A benchmark for elevator fault diagnosis: Multi-level tasks from perception to diagnostic reasoning

Elevator fault diagnosis is a safety-critical problem that requires capabilities beyond isolated fault classification, including knowledge grounding and structured reasoning for maintenance support. In this work, we present a benchmark for elevator fault diagnosis and organize evaluation into three complementary task groups: (A) sensor-based fault detection and classification, (B) taxonomy-grounded fault classification and cause identification, and (C) reasoning with incomplete or constrained information. To enable reproducible comparison across heterogeneous model families, we standardize task definitions, input/output formats, and scoring protocols and report component-level predictive metrics. Experiments across representative language models show task-dependent performance: higher performance on one task does not necessarily transfer to another. The benchmark is a reproducible testbed for studying diagnostic-assistance methods, not a validation of safe operational deployment.

Authors

Institutions

Publication Details

Journal
PLoS ONE
Published
2026-11-18
DOI
https://doi.org/10.1371/journal.pone.0362972
Primary Topic
Elevator Systems and Control
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

A benchmark for elevator fault diagnosis: Multi-level tasks from perception to diagnostic reasoning

Qiang Zhang, Junjie Huang, Haozhe Sun, Xinda Wang
1 links
PLoS ONE
Elevator Systems and Control
article

A benchmark for elevator fault diagnosis: Multi-level tasks from perception to diagnostic reasoning

Qiang Zhang, Junjie Huang, Haozhe Sun, Xinda Wang
article en

Abstract

Elevator fault diagnosis is a safety-critical problem that requires capabilities beyond isolated fault classification, including knowledge grounding and structured reasoning for maintenance support. In this work, we present a benchmark for elevator fault diagnosis and organize evaluation into three complementary task groups: (A) sensor-based fault detection and classification, (B) taxonomy-grounded fault classification and cause identification, and (C) reasoning with incomplete or constrained information. To enable reproducible comparison across heterogeneous model families, we standardize task definitions, input/output formats, and scoring protocols and report component-level predictive metrics. Experiments across representative language models show task-dependent performance: higher performance on one task does not necessarily transfer to another. The benchmark is a reproducible testbed for studying diagnostic-assistance methods, not a validation of safe operational deployment.

PLoS ONEVol. 21(11)
Xi’an Jiaotong-Liverpool University (CN), Zhejiang University (CN)
Science and Technology Department of Zhejiang Province
Elevator Systems and Control
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon