RSF Overview

How a Robot Should Die: Phase 5, Lifecycle KPIs, and One Welding Arm's Four Years

When is a robot's time up? RSF's four-dimension scoring matrix, Data Erasure Certificates, hazmat compliance, and five KPIs that measure the lifecycle.

Share
How a Robot Should Die: Phase 5, Lifecycle KPIs, and One Welding Arm's Four Years
Share

Domain 1 Article Series · Part 6 of 6  |  RSF Whitepaper v1.0 · §3.6–§3.7

Robots rarely die well. The common ending goes like this: the machine limps along on rising downtime until someone finally pulls the plug, it sits in a corner of the warehouse for a year, and one day it leaves on a scrap dealer's truck — with the customer's process parameters still in its controller, its service history scattered across three filing systems, and nobody quite sure who signed off on any of it.

The RSF curriculum is unambiguous about why this ending is unacceptable: Phase 5 is not optional — it is required by contract, environmental law, and data governance, and incomplete decommissioning creates legal and IP exposure. The last phase of the lifecycle is where obligations concentrate, not where they fade.

This closing article of the Domain 1 series covers three things: how RSF decides when a robot's time is up, what a compliant ending actually involves, and how the whole lifecycle gets measured. Then it follows one welding robot through all four years — because after five articles of framework, the best summary is a biography.

Deciding: The Four-Dimension Scoring Matrix

Retirement is a decision, and RSF refuses to let it be a gut call. The framework scores the machine on four dimensions, 1 to 10 each, with defined bands:

Dimension1–3 · Continue4–7 · Overhaul8–10 · Retire
Availability trend≥ 95%, stable85–95%, slow decline< 85% and deteriorating
Maintenance-cost ratio< 10% annualized, planned-dominated10–25%, planned + unplanned mix> 25%, unplanned-dominated
Spare-parts availabilityFull OEM supplySome spares EOL or long lead timeCritical spares EOL, no substitute
Task suitabilityMeets current & planned tasksNeeds downgrade or supplementary kitCannot meet tasks; constrains the line

The dominant band across the four dimensions drives the call: continue, overhaul, or retire. Notice what the matrix runs on — availability trends, cost ratios, failure history. Every input comes from the records the earlier phases were told to keep. The PM snapshots from Part 4 of this series, the corrective work orders from Part 5: this is the decision they were feeding all along. An organization that skipped the record-keeping now faces this call with opinions instead of evidence.

Worth noting: retirement is one of three outcomes, not the default. A machine scoring in the middle band gets an overhaul and a new baseline. Even a retiring machine may be a candidate for downgraded redeployment — a robot no longer accurate enough for welding may have years left in material handling. Phase 5 opens with an assessment, not a forklift.

Ending Well: Three Sets of Obligations

When the decision is retire, the curriculum organizes the work into three obligation sets, and none of them is ceremonial.

Data erasure and IP protection. A modern robot's controller holds the customer's process parameters, and increasingly holds AI model weights and training data. All of it gets wiped, remote-access credentials revoked, VPN certificates invalidated — and the wipe gets documented in a signed Data Erasure Certificate (DEC), a hard requirement in regulated industries. A robot that leaves the building with its memory intact is a data breach on wheels.

Hazardous materials compliance. Lithium batteries, servo-drive capacitors, reducers full of oil and grease — identified, segregated, quantified per local environmental regulations, and routed to certified disposal or recycling. The chain-of-custody records stay on file for seven years. This is the part of Phase 5 where environmental law is watching, whether anyone else is or not.

The archive package. A final IPR snapshot capturing the machine's last-known state, the complete service history — every PM record, every corrective event — and every Known Fault Library entry the machine generated in its working life. The customer signs for the handover, paper and digital. The machine leaves; what it taught the organization stays.

Measuring: Five Numbers Across the Whole Lifecycle

The last piece of Domain 1 is the measurement layer — a standardized KPI set spanning every phase:

KPIPhaseDefinitionTarget
MTTRP4Total repair time ÷ number of repair events< 4 hrs (L2 SLA) / < 2 hrs (L3 SLA)
MTBFP3/P4Total operating hours ÷ number of failuresTrending upward over a 12-month rolling window
PM CRP3PM tasks completed on schedule ÷ total scheduled> 95% — non-negotiable for L3/L4 SLA tiers
WPFP4Permanent fixes ÷ total corrective events> 90%; < 10% workarounds unresolved at 30 days
OEESLAAvailability × Performance × QualityPer contract tier

One of these deserves a pause: WPF, the workaround-to-permanent-fix ratio. Part 5 argued that labeling workarounds keeps an organization honest about what is actually solved. WPF is that honesty turned into a number. A team quietly living on temporary fixes cannot hide it here — the ratio says so, thirty days at a time.

