Solution overview

The renewable energy sector is rapidly evolving with advancements in AI and machine learning, offering significant potential for efficiency gains and cost reductions, yet much innovation remains untapped. By integrating AI into renewable energy systems, new opportunities for efficiency gains and cost reductions emerge.

Companies are increasingly turning to predictive maintenance to increase efficiency and reduce costs. However, this strategy comes with significant challenges:

• The integration of heterogeneous data of diverse formats and sources
• Scalability becomes an issue due to the volume of IoT signals generated
• Real-time analysis of vast data sets
• Acquiring and effectively utilizing unstructured data, hindering the development of robust predictive maintenance models

Our solution explores the application of AI in real-time anomaly detection using sound input, highlighting the impact of MongoDB Atlas Vector Search. This approach offers several benefits:

• Document data model: MongoDB's BSON (binary JSON) format stores diverse data types, including unstructured data, simplifying maintenance and enabling faster responses to changes.
• Time series collections: MongoDB efficiently handles time-series data, crucial for real-time monitoring in predictive maintenance, ensuring timely interventions.
• Real-time data processing: MongoDB enables immediate diagnostics and responses, vital for proactive maintenance to prevent costly repairs.
• Data aggregation: MongoDB's powerful aggregation capabilities provide comprehensive insights into fleet-wide performance trends.
• Atlas Vector Search: MongoDB Atlas facilitates effortless search of unstructured data, enabling robust predictive maintenance solutions with features like vector indexing and retrieval.

Other applicable industries and use cases

Manufacturing: Implementing real-time anomaly detection in manufacturing plants to predict equipment failures and optimize production processes, leading to reduced downtime and increased productivity.

Transportation: Utilizing AI and Atlas Vector Search for predictive maintenance in vehicles, aircraft, and logistics equipment to anticipate maintenance needs, minimize disruptions, and improve fleet management.

Healthcare: Applying real-time anomaly detection in medical devices and equipment to identify potential issues early, ensuring patient safety, and optimizing healthcare operations.

Reference architectures

With MongoDB: The solution demo is divided in two parts:

Audio preparation

We begin by capturing the audio from the equipment in different situations (normal operation, high/low load, equipment obstructed, not operating).

Once each sound is collected, we use an embedding model to process the audio data and convert it to vector embeddings (numerical representations of data). This step is crucial — by generating embeddings for each audio track, which are high-dimensional vector representations, we are essentially capturing the unique characteristics of the sound.

We then upload these vectors to MongoDB Atlas, which is where the vector search magic begins. With just a few examples of sounds in our database, they are ready to be searched (and essentially compared) with the sound emitted by our equipment during its operation in real-time.

Audio-based diagnosis

Next we put our equipment through normal operation and start capturing the sound it is making in real time. In this demonstration we capture one-second clips of audio. Then, with the same embedding model used before, we take our audio clips and convert them to vector embeddings in real-time. This process happens in milliseconds, allowing us to have real-time monitoring of the audio. The one-second audio clips now converted to vector embeddings are then sent to MongoDB Atlas Vector Search. At this step, the Atlas Vector Search algorithm searches for what is the most similar sound from the ones we previously recorded in our audio preparation phase. The result is given back with a percentage of similarity and it enables a very accurate diagnosis of the current status of the operation of the wind turbine.

These steps are performed repeatedly every second, leveraging fast embedding of vector data and quick searches, allowing for real-time monitoring based on sound.

Data model approach

The data model for the vectorized audio collection is simple, we have our audio preparation collection called “sounds,” which contains documents including a status and the url of the related GIF to be displayed on the UI. Once the reference audio for each status is vectorized, the embedding is added to the document as an array alongside existing fields.

On the reading end, the one-second audio clips we record in real-time are vectorized and sent to MongoDB Atlas Vector Search to be compared against embeddings from the “sounds” collection above.

Building the solution

Step 1: MongoDB Atlas Connection

Create a file called .env in the main directory alongside the add_audio.py file and add your atlas connection string in the following format:

MONGO_CONNECTIO_STRING="mongodb+srv://connectionstringfromatlas"

Then copy this file into the nodeUI directory.

Step 2: Install Python Modules

Install the required python modules listed in the Github repo.

Step 3: Record Audio Files

1. Run python3 add_audio.py
2. Select the audio input by typing the relevant number and then press enter. Record each sound in sequence.

Step 4: Create a Search Index

Go to MongoDB Atlas and create an Atlas Search Index in the audio database sounds collection using the content of searchindex.json.

Code Snippet

Step 5: Query the Database

Run python3 live_query.py and place your microphone next to the fan.

Step 6: Run the Frontend

1. Switch to a new console and cd to the 'nodeUI' directory.
2. Run npm install
3. Run node nodeui.js
4. Use a browser to open the link http://localhost:3000/

Key Learnings

• Understand the role of AI and machine learning in revolutionizing predictive maintenance in the renewable energy industry.
• Explore how MongoDB Atlas Vector Search facilitates real-time anomaly detection and addresses challenges faced by businesses and development teams.
• Creation of a Vector Search index in Atlas — It is now possible to create indexes for local deployments.

Technologies and products used

MongoDB developer data platform:

• Atlas Database
• Atlas Vector Search

Partner technologies:

• Audio embedding generation with panns-inference.

Authors

• Ainhoa Múgica, MongoDB
• Arnaldo Vera, MongoDB
• Dr. Humza Akhtar, MongoDB
• Dr. Han Heloir, MongoDB
• Ralph Johnson, MongoDB