Skip to content

PeriodSummaryEvents¤

Weekly and monthly production summaries and a comparison of two date ranges from a production counter, or a weekly/monthly roll-up of any daily metrics table.

ts_shape.events.production.period_summary · Production guide · API reference

Example¤

import numpy as np
import pandas as pd
from ts_shape.events.production.period_summary import PeriodSummaryEvents

t = pd.date_range("2024-01-01", periods=60 * 144, freq="10min")  # 60 days
df = pd.DataFrame({"systime": t, "uuid": "line1:parts", "value_integer": np.arange(len(t)) * 3})

summary = PeriodSummaryEvents(df)
result = summary.weekly_summary(counter_uuid="line1:parts").head()
print(result[["week_start", "total_production", "daily_avg", "production_days"]])
monthly = summary.monthly_summary(counter_uuid="line1:parts")
comparison = summary.compare_periods(counter_uuid="line1:parts",
                                     period_a=("2024-01-01", "2024-01-31"),
                                     period_b=("2024-02-01", "2024-02-29"))

Methods¤

Method Returns
weekly_summary(counter_uuid, ok_counter_uuid=None, nok_counter_uuid=None) per week: week_start, week_end, total_production, daily_avg, production_days (+ ok_parts, nok_parts, quality_pct, resets)
monthly_summary(counter_uuid, ok_counter_uuid=None, nok_counter_uuid=None) per month: year, month, same metrics
compare_periods(counter_uuid, period_a, period_b) one row per metric: metric, period_a_value, period_b_value, change, change_pct
from_daily_data(daily_df, window='W', date_column='date', aggregations=None) (static) start, end, days and the aggregated numeric columns

Notes¤

  • Weeks run Monday to Monday midnight; months are calendar months. Partial weeks or months at the edges of the data lower the averages.
  • from_daily_data takes any daily table with a date column (for example QualityEvents.daily_quality_summary() merged with DowntimeEvents.availability_trend(window="1D"), or OEEEvents.oee() with date_column="start"). window='W' gives Monday-to-Sunday weeks, 'W-<DAY>' weeks ending on that day, 'MS' calendar months.
  • Each numeric column has a declared roll-up rule. Counts, durations and energy are summed. A ratio KPI is recomputed from the period's summed basis columns (ISO 22400-2), so the daily tables the library produces roll up exactly:
Column Period value
quality_pct, first_pass_yield_pct 100 × Σgood / Σtotal (good_parts/ok_parts, total_parts or ok + nok)
nok_rate_pct 100 × Σnok / Σtotal
availability_pct 100 × Σrun_seconds / Σplanned_seconds (OEE), Σuptime / Σ(uptime + downtime) (downtime), Σoperating_hours / Σ(operating + repair) (reliability)
performance_pct 100 × Σactual_parts / Σideal_parts
oee_pct availability × performance × quality of the period
mttr_hours, mtbf_hours Σrepair_hours / Σfailures, Σoperating_hours / Σfailures
mtbf_iso22400_hours, failure_rate_per_hour Σoperating / (Σfailures + 1), Σfailures / Σoperating
enpi, sec, efficiency Σenergy_kwh / Σunits_produced, Σtotal_energy / Σtotal_output, Σunits / Σenergy (ISO 50006)

Without the basis columns a ratio is the mean weighted by its denominator when that column is present (e.g. MTTR by failures, quality by total_parts, which is exact), availability, OEE and other pct columns by the rows' duration_seconds, and otherwise the plain mean with a DataQualityWarning naming the columns. Other ratio-like columns (avg, mean, rate, ratio, cycle times) are averaged the same way, never summed. A numeric column that matches no rule is dropped with a DataQualityWarning; aggregations={"col": "sum" | "mean" | "min" | "max" | "first" | "last" | "drop"} sets the rule per column. - Counters are reset-aware by default (handle_resets=True); increments are booked to the day of the reading that reports them.