Build a predictive model able to collect, process, analyze and visualize operational data in order to reduce non-quality and non-conformity on a supply chain site.



DAHER is an aeronautic actor with five main activities: 

  • Aeronautic constructor with TBM series
  • Aerostructure and systems (Belly fairing, trapdoor…)
  • Logistics
  • Nuclear services
  • Valve constructor

Daher employs around 8500 co-workers and its revenue in 2016 was over 1 billion euros. 

The challenge will take place in the industrial site of Marignane, where DAHER does the logistics of an aeronautic client.

 Platform in the Marignane site of DAHER

Platform in the Marignane site of DAHER

One of the activities of the site is to receive pieces from the client and suppliers, stock and finally pack them in batches in order to send the batches to the border assembly line.

During this process, the occurrence of quality problems results in damaged pieces or wrong parts dispatched. Management of workflows in ramp up periods or emergency situations often leads to mistakes.

The cost of non-quality at the Marignane site represents thousands of euros per year.

Today, Daher uses classic tools to reduce non-conformity and non-quality:

  • AMDEC and Risk analysis to understand the origin of issues and solve them
  • Reports and daily routines to highlight what doesn’t function well in different processes. They are included in SEED which is an operational excellence program.

These tools enable Daher to communicate internally and to constantly improve its processes. Thanks to new technologies, Daher hopes to intensify its efforts to reduce non-quality and non-conformity, thus making the company more competitive.




 Automatized stocks in Marignane

Automatized stocks in Marignane

The main objective of the project is to lead to a quantifiable reduction of non-conformity in the Marignane operations.

One anticipated way of achieving this is to have a software that:

  • Based on historical data, determines key parameters in past non-quality
  • Detects in the present if these key parameters show a combination indicating a risk to quality
  • Would signal the risk to quality to the concerned manager who could check and set up a counter-measure.

    The aim of the project is to create a predictive model developed on Predix to identify an indicator derive, in order to schedule an intervention before the occurrence of a quality problem. To create this model, different types of data have to be collected, treated and analyzed. Above all, DAHER is seeking an algorithm able to make correlations between all data



    The data analysis algorithm must be ported on the Predix cloud, and accessible via a user interface and data visualization platform hosted on a Predix server.

    Success indicators

    The success will be measured with:

    • A relevant model in which all characteristics presented above are taken into account.
    • Rate of success in predicting a non-conformity from input operational variables.


    Data is in Excel and txt files which are extracted from the ERP where the data is stored.

    The following data is available as input variables for a predictive algorithm.

    1. Operational Data (extracted each week, txt file)

    • Reception/expedition voucher
    • Quality Note to get information about suppliers
    • Stock data

    2. Tracking tool (connection in real time)

    • Part tracking
    • Information about pieces in batches

    3. Non-conformity tool (historical data)

    • Problem typology
    • Place in the flow where the problem is detected
    • Person in charge of the problem resolution

    4. Emergency workflow (extracted each week into an excel file, tracks the emergency tasks)

    • Part tracking
    • Object time of stay by workflow steps

    5. Reception/delivery voucher (extract each week)

    • Truck tracking

    6. External parameters (each week)

    • Weather
    • Traffic
    • Events

    business scenario

    The solution developed will be assessed according to the above-mentioned success indicators. Other big data projects are planned, so if the app is a success, a collaboration between Daher and the startup is worth considering. 

    The contract between Daher and the startup could be a licence to use the software program developed. 


    The startup team will have access to operational experts as well as the innovation and data team within DAHER. Access to the Marignane site for observation and operational analysis can be granted according to the needs of the project.

    GE Predix technical experts will also be available to answer to the startup’s questions from the selection phase on.


    Capture d’écran 2017-09-15 à 10.51.25.png