Unleash the power of data and generative AI with Microsoft Cloud for Manufacturing

Microsoft

Nov 20, 2024

The manufacturing sector generates massive amounts of data, often siloed across machines, sensors, enterprise systems, processes, and human interactions. This data often remains underutilized due to fragmentation, proprietary formats, and lack of interoperability between information technology (IT) and operational technology (OT). These data silos make it difficult for manufacturers to gain real-time insights, optimize production, ensure workplace safety, and respond swiftly to operational challenges. Moreover, complex issues like unplanned downtimes, quality control problems, safety hazards, and unpredicted supply chain disruptions add to the need for agile, data-driven decision-making across both operational and safety domains.

Microsoft is addressing these challenges with manufacturing data solutions in Microsoft Fabric, and Factory Operations Agent in Azure AI, now in public preview. These capabilities enable manufacturers to:

Ingest and unify data directly from diverse sources across the factory ecosystem.
Standardize and enrich data according to the International Society of Automation (ISA-95) model for seamless interoperability.
Create agent experiences for accessing insights through conversational interfaces.

In addition, a new Factory Safety Agent in Copilot Studio is available in preview through Microsoft Copilot Studio. This Factory Safety Agent allows industrial organizations to use low code to quickly develop a customized agent, grounded in occupational health and safety data. Whether addressing publicly available standards like those from the Occupational Safety and Health Administration (OSHA) or specific protocols of the organization, this agent provides critical safety information and insights to factory workers. It streamlines critical activities such as safety inspections or incident reporting to reduce risks and ultimately accidents.

Data Ingestion, Transformation, and Enrichment in Microsoft Fabric

Microsoft Fabric's manufacturing data solutions offer a unified approach to data ingestion, transformation, and storage, adhering to the International Society of Automation (ISA-95) information model. Data from various sources, including sensors, manufacturing execution system (MES), enterprise resource planning (ERP), and OT systems, can be ingested into Fabric through connectors like Azure IoT Operations (streaming OPC Unified Architecture (OPC UA) for factory telemetry). Industry partners can also extend these first-party connectors or create custom connectors to integrate additional data sources.

Once ingested, the data is prepared for enrichment and standardization, enabling manufacturers to structure their data according to a coherent industry data model. This standardization facilitates seamless cross-domain analysis and insights across manufacturing operations. By providing a common data framework, manufacturing data solutions enables large language models (LLMs) and AI-based tools, like the Factory Operations Agent in Azure AI, to accurately reason for advanced scenarios, such as root cause analysis and exploratory analytics. This approach enhances data-driven decision-making, operational efficiency, and accelerates time to insight, empowering manufacturers to be more agile and responsive to operational challenges.

Deploying the solutions

You can find these Microsoft Cloud for Manufacturing solutions through Microsoft Solution Center. With Solution Center, manufacturers can deploy resources using a managed on-behalf-of (MOBO) model, which shifts data ownership and infrastructure consumption to the customer's Azure tenant, enabling partners to meet stringent data ownership and isolation requirements. The MOBO deployment model provides each customer with a user-defined resource group containing the manufacturing data solutions configurations and a managed resource group that holds dependent resources, ensuring strict governance. The managed resource group hosts critical manufacturing data solutions components, such as Azure Kubernetes Service (AKS) for the agent configuration and API hosting, while locking down user access to prevent unintentional configuration changes that could impact security or future upgrades. Lastly, this deployment model ensures Microsoft engineers can collect service telemetry for diagnostic purposes, providing support and reliability without compromising customer data privacy or control.

Factory edge to manufacturing data solutions in Microsoft Fabric reference architecture

Unlocking operational efficiency with the Factory Operations Agent in Microsoft Azure AI

The Factory Operations Agent, together with manufacturing data solutions in Microsoft Fabric, addresses key operational challenges in the manufacturing landscape by democratizing insights and enabling data-driven decision-making. Machine operators benefit from real-time, contextualized data on equipment conditions, improving their response times and overall performance on the factory floor. Maintenance and quality teams can leverage the Factory Operations Agent for root cause analysis to quickly identify and address production issues that cause rate loss, ultimately boosting critical metrics like overall equipment effectiveness (OEE). Lean managers benefit from shift end reporting and waste reduction insights, optimizing processes, tracking quality metrics, and minimizing waste.

