Automated quality control for industrial efficiency at the service of the end customer

Automated quality control for industrial efficiency at the service of the end customer

As we know, fashion is an industry with a complex structure, characterized by incessant rhythms and the sudden arrival of requests that often force us to review the priority objectives to be achieved in the short term. Where the need to control multiple dynamics pushes the system more and more to interface with the world of technology, in order to enhance its operational capacity, excelling in the market for competitiveness and ability to satisfy the final consumer.

The constant updating of skills is necessarily accompanied by the ability to be flexible and reactive.

It is precisely with the intention of enhancing these strengths that the fashion industries choose to turn to Automate in order to integrate automation within their processes and, thanks to it, optimize control over the quality of the goods at several levels: from the single operations relating to the manufacturing of the products, up to the management of the warehouse, where checks are carried out both on the incoming raw material and on the flows of products supplied by third parties.

Specifically, the main need exposed by the company involved was to reduce the economic investments destined for quality control, without however renouncing the efficiency and reliability of the systems used. Dealing with the production and distribution of clothing, the requests were mainly two:

• Reduce the workload for operators, so that they can be engaged in operations to set high quality control objective.
• Speed up the control phase for some types of garments articles purchased from external suppliers.

Considering these premises, the first action undertaken from the point of view of technological updating concerned the replacement of the personnel employed in the system in use with software capable of interpreting and confirming the quality of a product with the same accuracy. Thus, if before the intervention of Automate the two semi-automated lines, equipped with X-Ray machinery and a special system designed to identify and divert outgoing garments that do not comply with the established standards, needed the presence of an operators for each machine, subsequently the arrangement of an ad hoc designed system made it possible to speed up the entire operation.

That is, by automating the process of verifying the images, sent directly from the machines to the software programmed in such a way as to be able to assimilate human experience with the ability to recognize shortcomings and defects, it was possible to dispense with the employees in charge of inspecting, for example, the absence of defects or residues-such as needles in pockets-and the correct presence of labels, buttons, pockets, zippers, or any other expected detail.

In order to ensure high performance and minimize possible future critical issues, the three points most carefully addressed by the team, prior to project implementation, were:

• The training of Machine Learning algorithms by an adequate amount of collected images.
• The development of software that can handle the "doubt" factor and respond with consistent actions, coming as close as possible to the level of accuracy of the human eye.
• Accompanying operators to the knowledge and understanding of the prepared innovation, so as to facilitate the gradual integration of modern technologies into the work environment.

While a further reflection that matured from a strategic point of view at this stage, however, concerned the choice to design the software component as a modular system, relying on specific functions, such as: collection and historicization of outcomes provided by operators, recognition and counting of expected details, residual identification. An advantageous approach for achieving important objectives, such as:

• The possibility of proceeding in incremental mode, defining specific releases associated with different module functionalities.
• The optimization of the time to be allocated to data collection for algorithm training, thanks to the parallel proceeding of several distinct computation engine development activities.
• The achievement of a linear and adequate quality throughout the process, being also able to enjoy the benefits of adopting a flexible, detailed and customized approach, subject to continuous improvements in response to customer needs.

The intervention prepared by the working group consisted of four phases, set in temporal order as follows:

1. The first step focused on the introduction of useful technologies to automatically transfer the collected images and historicize the outcomes previously provided manually by staff.
2. Following the data collection, after a few weeks the development of the first version of the algorithms to be used was continued, evaluated by the same operators involved in the quality control process.
3. Subsequently, by reviewing the operations and information received, efforts were made to improve the engines, including the introduction of additional components aimed at handling rare cases.
4. Finally, the last phase, aimed at the total replacement of the operators with the technological tool, involved optimizing the calculation times of the engines, making sure that the technologies used along the entire line were integrated.

Analyzing the work performed from a broader perspective, the level of complexity faced becomes evident, due in part to the need to design a system equipped with extreme accuracy, considering rather rapid action times and the presence of anomalous situations impossible to predict. TThe same complexity is to be considered as a strong point testifying to the solidity and reliability of a project that is the result of the convergence of different skills made available by a team of experienced professionals - so it was possible to act in parallel on several fronts - and strong in the choice of using an optimal Agile methodology to lead the project with flexibility and readiness.

The benefits of this intervention were not slow to manifest themselves, referring primarily to the optimization of the time allocated to the operation. By moving personnel to the end of the line, where the intervention carried out by the devices remained limited, the result achieved not only concerned the gain of greater control, but also an average 40% reduction in the time allocated to this activity.

On the other hand, the high performance provided by the algorithm programmed at the outset was further improved in the process, as the intervention of the engine was gradually refined enabling it to achieve a level of accuracy comparable to that of specialized personnel, and which is expected to be further refined in the future due to the availability of an increasingly large and detailed amount of data.

Guided by ambition and the desire to better meet customer needs, the next goal will be to introduce advanced algorithms and mathematical models designed to control the quality of raw material supplied by third parties.