[Guide 05] Production Management, OEE & Bottleneck (UPH · WIP) Analytics
AI closed-loop optimization is a preview. Its real AI provider and DSL simulation backend are not connected yet, so optimization and Apply are disabled. This modal neither produces verified improvement KPIs nor writes HMI commands. Existing MCP DSL authoring, compilation and execution verification remain separate features.
Production & OEE Guide (
v2.0.0)Estimate your bottleneck step (
[BN]), steady-state hourly output (UPH), and required work-in-progress (WIP) directly from a.ds2model before wiring, then connect read-only shop-floor telemetry to track live OEE (Availability × Performance × Quality) and defect-causing actuators.
1. The 3 Pillars of OEE at a Glance
OEE (Overall Equipment Effectiveness) multiplies three clear ratios so teams immediately see whether output is lost to stops, slowdowns, or defects.
OEE Factor · Plain Question · Typical Losses · Precision Calculation Rule
1. Availability *(A)* · How much of planned time did the line run without stopping? · Fault stops, alarm waits, changeover · Overlapping stop or alarm intervals are deducted once as a combined union so Availability is never penalized twice. {{Details::Fair Deduction: Concurrent alarms collapse into a single net stop window before calculating Availability.}}
2. Performance *(P)* · While running, how close was speed to the standard cycle time? · Micro-stops, cylinder drift, idling · Speeds exceeding target cycle time (P > 1) are shown honestly instead of clipping at 100%. {{Details::Exact Storage: Raw floating-point ratios stay unrounded in exports while UI badges format cleanly to two decimals.}}
3. Quality *(Q)* · Out of all exited parts, what share passed as good units? · Startup rejects, wear-induced defects · Even without analog sensors, cylinder timing trends locate upstream actuator wear and trigger early alerts. {{Details::Proactive Hold: Flags preceding units early once timing trends cross statistical wear limits.}}
💡 One-Line Formula:
OEE (%) = Availability (A) × Performance (P) × Quality (Q)— even three solid90%scores multiply to72.9%, making this 3-way split essential for shop-floor focus.
2. Four Production KPIs Derived Before Wiring ([BN] · UPH · WIP · VA/NNVA)
From .ds2 duration values and causal arrows alone, PLAY Web Studio (Views 1/8 SCADA · 6/8 Schedule · 8/8 Causal) and DS Pilot (Plan · Dashboard · Equipment Doctor) design integration surface four practical production metrics upfront.
Production Metric · What It Reveals · How Engineers Use It
Bottleneck Step ([BN]) · The slowest critical-path step governing line pace · Focus improvements on [BN] first while isolating pure motion time from PLC scan delays. {{Details::Scan Split: Reports mechanism cycle time separately from scan-hop latency on large lines.}}
Throughput (UPH) · Units Per Hour exiting the line in steady state · Uses the mean interval between steady Production Exit completions so throughput targets match continuous production. {{Details::Formula: UPH = 3,600,000 / bottleneck cycle during upfront design analysis.}}
Critical WIP · Minimum in-flight parts needed so the bottleneck never starves · Sizes buffer conveyors and pallets from total path time divided by bottleneck cycle time (Little's Law). {{Details::Buffer Sizing: Prevents bottleneck starvation without bloating factory floor inventory.}}
VA vs. NNVA Split · Value-Added processing vs. return, clear & wait (NNVA) share · Suggests non-value-added overlaps to shave cycle time while automatically blocking unsafe parallel moves. {{Details::Safety Filter: Rejects optimizations that would break `<\ · >` mutual exclusion or safety interlocks.}}
3. Why Part Counts Never Double-Count: Single Production Exit Rule
Even in mixed-model lines with parallel branches and skip when bypasses, DS Platform prevents inflated production counts through a strict exit commit rule.
Step Kind · Counted as Output? · Why
Declared Production Exit · Yes (+1, or +K batch) · Counts finished units exactly once when the terminal exit step finishes with all parallel branches merged. {{Details::Clean Tracking: Releases active WIP capacity on exit commit so counts never inflate.}}
Auxiliary Reset (helper) · No (0) · Re-arming helpers (such as ResetReady in `Capture =\ · > ResetReady;`) clean up state for the next cycle and never count as finished products.
Variant Bypass (skip when) · No (0ms pass) · Optional stations skipped for a given variant never distort station cycle-time averages or quality denominators.
Stop / Manual Jog / Homing · No (0) · Interrupted runs, manual jog motions, and origin flow homing moves are excluded from NormalCompleted production counts. {{Details::Token & Stall Honesty: Recovery STOP/HOMING preserves WIP token payloads and Finish-waiting steps, while cyclic stalls record DS2-R504 without synthetic resets (CALL-TOKEN-BOUNDARY-1 parity).}}
4. Honest Boundary: Design Estimates vs. Live Field Telemetry
Reporting fake 0% or 100% reliability numbers before field sensors are connected misleads plant teams. DS Platform strictly separates design-time projections from measured shop-floor telemetry.
Stage · Available Immediately · Unmeasured Metrics Policy · Connection Mode
Pre-Wiring Design & SILS *(SILS / MILS)* · View theoretical UPH, bottleneck ([BN]), critical WIP, and the /play/production Smart Factory HMI immediately. {{Details::Honest Labels: Keeps estimated projections clearly separate from measured OEE // Privacy Guard: Public web tracks run production evaluation locally in your browser/PC without remote persistence}} · Reports demand-free Takt Time and 8 unmeasured maintenance KPIs (MTBF, MTTR, etc.) honestly as NOT_APPLICABLE. {{Details::No Fake Zeros: Utilization and failure metrics wait for real shop-floor telemetry instead of showing misleading zeros.}} · Generates /play/production, SCADA HTML/PDF reports, and strict bijective OPC UA (UANodeSet NodeId) from .ds2
Live Field Collection *(REAL_OBSERVE · DS Pilot)* · Tracks live OEE (A × P × Q), stop-loss breakdown, MTBF / MTTR, and per-variant output intervals. · DS Pilot (Plan · Dashboard · Equipment Doctor) uses CONCEPT_AND_SCOPED_DELIVERY, retaining physicalAcceptance = NOT_RUN until on-site sign-off completes. {{Details::Virtual Commissioning Boundary: Virtual Kit checks stay in virtual memory and mark missing baselines as inconclusive.}} · Zero-write read-only PLC adapter collects field timestamps, OEE, and signed logs safely without touching outputs {{Details::Safe Telemetry: Enforces zero write frames so running production PLCs are never modified.}}
5. Interactive Bottleneck ([BN]) & Production Exit Example (ds2 v2;)
Hover over the boxes (Work) in the diagram to spot which step governs the cycle time of this four-stage production cell.
ds2 v2;
project ProductionOeeCell {
active LineController;
product Part {
int32 lot = 1;
}
system ThermalPress {
flow Motion {
work BondPress { duration 600ms; }
work LiftWait finished { duration 200ms; }
BondPress <|> LiftWait;
}
api BondPress = Motion.BondPress;
api LiftWait = Motion.LiftWait;
}
system VisionChecker {
flow Motion {
work JudgeQuality { duration 200ms; }
work ClearResult;
JudgeQuality =|> ClearResult;
}
api JudgeQuality = Motion.JudgeQuality;
}
system LineController {
origin flow CellHome {
work PressLiftHome { ThermalPress.LiftWait reuse completed; }
}
flow ProductionFlow {
work FeedPart { duration 150ms; }
work BottleneckBonding {
ThermalPress.BondPress > ThermalPress.LiftWait;
}
work VisionCheck {
VisionChecker.JudgeQuality;
}
work ShipFinishedGood { duration 100ms; }
FeedPart => BottleneckBonding => VisionCheck =|> ShipFinishedGood;
}
api RunProduction {
command ProductionFlow.FeedPart;
observe ProductionFlow.ShipFinishedGood;
product Part;
}
}
}
- Spotting the Bottleneck (
BottleneckBonding=800ms) : CombiningBondPress(600msvalue-addedVA) andLiftWait(200msreturnNNVA) makes this the longest step — the exact place to target for higherUPH. - Clean Exit Accounting (
ShipFinishedGood) :ClearResultis only a0msre-arming helper; finished goods are counted cleanly once per cycle atShipFinishedGood.