Screening Chinese Suppliers for Inspection and Packing Automation

Project Overview

An anonymized European manufacturer of precision-molded components was evaluating Chinese suppliers for an inspection and packing system.

The plant produced approximately 1.8 million components per year across 14 product variants.

The existing process used six operators per shift to:

• Load parts from trays
• Confirm orientation
• Check molding features
• Inspect printed labels
• Separate defects
• Place accepted parts into final packaging

The buyer wanted to reduce labor dependence and improve consistency, but the defect definitions were not yet precise enough for a reliable turnkey machine quotation.

The Initial Request

The original request was:

“Please quote an automatic vision inspection and packing machine.”

The first supplier search produced proposals from:

• Machine vision companies
• Custom automation machine builders
• Robot integrators
• Packaging-equipment manufacturers
• Turnkey line integrators

Prices and concepts varied widely.

Some suppliers proposed a six-axis robot and camera.

Others proposed indexing conveyors, fixed cameras, and mechanical handling.

Several suppliers quoted before testing any samples.

The Real Risk Was Inspection, Not Robot Movement

The process needed to identify:

• Missing molded features
• Incomplete seating
• Incorrect orientation
• Weak print
• Label mismatch
• Surface contamination
• Small cosmetic defects

Several defects changed appearance depending on:

• Lighting angle
• Part color
• Surface reflection
• Mold batch
• Camera exposure
• Part orientation

Moving the product from one position to another was not the hardest part.

The project would fail if the inspection concept could not distinguish acceptable variation from real defects.

Supplier Categories Were Separated

Machine Vision Specialists

They were asked to prove:

• Lighting method
• Camera position
• Lens selection
• Defect visibility
• Image examples
• False reject assumptions
• Pass-and-fail logic
• Processing time

Custom Machine Builders

They were asked to explain:

• Product handling
• Tray unloading
• Fixtures
• Indexing
• Reject segregation
• Packing flow
• Guarding
• Maintenance access
• Changeover

Robot Integrators

They were asked to justify:

• Robot type
• Reach and payload
• End-of-arm tooling
• Part presentation
• Cycle time
• Safety
• Handoff between inspection and packing

Early Supplier Responses Changed the RFQ

Several suppliers offered complete machine prices within five days.

However, they had assumed:

• Perfectly clean parts
• One product color
• Fixed label position
• No borderline defects
• Manual confirmation of uncertain results
• Low false-reject sensitivity

The stronger suppliers did not immediately provide a final price.

They asked for:

• Good samples
• Known bad samples
• Borderline samples
• Different production batches
• Defect frequency
• Target false reject rate
• Image-storage requirements
• Product changeover expectations
• Required output

This was a useful difference.

The suppliers asking more questions were not slowing the project. They were exposing the real technical risk.

The RFQ Was Rebuilt

The revised RFQ separated the project into two stages.

Stage 1: Vision Feasibility

Suppliers were asked to provide:

• Proposed lighting setup
• Camera and lens selection
• Annotated sample images
• Detection results by defect type
• Known limitations
• False reject assumptions
• Required sample quantity
• Estimated inspection time
• Written acceptance criteria

Stage 2: Complete Machine

Only suppliers with a credible inspection method would quote:

• Handling
• Fixtures
• Inspection
• Reject flow
• Packing
• Controls
• Traceability
• FAT
• Documentation
• Installation

Recommended Path

The recommended next step was a structured, paid feasibility test with two shortlisted vision-capable suppliers.

The buyer would not compare full machine prices until the inspection method had been demonstrated.

This reduced the risk of selecting a visually impressive automation concept built around an unproven inspection system.

FAT Direction

The future FAT would need to include:

• All major product variants
• Samples from several production batches
• Known examples of every defect
• Borderline samples
• Repeated inspection passes
• False reject limit
• False accept limit
• Product changeover
• Image and result storage
• Reject-bin verification
• Continuous runtime

Case Study Takeaway

For inspection automation, the supplier that proves the defect method is more valuable than the supplier that sends the fastest turnkey quotation.

Prove inspection first. Design the complete machine second.