THINGPLE 炘璞智能科技中文
← All success stories

SUCCESS STORY / CONNECTED MES FOR OPTICAL CONNECTIVITY

From paper records to unit-level traceability

An optical connectivity manufacturer builds fibre, cable and connector components by hand, with no ERP and paper records at every step. Thingple is replacing the paper with a connected MES: a screen at every station, a clip-on RFID tag that follows each unit, and AI vision in place of the microscope.

Under way: the first version goes live in mid-November 2026, with full operation planned by the end of December.

Customer
Optical connectivity manufacturer
Industry
Optical components
Location
China
Solution
Manufacturing excellence · Connected MES
Status
Under way

THE SCOPE

A precise product, many steps

The first phase covers every existing production and inspection step on the pilot lines.

125 µmfibre diameter, so inspection tolerances are tight
20+production steps moving to screens and RFID
10+inspection steps feeding structured records

THE CHALLENGE

Paper records and eyes on a microscope

The line is labour-intensive and the product is small and valuable. Without an ERP or MES, nobody can see output or yield as it happens, and one unit cannot be traced.

Paper everywhere

Work instructions, job records and logbooks are on paper. Changing a line means swapping printed sheets by hand, and versions slip.

Judged by eye

Fibre end faces are checked by eye under a 400× microscope. It is slow, and one inspector's call can differ from another's.

Hard to trace

Products carry no unique code, so a customer complaint cannot be traced back to the steps one unit went through.

HOW IT WORKS

Screen, clip and camera at every station

Three simple parts carry the system. Line staff see one clear screen, and the checks happen behind it.

Station screen

Each station has an industrial display showing the digital work instruction and the job. Move the screen and the station moves with it, so lines can be reorganised without rewiring.

Hardware and MES

Clip-on RFID

A passive RFID clip fastens to each unit without damaging it, is read at each station to check the step, and is removed and reused before packing.

RFID

AI vision

Industrial cameras with a deep-learning model take over end-face and appearance checks, and keep a growing library of quality images.

Thingple AI

At every station: read the RFID clip, check the unit is in the right state, show the work instruction, do the work, then report. If a unit arrives at the wrong step, the system stops it.

Station loop: read RFID, check status, show instruction, do the work, report.

AI INSPECTION

Learns from the customer's own engineers

The model starts cautious and improves as samples build up.

Cautious start

The model leans towards false alarms rather than missed defects, then relaxes as samples grow.

Goal: 90%+

The aim is above 90% accuracy within about a month of go-live, with a few results checked by people.

Their engineers in control

Customer engineers can correct a result directly, and each correction retrains the model without a call to the supplier.

BUILT IN BLOCKS

Start with production. Add the rest when ready.

The system is modular. Later phases can be combined or run in parallel, without rebuilding the base.

Phase 1

Core production

Station screens, RFID unit traceability, AI inspection of key steps, step-by-step error-proofing, staff KPIs and a basic dashboard.

Phase 2

Management

Warehouse, scheduling and quality modules, links to ERP and purchasing, and a plant data platform.

Phase 3

Intelligence

Industrial AI models, predictive quality and maintenance, early risk warning and a digital twin.

THE PLAN

Pilot first, then copy

Everything is tested on two existing lines, so the new plant can reuse a proven design. At full scale it is meant to manage 300 to 500 work positions.

First version live

Screens, work instructions, job orders and a basic dashboard, ready to show.

Mid-Nov 2026

Full operation

Unit-level RFID traceability, AI inspection, error-proofing and equipment links.

End-Dec 2026

Copy to the new plant

The proven design is repeated across the new factory's lines.

2027

EXPLORE YOUR OPERATION

Where could better
data help your plant?

Tell us about your warehouse, your lines or your quality challenges.

Email info@thingple.com ↗