Citrus

Background

Modern buildings like malls, offices, hospitals, and schools are intricate ecosystems filled with numerous devices controlling everything from temperature to security. Managing these devices, especially specialized equipment in places like hospitals, is complex. Our solution leverages sensors, machine learning, and active anomaly detection to automate the monitoring and maintenance of these essential systems, ensuring smoother daily operations and enhanced safety.

What it does

Our solution harnesses real-time sensor data and employs an ARIMA model to predict the likelihood of device failures within the next month. This prediction enables users to take preventative action before issues arise. Additionally, the program offers a database-like interface that simplifies the task of monitoring and filtering through thousands of sensors, making it easier to manage complex building systems effectively.

Dataset & Backend

We started by creating a mock dataset with 1000 elements for each building type, resembling the dataset provided by CBRE and covering various building assets. We then used this data to run our ARIMA (AutoRegressive Integrated Moving Average) model to calculate the probability of asset failure within the next month. Based on these probabilities, a binary value was assigned to indicate the likelihood of impending failure. This processed data was then exported as a CSV for front-end integration.

Frontend

We implemented a charting system on the front end that visualizes specific CSV data points. Our ARIMA model then further processes this data to calculate failure probabilities and forecasts. Although our demo uses manually uploaded CSV files, the real-world application will auto-update this data. Features include a ranking system for assets based on their failure probability and search boxes for easy data filtering. A “Generate Report” button allows users to create an asset report, providing a streamlined overview of the asset conditions.

Key Features

• Real-time sensor data collection and monitoring
• Predictive maintenance using the ARIMA model
• User-friendly frontend with search and filter capabilities
• Report generation for quick asset overview
• Live deployment on the CentOS server

With this setup, our platform not only eases the burden of building management but also significantly enhances the reliability and efficiency of asset maintenance across different types of buildings.

Challenges Faced

Varying Building Types One of the major challenges we encountered was dealing with the inherent differences between types of buildings, such as hospitals and malls. Hospitals often have multiple rooms with specialized equipment, whereas malls tend to have open spaces. This discrepancy made writing dynamic code that could adapt to various building structures and asset types challenging.

Data Generation Limitations Our mock dataset was initially skewed towards higher probabilities of failure, which is not generally reflective of real-world conditions. This made the predictive model potentially over-alerting, risking the dilution of genuinely critical failure warnings.

Accomplishments that we’re proud of

We are proud of our capability to work together and create a fully operational product. While aesthetics may not be its strongest point at the moment, the essence of our platform lies in its robust modeling of internal building systems. The functional core of our solution demonstrates our commitment to solving real-world building and asset management challenges.

What we learned

We learned to utilize datasets to create a real time working model of our algorithms, we learned more about the implementation of the ARIMA model. This was our project that utilizes ai in a real time scenario. We also learned to upload our website onto a server, making it readily available to anyone. Overall we are very proud of our work that we committed to this project and enjoyed ourselves as we learn new concepts and push our boundaries and expand our computer science skill set and knowledge.

What’s next for Citrus

In the future, our company’s vision is to provide our customers with cutting-edge solutions that make their lives easier and more efficient. We are actively working on integrating an advanced scheduling system into our operations, allowing users to easily schedule maintenance and request assistance through a user-friendly chatbot interface. Moreover, our commitment to innovation extends to the development of a state-of-the-art CNN model that can detect real-time issues based on images or descriptions provided by our users. We also want to analyze our energy consumption to optimize energy use in buildings. We believe that these innovations will not only streamline our services but also enhance the overall experience for users.