Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Krisztián Horváth has a total of 9 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Production Gear Profile and Lead Variability: A Metrology Analysis and Hypotheses for Future Rolling-Test Vibration Validation

    by Krisztian Horvath, Ambrus Zelei

    Production gear inspection reports standard profile and lead deviations, but the amount of information retained in the dispersion of the displayed values beyond scalar magnitude is less clear. This study analyzes 65 inspection reports from 59 production helical gears with 23 teeth, comprising 4680 audited cell values; 4248 values from the latest report per component define the primary population. Six standard report parameters were separated from author-defined summaries of pooled magnitude, sampled dispersion, concentration, and flank asymmetry. Lead-form deviation ffβ showed the strongest magnitude–variability association: magnitude versus mean trace SD gave Spearman ρ = 0.738 (95% component-bootstrap CI 0.544–0.868), and magnitude versus normalized variability gave ρ = 0.675 (0.450–0.832). Total profile deviation Fα magnitude versus SD gave ρ = 0.618 (0.399–0.785). An empirical scale-coupling null indicated that the principal associations exceeded those expected from randomly pairing observed magnitudes with observed relative-shape factors. However, the six displayed positions are not verified sequential teeth; the localization statistic is sensitive to individual readings, and no exact same-part rolling-test records are available. The vibration-related mappings are therefore testable hypotheses for future matched-part validation, not demonstrated geometry-to-vibration transfer.

  • Open Access

    Sensor-Position Domain Shift in Gearbox Fault Diagnosis: Healthy-Reference Calibration and Frequency-Localized Transfer-Path Sensitivity

    by Krisztián Horváth

    Sensor relocation can induce structural domain shift because gearbox vibration depends on fault excitation and the transmission path. This study evaluates transferability using 180 source recordings representing ten states, six accelerometer positions, and three coupled operating conditions. In the independently reproduced 170-feature ExtraTrees evaluation, raw nine-fault balanced accuracy was 43.8% with a held-out sensor position versus 84.6% with a held-out operating condition; local healthy referencing increased cross-position accuracy to 61.7% but requires a target-position healthy baseline. Fault-by-position interaction increased from 9.7% below 2 kHz to 25.3% at 5–10 kHz. The original non-nested bandwidth sweep found a raw gear-fault model-mean maximum of 84.4% below 1 kHz (62.6% for 0–10 kHz). New nested source-position-only cutoff selection yielded 82.2% for gear faults versus 62.6% over the full band, selecting cutoffs between 1 and 2 kHz across folds and classifiers. Raw bearing-fault accuracy did not benefit consistently from nested cutoff selection. The findings establish a within-rig baseline for frequency-localized structural portability rather than a universal frequency limit or new classifier; independent machine-level validation remains necessary.

  • Open Access

    Geometry-Anchored Paired Component-to-System Order Analysis with Falsification-Controlled Interpretation for End-of-Line Noise, Vibration, and Harshness of Electric-Vehicle Drivetrains

    by Krisztian Horvath

    Component-level vibration tests are performed before final assembly, but how their order-domain signatures relate to later drive-unit end-of-line (EOL) noise, vibration, and harshness (NVH) remains uncertain. This study links two pre-assembly component-test states to four EOL channels for 107 electric-drivetrain units using exact genealogy. The paired state-difference spectrum, Δ(B − A), identified contiguous shaft-synchronous regions associated with later EOL nonconformity, while high-order associations to the fourth mesh-family harmonic (H4) formed an approximately one-order-spaced modulation comb. A four-feature topology model achieved area under the receiver operating characteristic curve (ROC-AUC) of 0.742–0.753 across random and process-grouped validation, close to PCA(10) on the full Δ-spectrum (0.785–0.788). A falsification-controlled kinematic audit rejected an attractive grid-derived carrier assignment. A secondary archive of 3157 machine-readable coordinate-metrology records provided a geometry anchor: in a predeclared 30-day one-to-one subset, the sixth circumferential pitch-error harmonic of the 28-tooth component family was strongly associated with the corresponding State-B symmetric sideband family (n = 14, Spearman ρ = −0.908, and FDR-adjusted q = 8.8 × 10−5) and remained directionally consistent across traceability and timing sensitivity analyses. Direct geometry-to-EOL H4 associations did not survive multiplicity correction. The results support staged geometry-to-component and component-to-EOL relationships without claiming complete same-unit causal mediation.

