Capgemini’s Lifecycle Optimization for Aerospace: Extending Component Lifespan Using AWS

By Claudio Pontili, AWS CoE S&C Europe – Capgemini
By Mathieu Demarquay, Solution Architect – Capgemini
By Cédric Bordonné, Sr. Partner Solutions Architect – AWS

Capgemini

One of the most rigorous operational activities in the aviation industry is the aircraft’s operational maintenance. When major aircraft maintenance or dismantling is involved, maintenance data from existing assets are used, changed, and need to be supplied with their complete history. This is a complex analytical task that’s performed manually.

To solve this challenge, Capgemini developed a solution called Lifecycle Optimization for Aerospace which aims to accelerate adoption of circular economy practices within the aviation industry.

Built on Amazon Web Services (AWS), the solution analyzes parts documentation and inspection processes, allowing optimization of its lifecycle by turning unstructured usage data into digital assets.

Lifecycle Optimization for Aerospace further optimizes maintenance approaches while promoting reuse of parts and assets, and consolidates historical operation data and reconstructs a complete traceability of all the constituent parts of an aircraft.

This solution has been developed as a result of the strategic collaboration between Capgemini and AWS, using AWS machine learning services to train and develop data models, based on standards from the Aerospace, Security, and Defense (ASD) Industries Association.

In this post, we will look at Capgemini’s solution which is built on serverless architecture and how it will help aerospace enterprises with sustainability and circular economy objectives.

Capgemini is an AWS Premier Tier Services Partner and Managed Services Provider (MSP) that’s at the forefront of innovation to address a breadth of client opportunities across cloud, digital, and platforms.

Solution Overview

Figure 1 shows the high-level architecture of the Lifecycle Optimization for Aerospace solution, as well as the AWS services used.

Figure 1 – Single-tenant serverless design.

The architecture is built on the following AWS managed services:

Amazon CloudFront: Delivers the frontend React website for users to interact with the solution.
Amazon S3: Stores static web pages for the frontend, presented to users via Amazon CloudFront. Amazon S3 is also used to store maintenance and operational documents relevant to the parts.
Elastic Load Balancing: Exposes GraphQL API for the frontend.
Amazon ECS: Hosts GraphQL APIs used by the frontend to optimize visualization and unzip process containers.
AWS Lambda: Hosts solution’s API logic that’s hosted behind Amazon API Gateway; also runs data extraction in AWS Step Functions.
Amazon API Gateway: Exposes REST API that manages document analysis workflows.
Amazon Cognito: Implements identity and access management.
Amazon Aurora: Stores relational information required to manage document analysis workflows.
Amazon OpenSearch Service: Stores document metadata extracted from document analysis workflows.
Amazon SageMaker: Used for machine learning (ML) training and inference; natural language processing (NLP) algorithms are used for page classification.
Amazon Textract: Extracts data documents using text, tables, and form APIs

The design follows a serverless and siloed multi-tenant architecture, with separation of duties between different operational teams.

Figure 2 – Serverless and siloed multi-tenant approach.

The solution achieves serverless and siloed multi-tenant architecture using AWS Control Tower. This enables the solution to be deployed on different AWS accounts for different customers using AWS CloudFormation, AWS Cloud Development Kit (AWS CDK), and AWS CodePipeline.

Crucial to this approach is a “Control Plane Tools” AWS account that provides mechanisms to control deployments, operations, while separating operational access and toolsets.

The architecture also enforces security by enforcing encryption in transit/at rest by leveraging AWS services like AWS Key Management Service (AWS KMS) and AWS Control Tower-provided security features.

Design Considerations

The asynchronous event-driven architecture of the solution enables it to scale to invocations by multiple sources or events. Such an architecture requires an orchestration mechanism and troubleshooting mechanism to ensure errors are handled with little to no supervision end to end.

Following are design considerations used in the solution:

Architecture decoupling: The solution handles a wide variety of maintenance documents containing thousands of pages, with their sizes reaching gigabytes. Workloads steps were designed using AWS Lambda and orchestrated using AWS Step Functions to adhere to Lambda’s 15-minute runtime threshold.
Scalability for simultaneous invocations: The solution analyses files using Amazon Textract to extract print text, and Textract throttles calls when transactions per second or the total number of concurrent jobs reaches the defined quota. AWS Step Functions controls the invocation rate and global number of concurrent Textract jobs .
Robustness against failures: In a distributed architecture, the amount of failures increases in the absence of development standards that develops idempotent code artefacts. This requires establishment of DevOps pipelines. The solutions’ development lifecycle enforces these standards by configuring Lambda error handling natively for asynchronous invocation, and by using of AWS Developer Tools portfolio services.
Operational troubleshooting: To improve operational root-cause analysis, a distributed architecture benefits from observability tooling. AWS enables observability using AWS X-Ray, Amazon CloudWatch RUM, and synthetic monitoring.

