返回

2023HUMS 数据挑战赛成果： LM-RMS-DAI团队

ID：72803

阅读量：1

大小：1.07 MB

页数：6页

时间：2025-01-03

金币：10

上传者：神经蛙1号
HUMS2023 Data Challenge Result Submission
Team Name: LM-RMS-DAI
Team Members: Nathaniel Rigoni, Daniel Wade
Institutions: Lockheed Martin Rotary and Missions Systems
Publishable: Yes
1. Summary of Findings
A neural network tuned for multiplexed time series Virtual Sensor (VS) anomaly detection is trained and
evaluated against a faulted epicyclic transmission in this paper. The VS is shown to provide good
estimation of fault presence and progression of an artificially induced planetary gear crack. The team used
three varieties of the same VS architecture: leave-one-out (VS-LOO), two-class (VS-2C), and unsupervised
operations clustering (VS-UOC). This paper focuses on the VS-LOO and VS-UOC.
Each aspect of the performance versus the fault detection and growth is tied to a combination of the VS
capabilities (Table 1). We claim detection (row 1) when the VS-LOO exceeds the threshold across multiple
readings. Due to the nature of our method, we do not claim a difference between multi and single sensor
detection (row 2, 3). We find that VS-LOO error increases as the fault progresses from the Day 22 minima
(row 4). We claim exponential growth detection and state change based on VS-UOC and VS-LOO (row 5).
Table 1 Summary of Analysis Results
#
Detection & Trending
Data file
name/number
Comments
1
Consistent detection on at least one signal channel; i.e. the
fault indicators remain consistently above the threshold.
Day025
20220111_100623
Figure 5,
reconstruction
error
consistently
greater than 40
2
Confirmed detection on at least two signal channels; i.e. the
fault indicators remain consistently above the threshold.
3
Clear multi-channel indication of the characteristic fault
features; i.e. faulty planet gear meshing with both the ring
and sun gears.
4
Confirmed trend of fault progression; i.e. a consistent
increasing trend started from which file number/name.
Day022
20211209_125436
Figure 5,
Error minima
5
Confirmed trend of accelerated fault progression; i.e. a
consistent exponential increasing trend started from which
file number/name
Day025
20220111_153659
Figure 5,
Cluster Change,
Error Change
资源描述：

当前文档最多预览五页，下载文档查看全文

此文档下载收益归作者所有

当前文档最多预览五页，下载文档查看全文
温馨提示：
1. 部分包含数学公式或PPT动画的文件，查看预览时可能会显示错乱或异常，文件下载后无此问题，请放心下载。
2. 本文档由用户上传，版权归属用户，天天文库负责整理代发布。如果您对本文档版权有争议请及时联系客服。
3. 下载前请仔细阅读文档内容，确认文档内容符合您的需求后进行下载，若出现内容与标题不符可向本站投诉处理。
4. 下载文档时可能由于网络波动等原因无法下载或下载错误，付费完成后未能成功下载的用户请联系客服处理。
最近更新
更多
大家都在看
近期热门

分享

购物车

关闭