AI powered Ionic app (MVP) for anomaly detection

Zsolt MolnarZsolt Molnar | | Freelance
Cover image for AI powered Ionic app (MVP) for anomaly detection

At Starschema, I had the opportunity to participate in the development of a cutting-edge Ionic mobile app for anomaly detection. The project aimed to assist field service engineers of our client, a big Fortune 500 company, in identifying heat losses with greater accuracy by leveraging thermal imagery and advanced data science techniques.

The Project Objective

The project's primary goal was to create a mobile application capable of detecting anomalies in power generation services using thermal imagery. Our Fortune 500 client needed a powerful solution for their engineers to identify malfunctioning machines and other assets by analyzing thermal images. In this context, heat loss was a significant symptom of these anomalies.

Technologies Used

To tackle this challenging project, our team leveraged a diverse technology stack, combining different tools and programming languages:

Amazon EC2: We utilized Amazon's Elastic Compute Cloud for secure, scalable computing resources.

Java: Java was one of the primary programming languages for backend development and algorithm implementation.

Python: We used Python for various data science tasks and developed a machine learning model.

Node.js: Node.js powered the server-side environment and facilitated smooth communication between the app and backend services.

Ionic: To ensure cross-platform compatibility, we employed the Ionic framework to develop the hybrid mobile application, making it accessible on iOS and Android devices.

FLIR SDK: The FLIR Software Development Kit enabled us to seamlessly integrate thermal imaging capabilities into the application.

Application Features

The mobile app we developed offered a wide range of features designed to streamline anomaly detection and foster collaboration among field service engineers:

Thermal Image Analysis: Engineers could capture thermal images using mobile devices and upload them to the app for analysis.

Anomaly Detection: Advanced deep neural networks powered the anomaly detection algorithms, allowing the app to identify potential issues within the thermal imagery.

Asset Identification: When the algorithm detected an anomaly, the app notified the specific asset experiencing the anomaly.

Collaboration: Engineers could collaborate by flagging images with corresponding tags, allowing others to view and contribute to the anomaly identification process.

Asset Search: All assets were searchable across multiple hierarchy levels and sites, enabling efficient asset management and identifying potential issues.

Lean Approach and MVPs

We followed a lean approach to adapt to the client's evolving business requirements, delivering several Minimum Viable Products (MVPs). This iterative development process allowed us to gather client and end-user feedback, enabling us to fine-tune the application and add new features as the requirements became more elaborate.

Importance in Power Generation Services

Anomaly detection and outage prediction play a crucial role in power generation services. The app we developed addressed this importance by providing engineers with a reliable tool to find and resolve issues proactively. By leveraging the power of thermal imagery and data science, we empowered the client to enhance asset management and reduce the risk of catastrophic failures.

Working as a developer at Starschema on the Ionic mobile app for anomaly detection was an exhilarating experience. By combining technologies like Ionic, Java, Python, and Node.js, we created a powerful and user-friendly app that enabled engineers to detect anomalies and collaborate effectively and easily. The project's success lies in improving power plant services, ensuring smooth operations, and safeguarding critical assets.

Remark: The project used to be showcased on the Starschema website as a success story, but it is now available only in the web archive.

  • angular
  • NodeJS
  • typescript
  • AWS
  • ionic
  • python