End-of-Line Vision Poka-Yoke System for Engine and Transmission Assemblies: Automated Detection of Missing and Incorrectly Installed Components

Overview

Missing or incorrectly installed components are among the most difficult assembly defects to identify manually at the end of an automotive production line.

For diesel engines, gasoline engines, transmissions and other automotive assemblies, an end-of-line vision poka-yoke system can automatically inspect critical components such as bolts, serpentine belts, nameplates, certificates, fans and filters.

By combining chain-plate conveying, industrial cameras, robotic inspection, code reading and image-based verification, the system provides an online quality checkpoint before the completed assembly leaves the production line.

The described system supports a 70-second production takt and a publicly specified 99% accuracy rate, with image retention, traceability and missing/incorrect assembly data analysis.

1. Why Missing or Incorrect Assembly Is a Critical End-of-Line Risk

Assembly defects are different from conventional dimensional defects.

A dimensional deviation can often be identified using a gauge or measuring system. Assembly errors, however, may require verification of whether a component is actually present, correctly positioned or correctly configured.

On an engine or transmission assembly line, operators may need to confirm:

• Whether bolts are installed
• Whether the serpentine belt is correctly installed
• Whether the nameplate is present
• Whether the certificate is attached
• Whether the fan is installed
• Whether the filter is installed
• Whether other specified components are present and correctly positioned

When production takt is short and the number of components is high, relying entirely on manual visual inspection creates a significant repetitive inspection workload.

Why End-of-Line Inspection Matters

If a missing or incorrectly installed component is not identified before the assembly leaves the line, the issue may only become apparent during downstream assembly, vehicle testing or field service.

An automated end-of-line inspection station provides a final verification point:

Assembly → Vision Inspection → OK / NG Result → Traceability → Release or Rework

The objective is not simply to replace operators. Instead, the system provides a consistent automated check for defined inspection items and creates a record of the assembly's condition at the end of production.

2. Equipment Configuration: Chain-Plate Conveyor + Robotic Vision Inspection

The described end-of-line inspection system uses a chain-plate conveyor and robotic inspection configuration.

Depending on the assembly structure and distribution of inspection points, the system can use either a single robot or multiple robots.

Chain-Plate Conveying

The chain-plate conveyor transports the engine or transmission assembly through the inspection station.

The conveying system provides a defined inspection position and integrates the inspection process into the production line.

Robotic Camera Inspection

Robots carry or position the inspection cameras around the assembly.

This is important because automotive powertrain assemblies have complex three-dimensional geometries. Inspection points may be distributed across:

• Top surfaces
• Side surfaces
• Front and rear areas
• End faces
• Recessed regions
• Other component-specific locations

A robotic configuration allows the inspection system to reach different viewing positions without requiring a large number of permanently fixed cameras.

3. What Can the System Inspect?

The system can be configured to detect missing or incorrectly installed components according to the customer's inspection requirements.

Typical inspection objects include:

Bolts

The system can verify whether specified bolt positions contain the expected components.

Whether the inspection needs to determine only presence or also evaluate tightening position, orientation or other characteristics must be defined during project specification.

Serpentine Belts

For engines equipped with serpentine belts, machine vision can verify whether the belt is present and whether its visible routing corresponds to the defined inspection condition.

Nameplates and Certificates

The system can verify the presence and location of identification-related components.

Code-reading functions can also be integrated when product identification or traceability is required.

Fans and Filters

The vision system can verify the presence and visible installation condition of specified components.

The actual inspection criteria depend on the component geometry and whether the relevant feature remains visible after assembly.

4. Beyond OK/NG: Code Reading and Image Traceability

One of the important features of an end-of-line vision system is that it does more than simply output an OK or NG result.

The described system can integrate:

• Code reading
• Machine vision inspection
• Overall assembly image capture
• Manual re-inspection
• Traceability
• Missing/incorrect assembly data analysis

Overall Image Retention

The complete assembly image can be retained as part of the inspection record.

This provides an important traceability function.

If a quality issue is identified later in downstream assembly or after delivery, the manufacturer can retrieve the recorded image and check the assembly condition at the time it left the production line.

This can help determine whether the condition was already visible at end-of-line inspection or whether it may have occurred during a subsequent process.

5. Technical Specifications

The following specifications are based on publicly available product information for the described customized system.

Item	Specification
Equipment type	Chain-plate conveyor + robotic vision inspection
Application	Commercial vehicles, passenger vehicles and new energy vehicle-related applications
Workpieces	Diesel engine assemblies, gasoline engine assemblies, transmission assemblies
Equipment dimensions	Approx. 11.23 × 5.04 × 3.14 m
Configuration	Single robot or multiple robots
Production takt	70 seconds, customizable according to workpiece
Publicly specified accuracy	99%
Industrial camera	20 MP
Lighting	Large high-brightness light source + high-brightness bar lights
Robot options	FANUC, ABB, KUKA, Yaskawa, SIASUN, ELITE and others
Control system	Siemens, Mitsubishi, Advantech industrial computer
Inspection functions	Code reading, component inspection, overall image capture, manual re-inspection, traceability and data analysis
Equipment design	Customized / non-standard

Actual camera, lighting, robot and software configurations should be selected according to the assembly structure and inspection requirements.

6. Key Selection Considerations

6.1 Define Every Inspection Item Before Designing the System

“Check whether the bolt is installed” and “check whether the bolt is correctly tightened” are two different inspection requirements.

Likewise:

• Is the belt present?
• Is the belt routed correctly?
• Is the nameplate present?
• Is the identification code readable?
• Is a pipe connected?
• Is the pipe connected in the correct orientation?

