Subtopic Deep Dive

← Advanced Statistical Process Monitoring

Change-Point Detection
Research Guide

What is Change-Point Detection?

Change-Point Detection identifies the exact time of abrupt shifts in statistical process parameters like mean or variance within Advanced Statistical Process Monitoring.

Researchers apply CUSUM, EWMA, and change-point methods to detect shifts in autocorrelated and profile data (Woodall 2007, 313 citations; Mahmoud et al. 2006, 280 citations). These techniques extend to multivariate and nonstationary processes (Liu et al. 1999, 643 citations; Zhao and Huang 2017, 273 citations). Over 50 papers since 1999 address surveillance and fault detection applications.

Curated Papers

Key Challenges

Why It Matters

In healthcare, change-point methods enable rapid outbreak detection via prospective surveillance (Sonesson and Bock 2003, 312 citations). Photovoltaic systems use these for fault localization, reducing downtime (Garoudja et al. 2017, 270 citations). Chemical processes benefit from cointegration-based monitoring of nonstationary dynamics, minimizing production losses (Zhao and Huang 2017, 273 citations). Precise localization supports root cause analysis in high-throughput screening.

Key Research Challenges

Autocorrelated Process Shifts

Detecting mean shifts in AR processes requires GLRT over standard Shewhart charts (Apley and Shi 1999, 142 citations). Standard CUSUM fails due to serial correlation. Methods must estimate shift size and location simultaneously.

Profile Data Change Points

Linear profiles need segmented regression for parameter constancy tests (Mahmoud et al. 2006, 280 citations). Multiple profiles complicate shift localization. Phase I analysis demands historical data stability.

Nonstationary Multivariate Monitoring

Cointegration and slow feature analysis handle time-variant chemical processes (Zhao and Huang 2017, 273 citations). Data depth measures outlyingness in multivariate samples (Liu et al. 1999, 643 citations). Full-condition monitoring challenges fault isolation.

Essential Papers

Multivariate analysis by data depth: descriptive statistics, graphics and inference, (with discussion and a rejoinder by Liu and Singh)

Regina Y. Liu, Jesse M. Parelius, Kesar Singh · 1999 · The Annals of Statistics · 643 citations

A data depth can be used to measure the “depth” or\n“outlyingness” of a given multivariate sample with respect to its\nunderlying distribution. This leads to a natural center-outward ordering of th...

Current research on profile monitoring

William H. Woodall · 2007 · Production · 313 citations

In many applications the quality of a process or product is best characterized and summarized by a functional relationship between a response variable and one or more explanatory variables. Profile...

A Review and Discussion of Prospective Statistical Surveillance in Public Health

Christian Sonesson, David Bock · 2003 · Journal of the Royal Statistical Society Series A (Statistics in Society) · 312 citations

Summary A review of methods suggested in the literature for sequential detection of changes in public health surveillance data is presented. Many researchers have noted the need for prospective met...

A change point method for linear profile data

Mahmoud A. Mahmoud, Peter A. Parker, William H. Woodall et al. · 2006 · Quality and Reliability Engineering International · 280 citations

Abstract We propose a change point approach based on the segmented regression technique for testing the constancy of the regression parameters in a linear profile data set. Each sample collected ov...

A full‐condition monitoring method for nonstationary dynamic chemical processes with cointegration and slow feature analysis

Chunhui Zhao, Biao Huang · 2017 · AIChE Journal · 273 citations

Chemical processes are in general subject to time variant conditions because of load changes, product grade transitions, or other causes, resulting in typical nonstationary dynamic characteristic. ...

Statistical fault detection in photovoltaic systems

Elyes Garoudja, Fouzi Harrou, Ying Sun et al. · 2017 · Solar Energy · 270 citations

Introduction to Statistical Process Control

Peihua Qiu · 2013 · 170 citations

Introduction Quality and the Early History of Quality Improvement Quality Management Statistical Process Control Organization of the Book Basic Statistical Concepts and Methods Introduction Populat...

