Zhengyang

—

Unclaimed · AI-structured profilename-matched identity

Zhengyang — first-author industrial AI researcher in deep RL control and predictive maintenance

papers here

reviewed by Evolve

first-author

corresponding

—

works · OpenAlex

—

h-index · OpenAlex

Bibliometric figures come from OpenAlex (2026-10-13) and cover the author's entire public record, not just the papers listed here. Reproducibility signal comes from Evolve's own reviews.

Research profile

Output—

Not scored: no publication-volume data in this snapshot beyond two reviewed 2026 preprints — no totals, venue record, or yearly counts. This axis would measure research productivity: number of outputs, where they appeared, and cadence over time.

Impact—

Not scored: the snapshot has no citation counts, h-index, or uptake metrics. This axis would measure scholarly influence — citations and adoption of the researcher's methods by other groups — and is not estimable from two brand-new preprints.

Focus0/100

Scored 0/100. The field distribution is an exact 50/50 split between 'machine learning (industrial AI / predictive maintenance)' and 'reinforcement learning for industrial control', derived from the two cataloged 2026 papers. With only two outputs sitting in two distinct subfields, no dominant specialty has emerged yet; the connective theme is applied industrial AI, but with a two-paper sample this axis should be read as early-stage rather than as evidence of scattered interests.

Collaboration46/100

Scored 46/100, computed from 9 distinct coauthors in the bibliometric record. The recurring roster — Joseph E. Hernandez, Thomas Cook, John Burtenshaw, Sean Scott, Chris Couch — suggests a compact, stable team, while both 2026 preprints are first-author and non-corresponding. The raw list also carries split tokens ('Zhengyang', 'Gu') and a duplicated 'Thomas Cook', so entity-resolution noise may shift the true network size in either direction.

Activity—

Not scored: the snapshot provides no per-year publication activity beyond two 2026 preprints — no yearly output counts or gaps. This axis would measure research tempo: outputs per year, recent momentum, and any inactive periods across the career timeline.

Reproducibility38/100

Scored 38/100 — the mean repro-readiness across 2 assessed papers (reproPapers: 2). Evolve rates the PID-deep RL hybrid preprint 'low', citing single-episode evaluation, no significance tests, and an unproven MIMO premise, and the predictive-maintenance benchmark 'medium', citing 700+ runs but synthetic linear targets and absent statistical testing. Missing significance tests in both studies are the common thread behind this being the weakest evidenced axis.

Research summary

Zhengyang's cataloged record so far consists of two 2026 first-author preprints across two branches of applied industrial AI. The profile is name-resolved from public bibliographic data; the coauthor roster contains split tokens ('Zhengyang', 'Gu'), hinting the surname Gu, but no affiliation, country, or ORCID is attached. Both outputs are arXiv preprints: one augments PID control with deep reinforcement learning, pairing a classical loop with TD3-discovered Gain/Shift parameters on the Industrial Benchmark, where the hybrid reportedly matches the best DRL agent (-203.46 vs -204.92); the other benchmarks hybrid LSTM-Transformer architectures for Industry 4.0 predictive maintenance over 700+ runs, claiming a noise-robustness advantage. Evolve's reviews credit scale and ambition but flag methodological gaps — single-episode evaluation and no significance tests in the control paper; synthetic linear targets and missing linear baselines in the benchmark — yielding a mean reproducibility score of 38/100 across the two papers, with readiness rated low and medium. The bibliometric field split is exactly 50/50 between reinforcement learning for industrial control and machine learning for predictive maintenance, producing a focus score of 0; collaboration scores 46/100 on 9 distinct coauthors. Output, impact, and activity remain unscored because the snapshot holds no citation, volume, or yearly-timeline data.

Collaboration pattern

Collaboration scores 46/100 from 9 distinct coauthors. The frequent list — Joseph E. Hernandez, Thomas Cook, John Burtenshaw, Sean Scott, Chris Couch — suggests a compact recurring team, and both 2026 preprints are first-author (non-corresponding) works. The raw roster also contains split tokens ('Zhengyang', 'Gu') and a duplicated 'Thomas Cook', indicating name-parsing noise, so the true collaborator count may differ. No senior-collaborator or corresponding-author evidence is available in this snapshot.

Research keywords

deep reinforcement learningPID controlIndustrial BenchmarkTD3hybrid controlreward shapingpredictive maintenanceIndustry 4.0LSTMTransformerhybrid deep learningnoise robustness

Highlights

Hybrid PID–TD3 controller matches deep RL on Industrial Benchmark
First-author 2026 arXiv preprint pairs a classical PID loop with TD3-discovered Gain/Shift parameters; Evolve's verdict reports the hybrid matching the best DRL agent (-203.46 vs -204.92), while flagging single-episode evaluation, absent significance tests, and an unproven MIMO premise. Reproducibility readiness: low.
700+ run benchmark of LSTM–Transformer hybrids for predictive maintenance
First-author 2026 arXiv preprint benchmarks hybrid deep learning architectures for Industry 4.0 predictive maintenance, claiming an LSTM-Transformer noise-robustness win across 700+ runs; Evolve notes the evidence is weakened by synthetic linear targets, missing significance tests, and absent linear baselines. Reproducibility readiness: medium.

Timeline

2026Posts a first-author arXiv preprint augmenting PID control with deep reinforcement learning, evaluated on the Industrial Benchmark.
2026Publishes a first-author benchmark of hybrid deep learning architectures for Industry 4.0 predictive maintenance, spanning 700+ runs.
2026Both preprints enter the Evolve review catalog with reproducibility readiness rated low and medium.

Papers in the Evolve catalog

2026-09-18first author
Augmenting PID Control with Deep Reinforcement Learning: A Hybrid Approach to the Industrial Benchmark
2026-09-18first author
Benchmarking Hybrid Deep Learning Architectures for Predictive Maintenance in Industry 4.0

Frequent co-authors

Zhengyang×2 Gu×2 Joseph E. Hernandez×2 Thomas Cook×2 John Burtenshaw×2 Sean Scott×2 Chris Couch×2 Gu×2 Thomas Cook×2

About this profile

This profile is AI-structured by Evolve.Pub from public bibliographic data (OpenAlex) and Evolve's own paper reviews. It is not written or endorsed by the researcher unless the profile is claimed. Radar dimensions are computed from real bibliometrics with fixed formulas — raw figures are shown alongside; missing dimensions mean insufficient data, never zero. Errors? Claim the profile to correct or hide it.

Bibliometrics are computed from OpenAlex-style public bibliographic data only; no citation counts, h-index, or venue history are present, so output, impact, and activity remain unscored.
Identity is name-resolved from the given name 'Zhengyang' alone — no ORCID, affiliation, or country is attached, and the coauthor roster contains split name tokens ('Zhengyang', 'Gu'), so person-level disambiguation is uncertain.
The evidence base is two 2026 arXiv preprints reviewed by Evolve; no peer-reviewed publications, funding, or award data are included.
Reproducibility, focus, and collaboration scores rest on 2 assessed papers and 9 coauthor records — a very small sample that may shift as the catalog grows.