Custom Embedded Software Development for Smart Manufacturing in Pune
Project Overview
EmbedCrest, a leading embedded software development services and industrial IoT solutions provider based in India, partnered with Sarvottam Precision Manufacturing in Pune, Maharashtra to deliver a comprehensive smart factory transformation across 180 CNC machines and 24 robotic welding cells. Our embedded systems solutions team designed custom edge computing nodes on STM32H7 microcontrollers running FreeRTOS, interfacing with legacy CNC controllers from Fanuc, Siemens, and Mitsubishi through OPC-UA and Modbus RTU protocols. Each edge node ran TinyML solutions for real time anomaly detection, identifying quality deviations and equipment degradation patterns before they impacted production. The manufacturing IoT software platform aggregated telemetry through on-premises gateways running on Raspberry Pi CM4 clusters, feeding AWS IoT Greengrass for cloud-based predictive analytics solutions and production intelligence dashboards. As an embedded systems company in India with deep manufacturing domain expertise, EmbedCrest delivered the complete solution from our engineering center in Pune, including on-site commissioning, operator training, and ongoing optimization support. The platform drove OEE from 58 to 78 percent, reduced scrap rates by 62 percent, and delivered full ROI within 7 months of deployment.
Problem Statement
Sarvottam Precision Manufacturing, a mid-size automotive parts manufacturer based in Pune, Maharashtra, operated 180 CNC machines and 24 robotic welding cells across two factory floors. Their legacy SCADA system provided only basic machine status monitoring with no real time production analytics, resulting in an average OEE of just 58 percent. Manual data collection and paper-based quality logs meant defects were often caught only at final inspection, leading to a 4.2 percent scrap rate and frequent order delays. They needed a comprehensive industrial IoT solutions provider to modernize their shop floor operations without replacing existing machinery.
Engineering Approach
EmbedCrest delivered a full stack manufacturing IoT software platform built on custom embedded hardware and cloud analytics. Our embedded developers designed compact edge nodes based on STM32H7 microcontrollers running FreeRTOS, each interfacing with existing CNC controllers via OPC-UA and Modbus RTU protocols. The edge nodes sampled spindle load, vibration, coolant temperature, and cycle time data at 1 kHz resolution, performing local feature extraction and anomaly detection using a lightweight TinyML model deployed with TensorFlow Lite Micro. A mesh of industrial Wi-Fi 6 access points backhauled data to an on-premises gateway running our custom industrial IoT software on a Raspberry Pi CM4 cluster, which aggregated and normalized telemetry from all machines before pushing it to AWS IoT Greengrass for cloud analytics. Our predictive analytics solutions team built real time dashboards showing OEE, cycle time variance, and quality trends per machine, shift, and product line. The embedded software development services included custom device drivers for interfacing with Fanuc, Siemens, and Mitsubishi CNC controllers, ensuring broad compatibility across Sarvottam existing installed base. Our software testing and automation team validated every protocol integration against real production data, with zero tolerance for data loss or latency spikes that could affect production decisions.
Measurable Outcomes
58% to 78%
OEE improvement
62%
Scrap rate reduction
45%
Unplanned downtime reduction
7 months
ROI payback period
Technologies Used
“EmbedCrest understood our shop floor challenges intimately. Their team spent weeks on our factory floor in Pune before writing a single line of code, mapping every machine interface and production workflow. The result is a system our operators actually trust and use daily. Our OEE jumped 20 points in the first quarter alone.”
Vikash Joshi
Director of Operations, Sarvottam Precision Manufacturing
Note: Client details have been anonymized to protect confidentiality. Project outcomes and technical specifications are representative of actual engagements.
Frequently Asked Questions
How does the system integrate with legacy CNC controllers from different manufacturers?
Our embedded software development services team built custom device drivers for Fanuc FOCAS, Siemens SINUMERIK OPC-UA, and Mitsubishi SLMP protocols. Each driver abstracts manufacturer-specific data formats into a unified telemetry schema, allowing the analytics layer to work identically regardless of machine brand. For older machines without digital interfaces, we deployed retrofit sensor kits with current transformers and vibration sensors that provide equivalent monitoring capability.
What happens if the network connection between edge nodes and the gateway fails?
Each STM32H7 edge node has 2 MB of onboard flash configured as a circular buffer, storing up to 48 hours of compressed telemetry data locally. When connectivity is restored, the node performs a synchronized bulk upload with deduplication. Critical anomaly alerts use a secondary LoRa radio link to the gateway as a failover path, ensuring safety-critical notifications are never lost.
How was the TinyML model trained for anomaly detection across different machine types?
We used a semi-supervised approach with a variational autoencoder trained on healthy operation data from each machine type during a two-week baseline period. The model learns the normal vibration and electrical signature patterns for each specific machine and flags deviations that exceed learned boundaries. Transfer learning accelerates deployment to new machine types, requiring only three to five days of baseline data. The model runs entirely on the STM32H7 Cortex-M7 core at 480 MHz, consuming under 200 KB of RAM.
Can the platform scale to additional factory locations?
Yes, the architecture is designed for multi-site deployment. Each factory runs its own on-premises gateway cluster for low-latency local processing, while AWS IoT Greengrass provides centralized fleet management, OTA firmware updates, and cross-plant analytics. Sarvottam is currently planning expansion to their third facility near Chakan, Pune using the same platform.