These requirements may require different imaging methods and inspection logic.

Therefore, the first step should be to create a complete inspection-item list and define the acceptance criteria for every item.

6.2 Robot Quantity Depends on Inspection-Point Distribution

A complete engine or transmission assembly has inspection points distributed across multiple surfaces.

A single robot may be sufficient for some configurations, while multiple robots may be required when:

• Inspection points are widely distributed
• Several viewing angles are required
• Certain areas are difficult to reach
• The production takt is tight
• Parallel image acquisition is required

Robot quantity should therefore be determined from the actual inspection-point map rather than simply from the overall workpiece size.

6.3 Image Storage Requirements Should Be Defined Early

Overall assembly image retention can provide valuable traceability, but it also creates data-storage requirements.

Before commissioning, the project should define:

• Which images need to be retained
• Image resolution
• File format
• Retention period
• Number of assemblies produced per day
• Storage capacity
• Data retrieval method

These requirements are best included in the technical specification before equipment delivery.

6.4 Public Accuracy Specifications Should Be Validated Against Actual Samples

The publicly listed system specification gives an accuracy of 99%.

This figure should be understood as a product-level published specification rather than a universal guarantee for every inspection application.

Actual performance depends on factors such as:

• Component visibility
• Camera resolution
• Lighting
• Surface condition
• Occlusion
• Product variation
• Inspection algorithm
• Sample quality
• Defined acceptance criteria

For safety-critical or high-consequence inspection items, the acceptance criteria should therefore be separately defined and validated using representative production samples.

7. What Happens When the System Cannot Make a Clear Decision?

No machine vision system should be designed around the assumption that every image will always be perfectly classifiable.

In real production environments, components may be partially blocked, lighting may vary, surfaces may reflect light, or an inspection feature may simply not be visible from the available angle.

A practical system can therefore include a manual re-inspection process for defined uncertain or non-conforming cases.

The workflow can be:

Automatic inspection → Clear OK → Release

Automatic inspection → Clear NG → Rework / Reject

Automatic inspection → Uncertain result → Manual re-inspection

This approach allows automation to handle repetitive inspection while preserving a controlled method for exceptional cases.

8. Multi-Model Production and Product Changeover

Automotive plants commonly produce multiple engine or transmission models on the same production line.

An end-of-line vision system can integrate code reading and product-program switching so that the appropriate inspection program is selected according to the identified model.

A typical process is:

Product identification → Program selection → Robot path selection → Model-specific inspection → Result recording

However, different models may have significant differences in:

• Overall dimensions
• Component locations
• Inspection points
• Camera positions
• Robot paths
• Acceptance criteria

Therefore, the complete product list should be evaluated before confirming multi-model compatibility.

9. Vision Inspection Capabilities at Big Bird Industrial / CN ISEE

Big Bird Industrial, formerly Harbin Shimada Big Bird Industrial, provides integrated industrial cleaning and machine vision solutions for automotive and precision manufacturing.

Its specialized machine vision business, CN ISEE, covers six major solution areas:

• Orient See — dimensional and geometric inspection
• Surface See — surface defect inspection
• Assembly See — assembly inspection and error-proofing
• Inner See — internal-hole and inner-wall inspection
• Paint See — paint-surface defect inspection
• Algorithm & Software Platform — machine vision software and algorithm solutions

The Assembly See series focuses on assembly verification and error-proofing applications, including end-of-line inspection for engines, transmissions and other automotive assemblies.

Depending on the project, the system can combine machine vision, code reading, robotic inspection, image retention, manual re-inspection and traceability functions.

10. FAQ

Q: What happens when the published accuracy is 99%? What about the remaining cases?

The published 99% figure should not be interpreted as meaning that every possible inspection condition will achieve exactly the same result.

Uncertain cases can result from occlusion, lighting conditions, component variation or features that are not sufficiently visible.

A practical system can therefore include manual re-inspection for defined uncertain cases, while collected samples can be used to improve inspection parameters and algorithms.

Q: Why would multiple robots be necessary?

Automotive assemblies are three-dimensional and inspection points can be distributed across several directions.

Multiple robots can divide the inspection workload by viewing area or inspection group, which can help reduce robot travel and maintain more consistent imaging conditions.

The final robot configuration should be determined according to the inspection-point distribution and required production takt.

Q: What is the practical value of retaining an overall assembly image?

The overall image provides an original visual record of the assembly's condition at the end-of-line inspection point.

If a quality issue is identified later, the manufacturer can retrieve the corresponding record and review the assembly condition at the time of inspection.

This can support traceability, quality analysis and process improvement.

Q: Can the system inspect different engine models?

Yes, the system can be designed for product identification and multi-model program switching.

However, actual compatibility depends on differences in assembly dimensions, component locations, inspection requirements and robot paths.

A complete list of engine and transmission models should therefore be provided during system design so that the required inspection programs and robot configurations can be evaluated.

Conclusion

An end-of-line vision poka-yoke system provides an automated quality checkpoint for complex automotive assemblies.

For engines and transmissions, the system can combine:

Code Reading + Robotic Vision + Component Verification + Overall Image Retention + Manual Re-inspection + Traceability

The key to a successful system is not simply selecting a high-resolution camera. The inspection requirements must first be translated into clearly defined inspection items, acceptance criteria, camera positions, robot paths, lighting conditions and data-retention requirements.

With the appropriate configuration, automated vision inspection can provide a consistent final verification stage before an engine or transmission assembly leaves the production line.

Big Bird Industrial / CN ISEE can develop customized end-of-line machine vision and assembly error-proofing systems according to workpiece structure, inspection items, product variation and production takt.