consulting<\/a> to our client on optimal technologies<\/b> that would allow them to launch their project right away. We went with an open source IoT platform that provides critical user management functionality and is sufficiently optimized and scalable. In the course of the project, Intellias engineers completely customized the platform for our client\u2019s needs by reworking most of its components to fit project requirements.<\/p>\nOur team came up with a product development strategy<\/b> that outlined the product requirements and the goals for implementing the platform. We built the design and architecture of the system from scratch, developed a proof of concept and prototype, tested the product on end users, and pivoted it based on user feedback and insights from senior management. Together with our client, we held a prioritization workshop where we mapped out the next steps toward MVP development and what functionality needed to be added during MVP and post-MVP stages.<\/p>\nTeam composition<\/h3>\n
After finalizing the development pipeline, we built an end-to-end engineering team of diverse competencies by bringing together experts from across the technology spectrum. Our team consists of solution architects, UI\/UX designers, frontend and backend developers, AQA engineers, business analysts, a Scrum master, DevOps engineers, and a delivery manager.<\/p>\n
Over two and a half years of partnership with our client, our managed delivery team has grown from 6 to 30 people and has been divided into three subteams, each responsible for a different product component: a platform, an agent and data acquisition, and a portal.<\/p>\n
Platform implementation<\/h3>\n
The system we\u2019ve developed is an intelligent predictive maintenance solution for remote facility monitoring and management in the manufacturing industry. It provides predictive analytics to eliminate equipment failure and unscheduled downtime and helps manufacturers save significant costs on assets and resources through efficient asset use and maintenance.<\/p>\n
Through a network of sensors<\/b> embedded in mission-critical assets and a set of interactive dashboards<\/b>, plant operators and maintenance crews can continuously monitor the health of equipment across multiple locations. Dashboards provide real-time visibility into asset conditions, supply chains, loss prevention, financial savings, incident-free production cycles, spare parts replacements, and more.<\/p>\n\n- \n
Edge computing<\/h4>\n<\/li>\n<\/ul>\n
Our team built an agent<\/b> and a server<\/b> of the edge computing system<\/b> that captures and reads analog sensor signals on edge gateways, generates and stores data on the condition of plant assets, and structures this data to make it ready for further analysis. The intensity of data streams reaches 5 million entries per second.<\/p>\n
The agent is hosted on edge devices and provides a data acquisition flow by collecting data from sensors, processing and optimizing it, and sending it to the server through a secure communication channel. This data is then combined with other data received from plant control systems, stored, and analyzed by algorithms.<\/p>\n
\n- \n
Predictive algorithms<\/h4>\n<\/li>\n<\/ul>\n
Our client\u2019s IoT predictive maintenance system is powered by physics-based forecasting algorithms and models<\/b> developed by our client\u2019s team of data scientists. Unlike ML-based algorithms that require large volumes of data for precise predictions, physics-based algorithms enable accurate forecasts even in non-data-rich environments, such as plants that have been operating for decades. Using processed sensor data, the algorithms determine the remaining useful life (RUL) of an asset and generate systematic alerts and notifications<\/b> for hidden defects and unexpected events to prevent incipient issues and breakdowns.<\/p>\n
Our team was closely involved in developing and testing the predictive algorithms so they can be efficiently integrated into the whole system and produce accurate results. To this end, we came up with an algorithm manifesto<\/b> that details ways in which data should be passed to algorithms, how algorithms should perform, and what results should be generated on the output. The manifesto was approved by the client and simplified integration of algorithms with the edge computing system.<\/p>\n
![\"Industrial](\"https:\/\/d17ocfn2f5o4rl.cloudfront.net\/wp-content\/uploads\/2020\/07\/Pic-for-text.jpg\")
<\/p>\n