Real time analytics at the edge for manufacturing

Solving critical production & safety challenges to drive unexpected growth

Industry4.0 powered by machine learning

Combining human experience, insight, and AI techniques, manufacturers are discovering new ways to differentiate themselves while driving down costs, protecting employees and increasing margins. Over the last 5 years, manufacturers drove massive data collection, major progress has been made on the production line, however drivers of productivity (quality, time, automation, etc.) is still scarce.

By 2035, AI-powered technologies could increase labor productivity by up to 40% in manufacturing. (Accenture and Frontier Economics)

By 2022, 99% of video/image content captured for enterprise purposes will be analyzed by machines rather than humans. (Deloitte)

of CEOs of large companies consider artificial intelligence as a key technology.

Atos Computer Vision Platform use cases

Quality control

Your challenges

  • Quality inspection is a vital part the production process – it is also manual labor intensive and prone to human error.
  • Some flaws are too subtile & too small to be detected by the human eye
  • Strict regulatory environment to ensure consumers safety & guarantee standards of quality
  • It is challenging to find qualified controllers, whilst keeping the costs of quality control reasonable.
  • Detect surface defects, missing parts, mismatched parts, correct fitting and many

Our solution

On the production line, cameras scan the product in 360° simultaneously, then the edge computing server collects, processes data in real time. BullSequana Edge offering the highest inference capabilities outside the datacenter in the plant. This solution dramatically cuts the costs of real time in-line inspection are answers to these use cases:

Key features:

  • Object detection and inspection
  • Malfunction detection
  • PPE detection
  • People compliance
  • Anomalies & outages
  • CCTV + IoT cooperation

How to start?

Worker safety

Your challenges

There are about 100 deaths per month on the job in 2019 in the USA, which has a direct impact on the company’s reputation, attractiveness, but moreover on employee’s safety feeling & productivity. It’s a high priority for manufacturers to ensure safety at all stages. The key is to ensure compliance with safety standards to prevent workplace accidents.

Our solution

A set of cameras is connected to BullSequana Edge servers, in case of a detection of a worker is not wearing his/her personal protective equipment (PPE) like ear plugs, helmet, gloves.. the server analyzes this information in real time and triggers an alert to production site managers. It can also detect:

  • Workers are in a hazardous and life-threatening situation
  • Environmental risks or hazards at the right time
  • Real-time abnormal situation (People on the ground..)
  • Dangerous driving situations with forklifts, trucks…

Outcome-driven AI Platform use cases

Predictive maintenance

Your challenges

  • Machines going unexpectedly out-of-order can be costly for smooth business operation. Knowing when a part is going to fail can prevent a lot of headache with delayed deliveries, machinery failures or unhappy clients.
  • Successful predictive maintenance enables to schedule maintenances smartly instead of having to wait for a failure.

Our solution

  • We offer predictive maintenance solutions for a wide range of applications – from real time monitoring using IoT to smarter maintenance based on historical data.
  • Our solution will help you reduce production disruptions, reduce costs and extend longevity of your equipment.

References

  • Cetin – predictive maintenance for in-the-field repairs optimalization for the largest Czech cable provider.
  • Flowpay – prediction of company’s healthiness for prioritization of investment.
  • Train producer – predictive maintenance of rolling stock bearings using real time monitoring (currently in RFP phase)

Digital twin

Your challenges

  • operational efficiency has plateaued
  • operators lack full visibility and control
  • occupants aren’t satisfied with their space
  • lack the ability to predict and preempt events.

Use cases

  • Using predictive maintenance to maintain equipment, production lines, and facilities
  • Getting a better understanding of products by monitoring them in real-time as they are used by real customers or end-users
  • Manufacturing process optimisation
  • Enhancing product traceability processes
  • Testing, validating, and refining assumptions
  • Increasing the level of integration between unconnected systems
  • Remote troubleshooting of equipment, regardless of geographical location
Use case brochure

Our solution

A digital twin is a virtual replica of a physical product, process, or system. A digital twin acts as a bridge between the digital and physical worlds, using connected sensors and IoT devices to collect real-time data about physical items. This data is then processed within a server at the edge (BullSequana Edge or BullSequana SA) and used to understand, analyze, manipulate, and optimize the item.

Demand forecasting

How do you know how much you will sell, how many machines or how much energy will you need? Inaccurate forecasting in manufacturing and warehouses causes a 5-10% loss on the margin. In the worst case, it can slow production due to out-of-stock.
With our powerful models, you can go from simple predictive methods to cutting-edge algorithms, giving you the power to account for seasonality, trends and holidays. On top of that, we offer our expertise by combining the historical data of production or sales with a wide range of data sources, such as geolocation data, promotions, competition and many more.
Join our satisfied customers like Asahi, who moved to data-driven demand forecasting with us.

Automated document reading

Your challenges

  • Many reports are still kept on paper – this makes it time consuming, prone to mistakes and hard to keep track of history.
  • Digitized documents enable advanced analytics on your documents – from fraud detection to performance prediction.

Our solution

  • Using experience across many long-running projects we offer end-to-end automated document reading solutions.
  • Our solution decreases costs associated with invoices by 60 %.
  • We can combine understanding and analysis of the text with anomaly detection as well as other sources of data – tabular data, images.

.

References

  • Kiekert – anomaly detection of incoming invoices
  • Digitoo – automated data extraction tool for administrative data used for invoice control
  • Intertek – analysis of handwritten reports of clothing goods

Our experts

Image of an expert

Pierre Jarrige

Manufacturing Business Developer, AI Solutions & 5G

pierre.jarrige@atos.net

Image of an expert

Jakub Štěch

Innovate tribe lead

jakub.stech@datasentics.com

Do you want more information?