Reading Guide

Foundational Papers

Start with Liu et al. (1999, 643 citations) for multivariate data depth concepts; Woodall (2007, 313 citations) for profile monitoring overview; Qiu (2013, 170 citations) for SPC basics including change-point ideas.

Recent Advances

Zhao and Huang (2017, 273 citations) on nonstationary chemical processes; Garoudja et al. (2017, 270 citations) on PV fault detection; Apley and Shi (1999, 142 citations) GLRT remains relevant for autocorrelation.

Core Methods

Segmented regression (Mahmoud et al. 2006); GLRT for AR processes (Apley and Shi 1999); cointegration with slow features (Zhao and Huang 2017); CUSUM/EWMA (Noyez 2009).

How PapersFlow Helps You Research Change-Point Detection

Discover & Search

Research Agent uses citationGraph on Woodall (2007) to map 313-cited profile monitoring works, then findSimilarPapers for change-point extensions in autocorrelated data. exaSearch queries 'CUSUM change-point autocorrelated processes' to surface Apley and Shi (1999). searchPapers with 'nonstationary change-point detection' retrieves Zhao and Huang (2017).

Analyze & Verify

Analysis Agent applies runPythonAnalysis to simulate GLRT on AR(1) data from Apley and Shi (1999), verifying ARL performance with NumPy/pandas. readPaperContent extracts segmented regression formulas from Mahmoud et al. (2006); verifyResponse (CoVe) with GRADE grades shift detection claims against empirical results. Statistical verification confirms CUSUM sensitivity via Monte Carlo runs.

Synthesize & Write

Synthesis Agent detects gaps in autocorrelated profile monitoring post-Woodall (2007), flagging underexplored variance shifts. Writing Agent uses latexEditText for CUSUM equations, latexSyncCitations for 10-paper bibliography, and latexCompile for camera-ready review. exportMermaid generates change-point timeline diagrams.

Use Cases

"Simulate GLRT vs CUSUM for AR(1) mean shift detection at tau=50"

Research Agent → searchPapers(Apley Shi 1999) → Analysis Agent → runPythonAnalysis(AR1 simulation, ARL computation, matplotlib plot) → matplotlib figure of run-length distributions.

"Write LaTeX review of change-point methods in profile monitoring"

Research Agent → citationGraph(Woodall 2007) → Synthesis → gap detection → Writing Agent → latexEditText(intro), latexSyncCitations(5 papers), latexCompile → PDF with equations and citations.

"Find GitHub repos implementing change-point detection for SPC"

Research Agent → searchPapers(change-point SPC) → Code Discovery → paperExtractUrls → paperFindGithubRepo → githubRepoInspect → List of 3 repos with CUSUM/GLRT code examples.

Automated Workflows

Deep Research workflow scans 50+ papers via searchPapers on 'change-point detection process monitoring', producing structured report with citation clusters around Woodall (2007). DeepScan applies 7-step analysis: readPaperContent on Zhao (2017) → runPythonAnalysis(cointegration test) → CoVe verification → GRADE scoring. Theorizer generates hypotheses for Bayesian change-points from Sonesson and Bock (2003) surveillance methods.

Try Doxa for Change-Point Detection Research

Frequently Asked Questions

What is Change-Point Detection?

Change-Point Detection locates the exact time of parameter shifts like mean or variance in sequential data. It uses methods like segmented regression (Mahmoud et al. 2006).

What are key methods?

GLRT for autocorrelated processes (Apley and Shi 1999); data depth for multivariate outliers (Liu et al. 1999); CUSUM/EWMA in profiles (Woodall 2007).

What are key papers?

Liu et al. (1999, 643 citations) on data depth; Woodall (2007, 313 citations) on profile monitoring; Mahmoud et al. (2006, 280 citations) on linear profile change points.

What are open problems?

Nonstationary multivariate shifts (Zhao and Huang 2017); risk-adjusted charts for heterogeneous data (Grigg and Farewell 2004); real-time localization in high-dimensional processes.