Next, let’s look at the document analysis workflow and how AWS Step Functions orchestrates the flow.

Document Analysis Workflow

Figure 3 shows how different components within the document analysis workflow are stitched together using AWS Step Functions.

Figure 3 – Document analysis workflow opened with AWS Step Functions Workflow Studio.

Using AWS Step Functions Workflow Studio enables developers to use it as a draft editor before transforming it into a CDK code, while also assisting in visualizing the final workflow.

Key benefits of the editor are:

Validate UI executions in real-time: This allows developers to validate the flow and identify failure points.
Provides error details: This includes number of retries, Lambda logs, and redirection to sub-execution, which can be used towards simplifying troubleshooting.
Handling state failures with retries policies: This mechanism can be used to set up retries for a Lambda function reaching its reserved concurrency limit, or to update a document status in the database when an error is detected.
Split computational workload into smaller chunks: This allows users to scale and handle a huge quantity of pages, and the solution divides the workload into batches for processing using the Distributed Map feature of AWS Step Functions. This approach manages Lambda’s 15-minute runtime threshold, while enabling optimal resource utilization and resiliency in the architecture.
Sequencing AWS Lambda calls: This improves Lambda reusability by creating small reusable building blocks, and further eases fixing or developing the workflow by manipulating simple units of code.
Parallelizing workloads: This is done with maps or parallel branches to improvise scaling capabilities.

Figure 4 – Document analysis workflow with AWS Step Functions.

Following is the way the document analysis workflow works:

The user, via the website, triggers an Amazon API Gateway endpoint to process a new package of documents. This calls a Lambda function that triggers a new asynchronous process, using AWS Step Functions.
The called “unzip” workload is responsible for instantiating an AWS Fargate container to decompress the zip into an Amazon S3 bucket and send an event to a custom Amazon EventBridge bus for each unzipped document.
For each unzipped document, the bus will trigger a new document analysis workflow execution that will:
- Start a semaphore-protected sub-workflow to recognize the characters in the document with Amazon Textract.
- Process optical character recognition (OCR) results by splitting it per pages.
- Start a sub-workflow using a Distributed Map: for each page, classifying it with an Amazon SageMaker algorithm and indexing its OCR content in an OpenSearch Service index to allow full text searches in the package.
- Once the document is classified, a new state machine is triggered to extract information from OpenSearch Service and structure them.
Per-project analysis is run on the full documentary package (the previous workflow is concentrated on a given document) to perform business-specific information reconciliations and inter-document consistency checks.

Customer Benefits

Capgemini’s Lifecycle Optimization for Aerospace digitizes hundreds of thousands of interrelated engineering documents that define an aircraft through its component parts.

The solution uses innovative AI to accelerate the time to carry out inspections and readily provide all of the relevant documents to permit the resale and re-use of parts, thereby promoting additional revenue and reduced carbon footprint for the Aerospace industry.

This serverless solution reduces costs, accelerates speed to market, and supports the aviation industry in their sustainability efforts better.

Conclusion

In this post, we described how Capgemini’s Lifecycle Optimization for Aerospace solution establishes a circular economy solution to optimize analysis of operated and end-of-life aircrafts and improve part reuse rates, resell, and recycling. The solution enables you to reimagine how the maintenance of your aircrafts can benefit from the technology, and helps in reducing maintenance time, sustainability goals, and cost of aircraft inspection.

Capgemini has observed the solution enables customers to save 30-50% of the time of an expert operator in the analysis of these documents, depending on the customer customizations to the processes.

To learn how the solution can be implemented, contact Capgemini.

.

.

Capgemini – AWS Partner Spotlight

Capgemini is an AWS Premier Tier Services Partner and MSP that’s at the forefront of innovation to address a breadth of client opportunities across cloud, digital, and platforms.

Contact Capgemini | Partner Overview | AWS Marketplace | Case Studies

AWS Partner Network (APN) Blog