Skip to content

ts-shape¤

Timeseries analytics for manufacturing — from raw signals to production KPIs.
PyPI Downloads Python License

Get Started Pipelines API Reference

Why ts-shape?¤

  • DataFrame-First

    Every operation accepts and returns Pandas DataFrames. No proprietary formats, no lock-in — plug into any notebook or dashboard.

  • Modular Design

    Use only what you need. Loaders, transforms, features, and event detectors are fully decoupled.

  • Multi-Source Loading

    Load from Parquet, S3, Azure Blob, or TimescaleDB with a unified interface. Enrich with JSON metadata.

  • Manufacturing-Ready

    OEE, SPC, cycle times, downtime tracking, shift reports — built-in modules for real production use cases.

From Signals to Insights¤

from ts_shape.loader.timeseries.parquet_loader import ParquetLoader

uuids = ["machine-state", "part-counter", "temperature"]
df = ParquetLoader.load_by_uuids("data/", uuids, "2024-01-01", "2024-01-31")

from ts_shape.transform.filter.numeric_filter import NumericFilter

clean = NumericFilter.filter_value_in_range(df, "value_double", 0, 500)

from ts_shape.features.stats.numeric_stats import NumericStatistics

stats = NumericStatistics(clean, "value_double")
print(f"Mean: {stats.mean():.2f}  Std: {stats.std():.2f}")

from ts_shape.events.production.machine_state import MachineStateEvents

mse = MachineStateEvents(df, run_state_uuid="machine-state")
intervals = mse.detect_run_idle(min_duration="30s")

Architecture¤

flowchart LR
    subgraph Input
        A1[Parquet]
        A2[S3 / Azure]
        A3[TimescaleDB]
    end

    A1 --> L[Load & Enrich]
    A2 --> L
    A3 --> L

    L --> T[Transform & Filter]
    T --> F[Features & Statistics]
    F --> E[Event Detection]
    E --> O[Production KPIs & Reports]

End-to-End Pipelines¤

Each pipeline starts with an Azure connection, a UUID list, and a time range — and produces actionable KPIs.

  • OEE Dashboard

    Machine state + counters + rejects into daily Availability, Performance, Quality, and OEE by shift.

    4 UUIDs

  • Cycle Time Analysis

    Cycle triggers + part numbers into statistics, slow-cycle detection, and golden-cycle comparison.

    3 UUIDs

  • Downtime Pareto

    Machine state + reason codes into Pareto analysis, shift downtime, and availability trends.

    2 UUIDs

  • Quality & SPC

    Measurements + tolerances into outlier detection, SPC rule checks, and Cp/Cpk capability trending.

    1+ UUIDs

  • Process Engineering

    Setpoint + actual + state signals into setpoint adherence, startup detection, and stability scores.

    3 UUIDs

Core Modules¤

Loaders¤

  • Parquet — Local and remote files
  • S3 Proxy — S3-compatible storage
  • Azure Blob — Container layouts
  • TimescaleDB — SQL timeseries
  • Metadata JSON — Context enrichment

Transforms¤

  • Numeric Filter — Range, threshold
  • String Filter — Pattern matching
  • DateTime Filter — Time ranges
  • Boolean Filter — Flag filtering
  • Calculator — Derived columns
  • Harmonization — Multi-signal alignment

Features¤

  • Numeric Stats — min, max, mean, std
  • Time Stats — Coverage, gaps
  • String Stats — Value counts
  • Cycles — Extraction & processing

Events¤

  • Quality & SPC — Outliers, control charts, Cp/Cpk
  • Production — Machine states, downtime, changeovers
  • OEE — Availability, performance, quality
  • Traceability — Part tracking across stations
  • Engineering — Setpoints, startup, steady-state
  • Maintenance — Degradation, failure prediction

Data Model¤

ts-shape uses a simple long-format schema. Use only the columns you need.

Column Type Description
uuid string Signal identifier
systime datetime Timestamp
value_double float Numeric values
value_integer int Integer values
value_string string String values
value_bool bool Boolean values

  • Concept

    Architecture and design principles.

    Learn more

  • Guides

    Topic-focused guides from data acquisition to shift reports.

    See guides

  • Pipelines

    End-to-end workflows from Azure to production KPIs.

    View pipelines

  • API Reference

    Complete auto-generated API documentation.

    Browse API

  • Examples

    Runnable examples for every module category.

    Try examples

  • GitHub

    Source code, issues, and contributions.

    View source

MIT License — Built for the timeseries community