Skip to content

Latest commit

 

History

History
68 lines (46 loc) · 2.98 KB

ReadMe.md

File metadata and controls

68 lines (46 loc) · 2.98 KB

This is the code for the paper "Adaptive Partially-Observed Sequential Change Detection and Isolation". The paper is accepted in Technometrics and the Arxiv version is provided here.

Code

  1. srpabstract.py: A class that do SRP statistics
  • Initialization:
    • p: Number of dimensions
    • c: scale vector, Target meanshift is c * M
    • k: Number of failuer Mode
    • M: Failure Mode Mean Matrix of k failure modes: p * k
    • nsensors: number of selected sensors
    • Ks: Number of selected failure mode
    • L: control limit, set to -1 if not initialized yet.

You need to override the following functions

  • compute_log_LRT: Compute the log liklihood ratio,
  • compute_index: Compute the index function to decide the best sensing allocation

Other implemented method:

  • compute_monitoring_statistics: Computed SRP statistics using the Top-R rules
  • compute_monitor_batch: Compute the monitoring results for batch of samples
  1. TSSRP.py: Implement original TSSRP method

  • srpabstract initialization except

    • mode: which testing statistics to use, Default to T2

  • compute_index: Sensor index becomes the failure mode index

  1. ExtendedTSSRP.py

  • srpabstract initialization except

    • mode: which testing statistics to use, Default to T2

  • compute_index

    • Mode T2 Default, using summation of log SRP.
    • Mode T1 , using summation of SRP, no closed forms and using greedy algorithm
    • Mode T1_Max , using summation of SRP, no closed formsusing Max approximation
  1. spc.py: A generic class for process monitoring using simulation

  • Initialization:

    • monitor_statistics: given input of n_batch * Tmax * dimensions, return the monitoring statistics of size n_batch * Tmax denote
    • data_gen_func0: Generate normal samples, return n_batch * Tmax * dimensions
    • data_gen_func1: Generate abnormal samples, return n_batch * Tmax * dimensions

  • phase1:

    • Generate iterations (= number of seeds) of each data with n_batch size
    • Use Binary search to find the control limit L where the ARL0 is fixed

  • phase2:

    • Generate iterations (= number of seeds) of each data with n_batch size
    • Return ARL1

Dataset

There are three datasets that used in the paper:

  1. The dataset is offered in temperature.mat. Thermal imaging monitoring of 3D printing: The dataset used is publically available at figshare Scenario 3. data
  2. Tonnage dataset, which is offered in tonnage.mat.
  3. COVID-19 monitoring of different counties in WA. The dataset is derived from the JHU. The WA dataset used in the paper has been also stored in Infection rate.