  • Open Access

    Spatial  Harmonic Analysis of Production Gear Pitch Reports with Accumulation Weighting, Digitization Sensitivity, and Population Descriptors

    by Krisztian Horvath

    Production pitch inspection retains circumferential traces, but scalar tolerance summarizes and discards their spatial organization. This study evaluates a report-based metrology workflow on 65 raster inspection reports from 59 unique z = 23 gears. Analytical accumulation weighting, circular first-differencing and the directly reported single-pitch trace provide distinct checks of representation dependence. A fully specified extraction rerun and 17 alternative settings assess waveform, sampling, and overlay sensitivity. In the revised baseline, the lowest two harmonics dominate 54/59 direct-pitch spectra, although exact descriptor values and component rankings depend on baseline exclusion and trace selection. The five-descriptor Spearman matrix and principal component analysis show substantial redundancy: two components explain 92.0% of standardized variance and the participation-ratio dimension is 1.81. Representation and descriptor ablations quantify information retained by reduced summaries. Descriptor calculation is inexpensive relative to report decoding, but timing does not establish production readiness. The contribution is a reproducible analysis of reported pitch topology; manufacturing, clamping, measurement and correction effects cannot be separated without native numerical data and controlled metrology records.

  • Open Access

    Heat Treatment and Manufacturing Variability in Gearbox NVH : An Evidence‐Aware Critical Review and Validation‐Oriented Modeling Roadmap

    by Krisztián Horváth, Daniel Feszty

    Gear whine arises from mesh excitation, structural transfer paths, and acoustic radiation. Heat treatment and manufacturing determine the final tooth geometry, residual‐stress state, surface integrity, hardness profile, and unit‐to‐unit variability entering this chain. This structured critical review maps evidence linking phase transformation, heat‐treatment distortion, retained austenite, finishing, material damping, manufacturing scatter, and nonuniform wear of case‐hardened tooth flanks to transmission error (TE), vibration, and noise. Direct NVH evidence is distinguished from indirect evidence based on material state, geometry, durability, or modal behavior. The strongest evidence supports a process and finishing → final measured geometry and surface state → static or dynamic TE and dynamic force → structural response → sound pathway, whereas direct phase‐fraction or hardness‐to‐noise links in conventional case‐hardened steel remain weak. The review applies explicit source‐prioritization criteria and distinguishes representative evidence mapping from exhaustive systematic screening. An evidence‐derived modeling and validation roadmap is presented as a literature synthesis rather than a separately validated all‐in‐one model: final measured geometry is the baseline input, while thermometallurgical and wear‐evolution modules are activated only when their effects are independently characterized. The review additionally translates manufacturing variability into process‐capability‐informed geometric dimensioning and tolerancing (GD&T) for noise‐sensitive design.

  • Open Access

    From Gear Metrology to End-of-Line Order Topology: EOL-Blind Geometry Anchoring and Falsification-Based Validation in an Electric-Vehicle Reducer

    by Krisztian Horvath

    Manufacturing-induced gear deviations can generate ghost orders, but direct validation from measured geometry to assembled-drive end-of-line (EOL) vibration remains scarce. This study tests an EOL-blind, falsification-oriented workflow for an electric-vehicle reducer. Repeated coordinate-measuring-machine reports of one 24-tooth gear yielded tooth-indexed pitch deviations with a dominant tenth circumferential spatial harmonic (k = 10) on both flanks (spectral-energy cosine similarity 0.952–0.994). Before EOL amplitudes were inspected, k = 10 was frozen and mapped to a sideband-order offset of ΔO = 3.29125; adjacent k = 9 and 11 harmonics and ±0.25-order shifted families served as controls. Industrial testing included 8974 spectra from 8232 unique units. Across mesh harmonics p = 1–6, median symmetric prominence was 0.485, 0.750, and 1.341 dB for k = 9, 10, and 11, respectively. The p = 1–15 analysis suggested a k = 10 advantage over shifted controls, but the eight-bin shift coincided with the local-background boundary and cannot establish specificity. The k = 11 family remained stronger. Grouped holdout against a 31.625-order upper-tail endpoint yielded ROC AUC = 0.488. Independently measured geometry can define an a priori order family, but repeatable spatial harmonics are not necessarily unique or predictive assembled-drive noise, vibration, and harshness (NVH) markers.

  • Open Access

    Vízi és víz alatti járművek csatolt vibroakusztikai modellezése

    by Krisztián Horváth

    A víz alatti kisugárzott zaj (underwater radiated noise, URN) predikciója a haditechnikai alkalmazásokban és a tengeri környezet védelmében egyaránt stratégiai jelentőségű. A szakirodalom és a szabványosítási gyakorlat alapján az URN nem értelmezhető pusztán hidrodinamikai jelenségként: a gerjesztések, a szerkezeti rezgések és a vízterhelés csatolt vibroakusztikai multifizikai (multiphysics) rendszerként határozzák meg a távoli térben mérhető hangnyomásszintet. Az áttekintés a víz alatti hangterjedés releváns fizikai alapjait, az URN fő forrásmechanizmusait (propellerkavitáció, szerkezet által terjesztett zaj, áramlás okozta zaj), a csatolt modellezési módszereket (strukturális végeselemes modell, folyadék–szerkezet kölcsönhatás, CFD-alapú áramlásmodellezés, akusztikai analógia, FEM–BEM csatolás), továbbá a prediktív modellek hierarchiáját és a számítási költség–pontosság kompromisszumait tárgyalja. Központi elemként részletesen bemutatok egy, URN-predikcióra alkalmas vibroakusztikai multifizikai munkafolyamatot (workflow), valamint a validáció és a tengeri mérések bizonytalanságait az ISO 17208 szabványsorozat kontextusában.

