Chubu Electric Power built its case for restarting two reactors at the Hamaoka nuclear plant on earthquake calculations that, for years, did not do what the company told regulators they did. An independent special investigation committee, in a report the utility published on September 14, found the company had not run a single clean process for selecting the seismic waveforms used to set the plant's design-basis ground motion.
The committee's language is careful. It said it was not finding the conduct to have been appropriate, but it explicitly declined to simplify its findings using words like falsification, fabrication, underestimation, or data fraud, and it refused to produce a summary version of its report, arguing the full document must be read as a whole. The underlying findings are nonetheless specific. In at least 105 of 225 waveform-selection cases reviewed, engineers built many candidate sets of 20 simulated earthquake waveforms and then chose whichever set produced the representative wave they wanted, rather than the single set the company had told the Nuclear Regulation Authority it used. In a smaller number of cases tied to a contested fault near the plant, engineers went further: they picked a wave that was not close to the statistical average of the set, then swapped out the other 19 waveforms until the chosen one appeared to be the best fit after all.
| Practice | What happened | Scope found so far |
|---|---|---|
| Multiple-set picking | Built many candidate sets of 20 waveforms, then chose whichever set gave the preferred result, instead of the single set described to regulators | At least 105 of 225 cases, from 2012 onward |
| Non-average pick with 19-wave swap | Selected a wave that was not the statistical average, then swapped the other 19 waveforms until the chosen one looked like the best statistical fit | At least 3 confirmed cases in 2018, with related practices in more |
| Retrospective data-set creation (after Method 2) | Created after-the-fact calculation sets to make earlier non-average picks appear statistically justified once regulators asked for evidence | 6 of 7 cases using the non-average selection method |
| Retrospective data-set creation (2025 inquiry response) | Reconstructed data specifically in response to the 2025 regulatory inquiry | 68 of all 225 cases |
When regulators pressed for evidence in 2025, the company did not come clean immediately. According to the investigation, staff in the nuclear civil engineering department created backdated calculation sets designed to make the earlier waveform picks look statistically justified, a practice the report catalogs across dozens of the 225 cases: at least 58 of 63 non-average selections got this retrospective patch job, and 68 of all 225 cases saw some form of after-the-fact data reconstruction tied to the 2025 regulatory inquiry. Internal staff had flagged the mismatch between what was shown to regulators and what was actually done as early as 2018, but the company did not correct its review materials.
The stakes for Chubu Electric are large. Hamaoka's three shuttered reactors have cost the company roughly ¥240bn a year in lost financial impact at current fuel prices, and the utility has spent about ¥1.5tn maintaining the idle plant and ¥280bn on safety upgrades since the 2011 shutdown. The Nuclear Regulation Authority has already said it will not proceed with the pending restart review for units 3 and 4 given the credibility problems in the underlying application, effectively freezing the process the company had been racing to complete.
The company's own account, released alongside the investigation, apologizes to stakeholders and promises what it calls a "dismantling and rebuilding" of governance and organizational culture. What it does not offer is a clear statement on whether Hamaoka's actual seismic risk was ever properly assessed. The investigation itself says that question, whether honest calculations would have produced a different, and possibly larger, design-basis earthquake, remains unresolved.