Shipped

Manufacturing Downtime and Maintenance Cost Analysis

Where should a maintenance budget actually go? Failure-mode KPIs and a documented cost model show two failure modes carry 70% of a £2.3M cost.

View code on GitHub

£2.33Mmodelled cost of downtime in the window

70%of that cost from two failure modes

39.2 per 1,000failure rate on the low-spec line, nearly double the high-spec line

~19 pointsof budget a cost-share split moves into overstrain prevention

About the data. This project uses the AI4I 2020 Predictive Maintenance dataset from the UCI Machine Learning Repository. It is a public benchmark dataset that its authors describe as synthetic, built to reflect real predictive maintenance data. The results show the method working on a benchmark, not on a real plant.

The problem

Maintenance budgets are usually split evenly across failure categories because nobody has quantified where the money goes. I trained as a mechanical engineer, and this does the quantification on 10,000 machining cycles. It is the descriptive sibling of my predictive maintenance project: analysis first, prediction second.

What I built

A KPI engine

Failure rate, MTTR, downtime hours and cost per 1,000 cycles, by failure mode and by quality line.

A documented cost model

MTTR, parts cost and lost margin per failure mode. Every assumption is listed and tunable in one file.

Sensor signatures

The pattern each failure mode leaves in the sensor data, which turns counts into countermeasures.

How it works

KPI engineFailure rates, downtime and cost per mode and quality line
Cost modelMTTR, parts and lost margin per failure mode, assumptions in one file
SignaturesSensor pattern per failure mode
DeliverablesFormatted Excel KPI report, Power BI star schema and DAX, charts, from one command

What I found

3.39% of cycles end in failure, costing an estimated £2.33M in lost margin and parts.
Overstrain (39%) and power failures (30%) carry 70% of the cost. Tool wear is frequent but cheap (5%).
Four of five failure modes have clear sensor signatures, so they are preventable. Every heat-dissipation failure happened with a cooling gap under 8.6 K, and a tool change at about 204 minutes prevents roughly 90% of tool-wear failures.
A flat budget split is the wrong split: cost-share allocation moves about 19 points into overstrain prevention and 15 points out of tool-wear response.

Pareto chart of downtime cost by failure mode — Cost by failure mode.

Sensor signatures for each failure mode — The sensor signature of each failure mode.

Limits and honest notes

Failure counts and sensor signatures are measured. Pound figures are estimates from a documented cost model, labelled as estimates wherever they appear.
The failure-mode ranking is robust to the assumptions. The absolute figures are not, and the analysis never mixes the two.

Skills shown

PythonKPI designCost modellingExcel (openpyxl)Power BIDAXPareto analysisEngineering domain knowledge

Next projectPredictive Maintenance MLOps Platform →