October 8, 2024

Mercedes-Benz: Quantum Computing transforms vehicle production planning

Ayesa has developed a solution based on classic and quantum AI to automatically detect anomalies prior to vehicles entering the production line.

Ayesa, a global provider of technology and engineering solutions, has partnered with Mercedes-Benz Spain to develop an zero-error vehicle manufacturing process and a production system focused on excellence. This initiative, known as Q4Real, involves the Ibermática Fundazioa, through the Ibermática Innovation Institute (I3B), to create a Quantum Platform that automatically detects order anomalies several days before vehicles enter the production line.

The Mercedes-Benz production process is unique as every individual order contains a customised configuration based on customer preferences. Each order is entered into the system as a series of codes representing the complete set of specifications for that specific vehicle. Until now, the order system prevents code combinations that could disrupt production. However, due to the complexity and sheer number of variables, some combinations could still result in a vehicle that cannot be manufactured. Ayesa’s solution addresses this by analyzing these codes in advance, identifying any unmanufacturable vehicles before they enter production. The objective is to ensure zero false negatives—vehicles that can’t be produced but go undetected—and to minimize false positives.

 

 

Harnessing the power of quantum computing

The quantum model developed by Ayesa can automatically detect anomalous configurations at various levels, including orders, manufacturing, modifications and processes. It compares ‘anomalous statuses’ and ‘similar statuses’ for a given period.

The system verifies that the identified discrepancy is accurate and aligns with the relevant documentation. If a discrepancy is found, users are notified of a possible error, thereby avoiding disruptions to the production line. The ultimate aim is to ensure zero errors in the Mercedes-Benz manufacturing process.

The automotive industry operates with a highly complex system due to the large number of components that can be combined in various ways. This complexity leads to the creation of over 5,000 unique codes, representing many different types of vehicles produced each day. Traditional IT solutions cannot effectively handle this complexity or provide the necessary efficiency to support these operations.

 

Benefits

Establishing different profiles for vehicles based on their modules, as well as detecting, notifying and explaining possible anomalies (caused by unusual, new and uncommon combinations, etc.) identified in configurations is one of the biggest and most important challenges faced by the automotive industry. However, it is one that is yet to be effectively solved.

Having an anomaly detection system means the vast majority of vehicles are automatically approved, with staff only having to manually check a small few. Of the vehicles flagged as anomalous, some will be correct, while others will require certain components to be changed. The ability to detect anomalies in highly complex and variable environments using quantum technology and optimization solutions can also yield significant benefits for various sectors, including healthcare, energy, marketing, and distribution.

 

 

Use case

The proof of concept was carried out onsite at the Mercedes-Benz production plant in Vitoria. The system was trained using a database with 50,000 orders (50,000 rows), with each one comprising 706 different components. Each day, the test dataset of 14,000 orders are compared with the database in order to “train” the system how to identify anomalous orders.

Ayesa developed a programme that creates a quantum circuit using gate computers between 10 and 30 qubits. This was executed on both IBM simulators as well as actual IBM quantum computers, with the problem mapped out using Qiskit.

Since no anomalous orders were detected in the historical data and the solution focuses on processing new requests, the problem is inherently unsupervised. The results yielded by the quantum solution are evaluated on a daily basis and the system is constantly being improved. In fact, Ayesa has implemented both a classic AI model and a quantum AI model, which are currently being assessed and compared. Although both are proving effective, the quantum AI model can identify anomalies with greater accuracy.

We support your projects

We are here for you, to advise you personally and offer you the product you need.