The case of an industrial electronics enterprise in northwestern Italy

In the industrial electronics industry, component quality is a critical factor that directly affects product reliability, operating costs, and lead times.
This case study tells how a company in northwestern Italy, with 70 employees, has significantly improved the quality control of printed circuit boards (PCBs) thanks to a Machine Learning solution custom-developed by Widecons®.

The background and the starting problem

The company received PCBs from an external subcontractor and used an automated visual inspection for quality control at the inlet.

The data showed significant critical issues:

• Average failure rate of 3.2% During automatic visual inspection

• Failure rate of 9.8 percent at the next functional test

• Difficulty in detecting “latent” defects that emerged only in later stages

• High indirect costs related to rework, returns and delays

Existing quality control was not effective in predicting functional failures, generating inefficiencies throughout the production chain.

The Widecons intervention: from analytics to tailored Machine Learning

Widecons approached the project with an organic and progressive approach:

Smart Small Solution – Problem Analysis

In the beginning, Widecons pointedly defined the problem, analyzing:

• the historical visual inspection data

• The correlations between visual defects and functional failures

• The limitations of the existing control algorithm

Customized Solution – Machine Learning Algorithm Development.

Based on the analysis, Widecons designed and developed a dedicated machine learning algorithm, capable of:

• Learning progressively from defect data

• Improve the ability to identify relevant defects over time

• Increase the correlation between visual inspection and functional test results

• adapt to process and supply variations

Widecons worked on images and outcomes produced by the visual inspection system installed at the customer, linking them to functional test results to build the training dataset. The algorithm was released as a software component that could be integrated into the customer’s quality control flow: trained on historical and then used to classify new incoming PCBs. The solution was designed to be Scalable and improving over time., in line with the principles of Data Intelligence applied to industry.

The results: more defects intercepted earlier, fewer failures later

The benefits of the solution have emerged in the first few months.

After 4 months

• Failure rate detected at visual inspection.:
  3.2% → 6,3%

• Functional test failure rate:
  9.8% → 3,2%

The system could intercept many more defects first, preventing them from reaching the later stages.

After 12 months

• Visual inspection failures: 7%

• Functional test failures: 1,5%

An achievement that demonstrates the ability of Machine Learning developed by Widecons to Continuously improve the performance of quality control.

Benefits extended throughout the supply chain

In addition to numerical results, the company has gained strategic advantages:

• Greater clarity on defect data, useful for dialoguing with the subcontractor

• Improvement of the PCB quality at the source, thanks to shared assessments

• Progressive decentralization of visual inspection at subcontractors

• Elimination of internal visual inspection, with reduced operating costs

• Reduction in Volumes and costs related to returns to the subcontractor

• Reduced delivery time

• Increased productivity and quality of both PCBs and the finished product

Quality was not only controlled, but governed and increased through data.

A concrete example of Widecons Digital Solutions.

This intervention effectively represents the approach described in the Widecons Digital Solutions.:

• Machine Learning and Industrial AI applied to real-world cases

• Intelligent automation of quality control processes

• Using data to improve processes, delivery and strategic decisions

• Solutions customized, non-standard, built on customer needs

It is not about applying AI in an abstract way, but about transforming data into measurable value.

Conclusion: when AI becomes a competitive advantage

This case study demonstrates that Machine Learning can become a practical tool when it is applied wisely and correctly: with reliable data, clear objectives, and integration into the actual process. In these cases:

• improves the quality of the product

• reduces costs and rework

• increases reliability and punctuality

• Strengthens collaboration with suppliers

Thanks to Widecons Digital Solutions, even an SME in industrial electronics can adopt advanced technologies pragmatically, achieving quick and lasting results.