The Factory Operations Agent leverages conversational AI to streamline factory operations, enabling roles like production supervisors to ask operational questions in familiar terminology without requiring knowledge of underlying schemas or domain-specific languages. Integrating with Azure AI models, Semantic Kernel, and Azure Data Explorer, the agent offers customers control over their data while enhancing efficiency with industry-specific prompt templates and connectors.

Powered by retrieval-augmented generation (RAG) on a semantically connected data estate, the agent enhances response accuracy using predefined dictionaries, similarity searches, and schema mappings. It goes beyond simple data querying by incorporating advanced reasoning capabilities, answering "why" questions, and identifying correlations across data sets for in-depth root cause analysis. By leveraging historical data, the agent surfaces patterns and pinpoints causes for issues like machine downtimes, rate loss, or quality deviations, making connections that may not be apparent through traditional analysis methods.

Utilizing Kusto Query Language (KQL) and a cyclic process for validation, the agent dynamically constructs and refines queries, providing accurate answers to operational and business questions. The system disambiguates terms and aliases based on predefined dictionaries, and traversing temporal relationships in the data model to accurately reconstruct past states for root cause analysis.

Contextually grounded interpretation of complex questions

Factory Safety Agent in Copilot Studio

Ensuring a safe and compliant working environment is a top priority for manufacturing organizations. The Factory Safety Agent in Copilot Studio, which focuses specifically on the factory safety domain, enables industrial organizations to use low-code tools from the Copilot Studio to quickly develop a customized safety agent grounded in critical occupational health and safety (OHS) data. Whether addressing publicly available safety standards like OSHA or specific internal safety protocols, the Factory Safety Agent can offer factory workers simplified access to safety procedures and streamline essential activities such as:

Safety inspections
Incident reporting
Risk mitigation efforts

By providing immediate access to safety guidelines and hazard alerts, the Factory Safety Agent reduces the likelihood of accidents and helps ensure a safer workplace. It can also be tailored to address specific safety needs of the factory, driving a proactive safety culture that aligns both regulatory standards and organizational goals.

Factory Safety Agent creating screen

The low-code framework makes it accessible for safety officers and factory managers to quickly configure agents for their specific safety requirements. This solution can play a critical role in enabling factory operations teams to adhere to safety protocols, report hazards, and instantly escalate issues, helping to reduce risks and foster a culture of safety across the factory floor.

Customer innovations

Schaeffler, a global leader in motion technology, is partnering with Accenture and Avanade to democratize information access across its factory workforce using the Microsoft Cloud for Manufacturing data solutions in Fabric and the Factory Operations Agent in Azure AI. With an AI-ready data estate, employees can access key metrics—such as scrap rates, yields, and energy usage—via natural language queries, reducing onboarding and troubleshooting time. This approach also supports Schaeffler's goals of lowering costs and reducing carbon impact.

Bridgestone, also in partnership with Accenture and Avanade, is addressing critical challenges like disruptions, scheduling inefficiencies, and yield loss, which can lead to quality issues. Leveraging the Manufacturing data solutions in Fabric and Factory Operations Agent in Azure AI, they aim to create a natural language query system that empowers frontline workers with insights for quicker issue resolution. Bridgestone aims to efficiently gather and present key information from diverse sources, enhancing decision-making and operational agility across Bridgestone’s production ecosystem.

Partner solutions and offers

The Microsoft partner ecosystem plays a crucial role in enhancing the capabilities of Cloud for Manufacturing:

Accenture and Avanade have developed the Avanade Manufacturing Copilot, enabling organizations to harness data from MES, PLCs, sensors, and ERP systems. This copilot enhances natural language querying, empowering factory workers to gain operational insights quickly. As trusted Microsoft partners, they have been instrumental in implementing these solutions for private preview customers.

The Sight Machine Factory Copilot on Microsoft Cloud for Manufacturing integrates production data in Fabric, allowing manufacturers to correlate operational data with financial, supply chain, ERP, and MES insights. Its Factory CoPilot, leveraging the Factory Operations Agent in Azure AI provides plant managers with conversational data querying capabilities that speed up issue resolution.

The Litmus Edge Manager complements the manufacturing data solutions in Fabric by normalizing and modeling real-time machine data. This edge-to-cloud infrastructure enhances dashboards and machine learning capabilities, driving operational efficiency for manufacturers.