industrial-ai-troubleshooting-agent: Command for Claude Code

MSPC Analysis

Run Multivariate Statistical Process Control analysis on the boiler historian data using PCA-based monitoring with T-squared and SPE statistics.

Conventions

• Python: Always use the project's virtual-environment Python, never bare python (which on macOS may not exist or may point at a dependency-free system Python). Written as <venv-python> below — substitute .venv/bin/python on macOS/Linux or .venv/Scripts/python on Windows.
• Opening files: Use the command for the user's OS, written as <open> below — open on macOS, xdg-open on Linux, start on Windows.

Steps

1. Run the analysis

Run the MSPC analysis script:

<venv-python> mspc_analysis.py

Capture the JSON output from stdout. The script prints progress to stderr and the JSON summary to stdout.

If the script fails, diagnose the error and report it to the user.

2. Open key plots and fault report

All outputs are saved to plots/MSPC/MSPC_<date>/. Open the monitoring chart, scree plot, and fault report for the user:

<open> <monitoring_plot_path>
<open> <scree_plot_path>
<open> <fault_report_path>

3. Present model summary

Report to the user:

• Number of PCA components retained and cumulative variance explained
• Training set size (NOC rows vs total rows)
• Control limits (T-squared and SPE at 95% and 99% confidence)

4. Present anomaly results

Report:

• Percentage of observations exceeding T-squared and SPE limits
• Number and timing of anomalous periods
• Duration of each anomalous period

5. Investigate worst events

For each of the top 3 worst T-squared events from the JSON output:

1. Open the contribution plot: <open> <contribution_plot_path>
2. Call alarm_search_context at that timestamp to get alarm context
3. Call kg_get_upstream_sensors on the top contributing tag to trace process causality
4. Summarize: what tags drove the anomaly, what alarms were active, what equipment is upstream

Do the same for the top 3 worst SPE events (skip any that overlap with T-squared worst events).