022)
• Visi-Trak (viewed October 2026)
• Foundry-Planet: VDMA: die casting industry on the way to intelligent machine-to-machine communication using OPC UA (June 13, 2023)
• NHTSA: Part 573 Safety Recall Report 26V613, initial (filed September 24, 2026)
• NHTSA: Part 573 Safety Recall Report 26V613, amended (filed September 29, 2026)
• NHTSA: Part 573 Safety Recall Report 26V069 (filed February 5, 2026)
• Mahasarakham University PDPA site: Personal Data Protection Act B.E. 2562 (2019), unofficial English translation
• Suwanvara Law: PDPA for HR in Thailand (updated September 18, 2026)