Guides

From raw signals to production KPIs — follow the data through the plant.

---

Plant Data Pipeline

Every manufacturing analysis follows the same flow. ts-shape mirrors this with dedicated modules at each stage.

flowchart LR
    subgraph ACQ["<b>Data Acquisition</b>"]
        direction TB
        A1["Parquet / S3 / Azure"]
        A2["TimescaleDB"]
        A3["Metadata JSON"]
    end

    subgraph COND["<b>Signal Conditioning</b>"]
        direction TB
        B1["Numeric Filter"]
        B2["DateTime Filter"]
        B3["Timezone Shift"]
        B4["Calculations"]
    end

    subgraph FEAT["<b>Feature Extraction</b>"]
        direction TB
        C0["Cycle / Segment Cutting"]
        C0b["Feature Tables"]
    end

    subgraph ANA["<b>Signal Analytics</b>"]
        direction TB
        C1["Statistics"]
        C2["Cycle Extraction"]
        C3["Pattern Recognition"]
    end

    subgraph EVT["<b>Event Detection</b>"]
        direction TB
        D1["Quality & SPC"]
        D2["Production Monitoring"]
        D3["OEE & Plant Analytics"]
        D4["Traceability"]
        D5["Process Engineering"]
    end

    subgraph RPT["<b>Reporting</b>"]
        direction TB
        E1["Shift Handover"]
        E2["Period Summary"]
        E3["Target Tracking"]
    end

    ACQ --> COND --> FEAT --> ANA --> EVT --> RPT

    style FEAT fill:#1a3a4a,stroke:#f59e0b,color:#fef3c7

    style ACQ fill:#1a3a4a,stroke:#2dd4bf,color:#e0f2fe
    style COND fill:#1a3a4a,stroke:#2dd4bf,color:#e0f2fe
    style ANA fill:#1a3a4a,stroke:#2dd4bf,color:#e0f2fe
    style EVT fill:#1a3a4a,stroke:#2dd4bf,color:#e0f2fe
    style RPT fill:#1a3a4a,stroke:#2dd4bf,color:#e0f2fe

---

Guide Index

Pick the stage that matches where you are in your analysis.

• Data Acquisition

  ---

  Connect to historians, data lakes, and metadata stores. Load Parquet, S3, Azure Blob, or TimescaleDB into DataFrames.

  ParquetLoader | AzureBlobParquetLoader | S3ProxyParquetLoader | MetadataLoader | DataIntegratorHybrid

• Signal Conditioning

  ---

  Clean and prepare raw signals. Filter by range, time window, pattern, or boolean flag. Convert timezones and compute derived values.

  NumericFilter | DateTimeFilter | StringFilter | TimezoneShift | NumericCalc

• Feature Extraction

  ---

  Cut timeseries into repeatable units (cycles or segments), then build feature tables with statistical profiles per unit. Discrete parts or continuous orders.

  CycleExtractor | SegmentExtractor | SegmentProcessor | DescriptiveFeatures | ProfileComparison

• Pipeline Builder

  ---

  Chain any ts-shape class into a single reproducible pipeline. Understand add_step vs add_instance_step, wire DataFrames with sentinels, and debug with describe() and run_steps().

  FeaturePipeline | add_step | add_instance_step | $prev | $input

• Signal Analytics

  ---

  Compute descriptive statistics, extract process cycles, and detect patterns across signals.

  NumericStatistics | TimestampStatistics | CycleExtractor | PatternRecognition

• Quality Control & SPC

  ---

  Outlier detection (Z-score, IQR, MAD, IsolationForest), Western Electric Rules, CUSUM shifts, and Cp/Cpk capability indices.

  OutlierDetection | StatisticalProcessControl | ToleranceDeviation

• Production Monitoring

  ---

  Machine states, line throughput, changeovers, downtime, cycle times, shift reporting, and quality tracking on the shop floor.

  MachineStateEvents | DowntimeTracking | ShiftReporting | CycleTimeTracking | +5 more

• OEE & Plant Analytics

  ---

  Overall Equipment Effectiveness, alarm management (ISA-18.2), batch tracking, bottleneck detection, micro-stops, and duty cycles.

  OEECalculator | AlarmManagementEvents | BatchTrackingEvents | BottleneckDetectionEvents | +2 more

• Product Traceability

  ---

  Track parts through single stations, multi-step routings, or parallel process lines. Lead times, station dwell, and routing analysis.

  ValueTraceabilityEvents | RoutingTraceabilityEvents | MultiProcessTraceabilityEvents

• Process Engineering

  ---

  Setpoint change analysis with control quality KPIs (settling time, rise time, overshoot). Startup detection across machines.

  SetpointChangeEvents | StartupDetectionEvents

• Shift Reports & KPIs

  ---

  Performance loss, scrap costs, target tracking, setup time (SMED), operator comparison, rework analysis, and auto-generated shift handover reports.

  PerformanceLossTracking | ScrapTracking | TargetTracking | SetupTimeTracking | +4 more

---

How the Modules Connect

In a real plant, signals flow between systems. ts-shape classes connect the same way — outputs from one class become inputs to another.

flowchart TB
    subgraph FLOOR["<b>Shop Floor Signals</b>"]
        S1["Machine State<br/><i>run / idle / setup</i>"]
        S2["Part Counter<br/><i>integer increments</i>"]
        S3["Part ID<br/><i>string signal</i>"]
        S4["Quality Flag<br/><i>OK / NOK</i>"]
    end

    subgraph EVENTS["<b>Event Detection</b>"]
        MS["MachineStateEvents"]
        LT["LineThroughputEvents"]
        CO["ChangeoverEvents"]
    end

    subgraph TRACKING["<b>Daily Tracking</b>"]
        PT["PartProductionTracking"]
        DT["DowntimeTracking"]
        CT["CycleTimeTracking"]
        QT["QualityTracking"]
    end

    subgraph KPI["<b>KPIs & Reports</b>"]
        OEE["OEECalculator"]
        SR["ShiftReporting"]
        SH["ShiftHandoverReport"]
    end

    S1 --> MS
    S2 --> LT
    S3 --> CO
    S4 --> QT

    MS -->|"run/idle intervals"| DT
    LT -->|"part counts"| PT
    CO -->|"product changes"| CT

    DT -->|"availability"| OEE
    PT -->|"performance"| OEE
    QT -->|"quality rate"| OEE

    DT -->|"date, shift"| SR
    PT -->|"date, shift"| SR
    QT -->|"date, shift"| SR

    SR --> SH

    style FLOOR fill:#0f2a3d,stroke:#38bdf8,color:#e0f2fe
    style EVENTS fill:#1a3a4a,stroke:#2dd4bf,color:#e0f2fe
    style TRACKING fill:#1a3a4a,stroke:#f59e0b,color:#fef3c7
    style KPI fill:#1a3a4a,stroke:#ef4444,color:#fecaca

Merge Keys

Classes connect through shared DataFrame columns: date, shift, part_number, uuid. When one class produces a column, the next class can join on it. See each guide for specific merge key documentation.

---

End-to-End Pipelines

Looking for complete workflows from Azure to results? See the Pipelines section — step-by-step use cases where you define just a UUID list and get production KPIs out the other end.

Pipeline	What you get
OEE Dashboard	Daily OEE breakdown by shift
Cycle Time Analysis	Cycle statistics, slow cycle alerts, trends
Downtime Pareto	Top downtime reasons, availability trends
Quality & SPC	Control charts, Cp/Cpk, outlier detection
Process Engineering	Setpoint adherence, stability scores