Research Engineer Position - Etrel Project

31 Oct 2023

Job Information

Organisation/Company

UNIVERSITE DE TECHNOLOGIE DE COMPIEGNE

Department

Mechanical Engineering – Roberval Laboratory

Research Field

Computer science

Engineering » Computer engineering

Researcher Profile

First Stage Researcher (R1)

Country

France

Application Deadline

30 Nov 2023 - 17:00 (Europe/Paris)

Type of Contract

Temporary

Job Status

Full-time

Hours Per Week

37 : 30

Is the job funded through the EU Research Framework Programme?

Not funded by an EU programme

Is the Job related to staff position within a Research Infrastructure?

No

Offer Description

Research Engineer position, in partnership with AML Systems and as part of a “Les Hauts de France (BPI France/FEDER)” project led by DeltaCAD.

Type of contract and anticipated starting date Fixed term contract : as soon as possible and until 31/07/2024

Job location: Compiègne, université de technologie de Compiègne, AML Systems (Hirson 02 production site).

Gross monthly salary According to experience

Context

Visual inspection and detection of defects on manufactured products on a production line must be carried out in real time at very high rates (under a second). These activities are an integral part of the overall strategy to improve product quality, by making it possible to limit customer returns. They are specific to a typology of defects; in the context of this project, we will focus on the inspection protocols described below.

checking the presence of parts by enumerating components;

checking the appearance of the product: presence of flash, part alignment problems, deformations, contour defects, etc. ;

surface appearance check for defects such as scratches, impacts, cracks, opacities, etc. All products, including even the most standard ones, are susceptible to evolution (e.g. shapes, aspects, functions) to meet the specific needs of each customer.

In addition, in the context of customised products, new defects that are as yet unknown may appear. The production requirements, which are crucial contextual elements for this position are :

High production rate, with an execution time of less than one second for detection;

Incomplete knowledge of defect classes: we consider that the typology of defects is incomplete and potentially infinite. In this case, the aim is to detect defects that have not been catalogued and have never been observed or known about at the time the inspection system was developed.

Rare defects, the number of which is considerably lower than the number of correct products: for an automatic system to be able to maximise the detection rate, by default, we consider that it is more appropriate to focus on the modelling of correct products, which are much more numerous.

Acquisition data for inspection are of different types; we will consider two in particular: 2D (images, in black and white or colour, as well as video if the inspection is filmed by inspection cameras) and 3D (scanner type from tomography, structured light scanner or laser and combined CCD camera).

In summary, we observe that there are therefore currently as many visual inspections to reprogram as there are new product configurations, which is a brake on the deployment of Industry 4.0. As such, this position is in line with the strategy of the "Alliance industrie du futur". More specifically, we can refer to sheet 9 of the Future Industry Technologies guide, which deals with "innovative non-destructive testing" of the global theme "Plants and production lines connected, piloted and optimised". The use of big data technologies is recommended to increase detection rates (Alliance industrie du futur, 2018).

This position will be conducted within the framework of an BPI France/FEDER project called ETREL (« inspEction auTomatique de défauts en temps Réel et en ligne à partir de données multi-sources et via l'usage de machines apprEnantes : contribution à L'induStrie 4.0 »).

This project is being conducted by the software publisher (DeltaCAD) in partnership with AML Systems - Johnson Electric Group and the Roberval laboratory at université de technologie de Compiègne.

Working program

The candidate must propose an implementation methodology (a pipeline) for defects inspection. Deep Convolutional Neural Networks will be studied and probably exploited. These networks are non-explicit, which may, despite their high performance, be a potential hindrance to their use. In the context of this position, we will endeavour to bring applicability and to make the results obtained explicit. This will give a high level of confidence in the proposed method.

To do this, the candidate will use an experimental platform called AMS (Agile Manufacturing System) located at the UTC, as well as mechatronic components and inspection data from AML systems (Figure 1).

The aim is to provide very practical industrial POCs (Proof Of Concepts) for which the entire implementation process and sources will be made available on web and community exchange platforms. In this way, patents and scientific publications can be filed.

Advances in supervised and unsupervised AI, and the coupling of these methods with computer vision techniques, offer the prospect of an effective solution to the problems of multimodal inspection. More generally, AI research is helping to develop systems capable of handling complex behaviours that are not predefined. Deep learning neural networks have the potential to adapt to new inputs that have not yet been seen, making it possible, for example, to identify defects in images based on a restricted set of parameters. The drawback of these deep neural networks is that their process is not intelligible. However, the performance achieved is such that it is possible to integrate these neural networks into an inspection station. They would act as an aid to the operator, supporting and assisting him in his inspection.

The candidate may draw on the following literature to propose the implementation methodology:

State of the art on production image processing techniques and approaches (Carvalho et al., 2023)

Works dealing with multimodal inspection and integration techniques (Ferguson and Law, 2019; Weckenmann et al., 2009). A more detailed bibliography is given at the end of this document.

Scientific program

years

tasks

7 months

Proposal of a methodological framework on the complementarity of 2D/3D data for industrial fault monitoring (on-line/real-time).

TRL4 "demonstration of the technology in a real environment" tests to be carried out using data from AML Systems in Hirson.

Dissemination

During this position, the work resulting from the state of the art will be published in a conference or seminar with a national audience (S-MART conference, etc.). The work developed is intended to be published in international publications concerned with the implementation of software environments in engineering and production activities. In accordance with AML Systems' policy, the drafting of patents will be considered.

Requirements

Research Field

Engineering » Computer engineering

Education Level

Master Degree or equivalent

Skills/Qualifications

Skills required:

Applied artificial intelligence,

production engineering,

computer programming in Python.

Specific Requirements

Junior engineer profile accepted.

Driving licence required

Languages

FRENCH

Level

Good

Languages

ENGLISH

Level

Good

Research Field

Computer science

Additional Information

Benefits

Gross monthly salary According to experience

Selection process

Contacts

Alexandre DURUPT and Benoît Eynard, Roberval Laboratory, UTC

Hassan Koulouh, Industrial Manager, AML systems

Harvey Rowson, CEO, DeltaCAD

Application CV and covering letter to be uploaded to: https: // candidature.utc.fr/utc Following an initial screening of applications, the final decision will be based on a interview.

Work Location(s)

Number of offers available

1

Company/Institute

UNIVERSITE DE TECHNOLOGIE DE COMPIEGNE

Country

France

State/Province

HAUTS DE FRANCE

City

COMPIEGNE

Postal Code

60200

Street

Centre de recherche - rue Personne de Roberval

Geofield

Where to apply

Website

https: // candidature.utc.fr/utc

Contact

State/Province

HAUTS DE FRANCE

City

COMPIEGNE

Website

https: // www. utc.fr/en/research/utc-research-units/mechanics-energy-and- electricity-roberval/

Street

Avenue du Dr Schweitzer

Postal Code

60200

STATUS: EXPIRED