VIP20
本课程属于：005-Artificial Intelligence and Machine Learning Engineering English Technical Courses · 共 20 门相关课程，组合学习更完整。
查看所属VIP全部课程 →
← 返回所属目录 咨询 / 购买

Industrial Anomaly Detection and Unsupervised Machine Learning

30 Chapters
View Course Samples Online
01
Industrial anomaly detection landscape
definitions, failure modes, operational constraints, and the role of unsupervised learning.
Industrialdetection
02
Data acquisition in industrial settings
sensor types, sampling rates, synchronization, labeling scarcity, and dataset construction.
acquisitionindustrial
03
Exploratory data analysis for unlabeled industrial data
distribution profiling, temporal patterns, and anomaly candidate identification.
Exploratoryanalysis
04
Feature engineering for industrial signals
time-domain, frequency-domain, time-frequency, and domain-specific representations.
industrialtime-domain
05
Dimensionality reduction for anomaly detection
PCA, kernel PCA, autoencoders, and their industrial trade-offs.
Dimensionalityreduction
06
Density-based anomaly detection
k-NN, LOF, and isolation forest with practical parameter tuning on sensor data.
Density-baseddetection
07
Clustering-based anomaly detection
k-means, DBSCAN, GMM, and cluster validity for identifying abnormal operating regimes.
Clustering-baseddetection
08
One-class classification
OC-SVM, SVDD, and their application to machine condition monitoring.
One-classclassification
09
Reconstruction-based methods
autoencoders, variational autoencoders, and reconstruction error thresholds for anomaly scoring.
Reconstruction-basedautoencoders
10
Generative models for anomaly detection
GANs, normalizing flows, and likelihood-based scoring in industrial contexts.
Generativedetection
11
Deep representation learning for time-series anomaly detection
LSTM, GRU, temporal convolutional networks, and transformers.
representationtime-series
12
Self-supervised and contrastive learning for industrial anomaly detection without labels.
Self-supervised and contrastive learning for industrial anomaly detection without labels.
Self-supervisedcontrastive
13
Anomaly scoring and thresholding
statistical process control, extreme value theory, and operational decision boundaries.
thresholdingstatistical
14
Evaluation metrics for unsupervised anomaly detection
precision@k, AUC-ROC, AUC-PR, and domain-specific cost functions.
Evaluationunsupervised
15
Handling class imbalance and rare events
resampling, cost-sensitive learning, and anomaly ratio estimation.
Handlingimbalance
16
Concept drift and non-stationarity in industrial processes
detection, adaptation, and online learning strategies.
non-stationarityindustrial
17
Multivariate and multimodal anomaly detection
sensor fusion, correlation analysis, and joint embedding spaces.
Multivariatemultimodal
18
Scalable and distributed anomaly detection
streaming architectures, edge computing, and real-time inference constraints.
Scalabledistributed
19
Explainability and root cause analysis
feature attribution, counterfactuals, and diagnostic support for operators.
Explainabilityanalysis
20
Case study
anomaly detection in rotating machinery using vibration and acoustic signals.
detectionrotating
21
Case study
process industry monitoring with multivariate time-series and unsupervised methods.
industrymonitoring
22
Case study
semiconductor manufacturing fault detection using sensor traces and reconstruction models.
semiconductormanufacturing
23
Case study
power grid and energy systems anomaly detection with unsupervised learning.
detectionunsupervised
24
Case study
automotive and aerospace manufacturing quality control via image and signal anomaly detection.
automotiveaerospace
25
Deployment pipelines
model packaging, monitoring, retraining, and integration with SCADA/MES systems.
Deploymentpipelines
26
Robustness and adversarial considerations
sensor noise, missing data, and security of anomaly detection models.
Robustnessadversarial
27
Benchmarking and reproducibility
public datasets, evaluation protocols, and comparison of unsupervised methods.
Benchmarkingreproducibility
28
Human-in-the-loop workflows
active learning, operator feedback, and iterative model improvement.
Human-in-the-loopworkflows
29
Regulatory and safety standards
IEC 61508, ISO 13849, and compliance for AI-based industrial monitoring.
Regulatorystandards
30
Emerging directions
foundation models, federated learning, and edge AI for industrial anomaly detection.
Emergingdirections