  • Open Access

    Industrial Component Rolling-Test Order Spectra with Paired Mixed, Synchronous, and Master Exports

    by Krisztián Horváth

    Industrial rolling tests produce large collections of component-level vibration spectra, but their reuse depends on preserving measurement relationships and distinguishing repeat tests from repeated exports. This Data Descriptor presents 13 anonymized industrial component rolling-test Parquet exports acquired between January and April 2025, comprising 651,178 export rows and 645,951,488 spectral ordinates. The collection includes mixed and synchronous specimen spectra, maximum- and mean-labeled exports, and synchronous master spectra. Complete pseudonymous identifiers distinguish 54,920 Population A components and 54,396 Population B components. The workflow preserves native order axes, removes redundant exports, validates event-level joins, and selects the earliest valid event per component and channel. Across four representative channels, median coordinate-wise standard deviations range from 2.67 to 4.57 dB, and median central 90% spans range from 8.61 to 9.65 dB. These results describe observed component-to-component variability under production rolling-test conditions; undetermined bench repeatability and reproducibility prevent separation of manufacturing, measurement, fixturing, and operating contributions. Quality checks identify 42 rows without component identifiers and one nonfinite spectral ordinate. An illustrative bearing-order analysis shows that restricting mixed spectra to shared coordinates changes a narrow-window flag rate from 94.76% to 0.98%; these flags are methodological diagnostics, not verified fault labels. The dataset supports descriptive spectral analysis, processing comparisons, and leakage-aware model development, subject to the documented acquisition and metadata limitations.


  • Open Access

    A Large-Scale Industrial End-of-Line Vibroacoustic Dataset for Electric Drive Units Under a Variable-Speed Braking Condition

    by Krisztian Horvath

    End-of-line vibroacoustic testing identifies abnormal noise and vibration before an electric drive leaves production, yet public datasets rarely capture production scale, heterogeneity, repeated testing, and multiple signal domains. This Data Descriptor presents 58,249 processed measurements from 54,505 pseudonymized electric drive units acquired in one generalized variable-speed braking state. The release covers two anonymized product variants, four test benches, and 13 validated signal matrices on fixed-frequency, synchronous-order, and mixed-order grids. The measurements represent the complete industrial EOL measurement environment rather than metrologically isolated component radiation. Technical validation therefore separates data-file integrity from production-level statistical behavior. At mixed order 7.875, raw standard deviations were 7.99 dB for gearbox-housing acceleration and 5.59 dB for sound pressure; variant-and-bench stratification reduced the pooled within-variant-and-bench values to 1.61 and 1.16 dB. Across nonzero fixed-frequency bins, the corresponding medians were 3.15 and 1.67 dB. Consecutive retests showed 95th-percentile absolute changes of 2.23 and 1.76 dB, but these retests are not controlled measurement-repeatability trials. Calibration histories, controlled sensor reattachment tests, and acoustic background spectra were not available in the research export and are explicitly treated as limitations. Stable public measurement and specimen identifiers preserve join and retest relationships while protecting original industrial identifiers.

  • Open Access

    From Benchmark Accuracy to Industrial Evidence: A Registered Systematic Map and Nested Experiment-Level Audit of Validation Credibility in AI-Based Rotating-Machinery Fault Diagnosis

    by Krisztian Horvath

    Artificial intelligence (AI) methods for rotating-machinery diagnosis based on vibration and acoustic signals are often evaluated on benchmark splits that do not establish physical independence or industrial generalization. This registered study combined a PRISMA-aligned systematic map with a nested experiment-level audit of validation credibility. Searches identified 11,916 records, retained 7520 unique records, and placed 5159 reports in the full-text population. The operational 500-record Batch 005 block produced a non-probability 351-report full-text queue. A post-review eligibility-consistency check yielded 97 included studies, 13 exclusions, 241 unresolved reports, and 157 validation experiments. Within this access-influenced subset, 130 experiments (82.8%) remained at A0–A1 physical independence, 27 (17.2%) reached A2–A4, and 41 (26.1%) evaluated B2–B4 domain shifts. Under a narrower leakage definition, positive evidence supported High risk in 19 experiments (12.1%), whereas 93 (59.2%) were Unclear because reporting was insufficient. Seventy-seven experiments reported classification-like numeric headline scores, but metric types were heterogeneous and were stratified rather than pooled. Metadata diagnostics confirmed source-structured selection and access-related differences between resolved and unresolved reports. The Industrial Evidence Matrix provides a structured vocabulary for validation evidence; all proportions describe the nested audit, not the complete literature.