Clip-On Machine Downtime Tracker for Small Factories

The problem small factories have with downtime

Small and medium factories, such as textile weaving units, plastic moulders, auto parts workshops and furniture makers, often run machines that are twenty or thirty years old. The owner knows the output is below capacity but not why. Production is recorded by hand at the end of a shift, downtime reasons are guessed, and the numbers in the register are whatever the supervisor writes. Without real data, the owner cannot tell whether the biggest loss is breakdowns, waiting for material, long changeovers or no operator at the machine. Modern machines expose data through a PLC or controller; old ones expose nothing.

What already exists and the gap

• Guidewheel uses clip-on current sensors to capture each machine's electrical signature and reports machine state, downtime and OEE; the company says installation takes a few minutes per machine. It is a subscription service aimed at US and European plants.
• MachineMetrics reads data directly from CNC controllers and PLCs through edge devices; strong for modern machine shops, less useful for machines with no controller.
• Evocon and Tractian offer OEE and monitoring platforms with their own sensors and dashboards, again priced for larger plants.

The gap is a low-cost kit that a small workshop can afford, which combines the current signal with a simple way for operators to record why a machine stopped, and which produces the Pareto chart a supervisor needs in the morning meeting.

What you build

• Machine node: an ESP32 with a split-core current transformer clipped around one motor supply wire. No wiring is cut and no controller is touched.
• Operator box: large labelled buttons and an LED next to each machine; when the machine stops, the LED blinks until the operator presses a reason.
• Backend: MQTT ingestion, an event service that turns state changes into downtime periods, and OEE calculation per shift.
• Displays: a floor screen with machine colours and a supervisor dashboard with reports.

The technically interesting part

Deriving machine state and cycle count from current alone is the core problem. A moulding machine has a clear repeating pattern; a loom runs continuously; a lathe spindle varies with load. You will sample current, compute RMS over short windows, and build per-machine rules or a small classifier to separate off, idle and running, then detect cycles from the pattern. You also need to match a reason press to the right downtime event when an operator presses late or presses twice. Show evaluators the raw current traces from real machines, your labelled ground truth and the accuracy of state and cycle detection.

How to evaluate it

• State detection accuracy against manually labelled video or observation for at least four hours per machine type.
• Cycle count error against the production register or a manual count over several shifts.
• Reason capture rate: percentage of stops longer than two minutes that got a reason.
• A before and after conversation with the supervisor: what did the Pareto chart show that they did not know?

Suggested plan

1. Month 1: factory visit, choose three machine types, record current traces with a data logger.
2. Month 2: state and cycle detection offline on recorded traces.
3. Month 3: machine node and operator box firmware.
4. Month 4: backend events, OEE calculation and tests.
5. Month 5: floor display and supervisor dashboard.
6. Month 6 to 7: two-week trial, accuracy study and report.

Path to a product

The first customer is a workshop owner with ten to thirty machines who suspects losses but has no data, often reachable through a local chamber of commerce or industrial estate association. Charge a one-time hardware price per machine and a monthly fee per machine for the dashboard. Scale through industrial estates, where one success story travels fast, and add machine types as you go.

Scope warning

Do not connect to PLCs or CNC controllers; the whole point is machines that have none. Do not build a full production planning or ERP system. Do not promise predictive maintenance; keep it as a stretch goal and be honest about how little data you will have.

Learn what you need on Sefism

• Data analytics roadmap : turning downtime events into Pareto charts and shift reports people act on.
• Backend cheat sheet : event handling, APIs and database design for the server side.
• System design cheat sheet : designing an ingestion path that survives flaky factory Wi-Fi.
• PostgreSQL Documentation : window functions for turning state changes into durations.
• From Data to Viz : picking charts a factory supervisor can read at a glance.

Get help with this idea

Access to a real factory is the make-or-break part of this idea, so sort that out before the proposal. If you want a second opinion on scope, hardware choices or how to set up the evaluation before you commit, bring your plan and your bill of materials and we will pressure-test it together. New here? Create a free account first.

• Discuss this idea with a mentor
• Send a query to Tauseef
• Book a 1:1 session