One Robot, Four Years

The RSF curriculum closes its lifecycle module with a scenario: a welding arm at an automotive tier-1 supplier, traced end to end. It compresses six articles of framework into one machine's working life.

WhenWhat Happened
P1 · Week 1Delivery and installation. IPR completed — all axes, encoder zero points, software v3.2.1. RACI established, DAC signed, Stage Gate 1 approved.
P2 · Weeks 2–3Commissioning. Trial production at 60% speed; full safety verification — E-stop, fence interlocks. Benchmark locked: 18-second cycle, ±0.05 mm repeatability.
P3 · Month 2 onThe maintenance rhythm begins. L1 weekly inspections; first L2 PM at 1,200 hours in month 4; second L2 in year 2; L3 major overhaul — reducer rebuild — in year 3.
P4 · Month 14First corrective event: J3 position deviation. Diagnosis finds reducer backlash exceeding IPR tolerance. Permanent fix applied, KFL updated, RCA feeds a PM adjustment.
P4 · Year 3Second corrective event, different in kind: vision model accuracy drop — an AI-layer fault. Model retrained on client data; the RCA adds a training-data refresh cycle to the PM checklist.
P5 · Year 4Decommissioning. Data erased, DEC issued. Reducers to a certified recycler. Archive package — IPR, full service log, KFL entries — delivered and signed for.

The curriculum attaches one observation to this timeline, and it ties the whole series together: the IPR recorded in week 1 is the diagnostic reference for the month-14 repair — without it, the backlash deviation could not have been quantified, and the fix could not have been verified. A measurement taken on a Tuesday in week one settled a repair thirteen months later. That is the lifecycle argument in a single sentence: every phase is either producing evidence or consuming it, and the machine's whole working life runs on the exchange.

Look also at the two corrective events. Month 14 is a classic mechanical fault — wear, backlash, a rebuild. Year 3 is a vision model losing accuracy — no worn part, no error code, a failure mode that did not exist in the industry's service vocabulary a decade ago. One machine, one working life, both worlds. The engineer who serviced this robot needed both diagnostic languages — which is, in miniature, the case for the certification this series has been pointing toward.

Where the Series Lands

Six articles ago, this series opened with a claim: robot service has phases, and the phases have rules. The rules turned out to be concrete. An installation is finished when its records exist, not when the machine moves. A gate is a stop, and skipping one has a named owner. A maintenance interval should measure the thing that wears the machine. A repair that teaches the organization nothing was only half done. And an ending done properly — scored, wiped, certified, archived — protects everyone the machine ever worked for.

All of it is one discipline, and the discipline is now written down: RSF Whitepaper v1.0, Chapter 3, with the RSF Professional certification course built on top of it. If this series has been your introduction to Domain 1, the whitepaper is the full text — and the course is where it becomes a career credential.

Read the complete Domain 1 specification

RSF Whitepaper v1.0, Chapter 3 — all five phases, stage gates, the decommissioning procedure, and the lifecycle KPI framework. Free download.

rsf.robottoday.com

RSF (Robotics Service Framework) is an initiative of RobotToday.com. The scoring matrix, obligation sets, KPI definitions, and lifecycle scenario follow the RSF Professional training curriculum (Module 1), which operationalizes RSF Whitepaper v1.0 §3.6–§3.7. The welding-arm scenario is a training scenario, not a client case study. Regulatory retention periods and hazmat requirements vary by jurisdiction; local law is authoritative. This article does not constitute 

RobotToday Initiative

Robotics needs a service framework.

RSF defines a common language for robot service capability, lifecycle operations, certification pathways, and service-provider networks.

Share
Written by
RSF Research - Editor

RSF Research is the research and analysis team supporting the Robot Service Framework (RSF). Its work focuses on robot service engineering, lifecycle management, maintenance methodologies, workforce development, and industry benchmarking. Through evidence-based research, technical publications, and educational resources, RSF Research aims to accelerate the professionalization of robot service worldwide.

inJoin the RobotToday community on LinkedIn

Daily robotics news, in-depth analysis, conference highlights, and discussions with professionals worldwide.