OEM Data API and Model Repository for Autonomous Control Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Industrial automation environments face challenges in effectively and efficiently training machine learning models due to the large and diverse data generated by Original Equipment Manufacturer (OEM) assets, devices, and sensors, which hinders the integration of machine learning for autonomous control.
Innovation Solution
A system comprising a data aggregation component and a machine learning interface component that receives operational data from OEM devices, identifies device types, generates feature vectors, and transfers them to a machine learning model for training, enabling autonomous control outputs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If machine learning models are trained using large and diverse data from OEM devices, then model accuracy and autonomous control capability are improved, but data organization and processing complexity increases
Solution Approach 1:
The patent segments the large and diverse OEM data into structured collections organized by device type, asset type, and operational context. Each data collection contains standardized fields and schemas that simplify processing. This segmentation transforms unorganized raw data into manageable, structured units that can be efficiently ingested by machine learning models while maintaining high accuracy.
Solution Approach 2:
The patent introduces an intermediary data processing layer that sits between the OEM devices and the machine learning models. This intermediary layer includes data aggregation components, schema validation mechanisms, and feature engineering pipelines that automatically organize raw OEM data into standardized formats. This intermediary infrastructure handles the complexity of data organization, allowing models to access clean, structured data without directly dealing with the underlying complexity.
2Adaptability or versatility
If machine learning models are trained with diverse OEM data, then autonomous control capability is improved, but data collection and processing time increases
Solution Approach 1:
The patent implements preliminary action by pre-defining data schemas, collection structures, and processing pipelines before data ingestion. Data collections are pre-configured with appropriate fields, data types, and validation rules that match the requirements of specific machine learning models. This preliminary setup enables rapid data processing during training operations, as the infrastructure is already optimized for the expected data formats and processing workflows.
Solution Approach 2:
The patent employs parameter changes by dynamically adjusting data sampling rates, collection frequencies, and processing priorities based on model training requirements and operational conditions. The system can modify data ingestion parameters in real-time to optimize processing speed while maintaining the diversity and quality needed for effective model training. This flexible parameter adjustment reduces processing time without sacrificing the adaptability needed for autonomous control.
3Extent of automation
If machine learning models are integrated into industrial controllers, then automation extent is improved, but system complexity increases
Solution Approach 1:
The patent implements universality by creating a standardized data collection framework and processing infrastructure that can serve multiple machine learning models and control applications. The same data aggregation, validation, and feature engineering components are reused across different model training operations, reducing overall system complexity. This universal infrastructure allows the system to achieve high automation across multiple functions without proportionally increasing complexity for each individual application.
Data Source
AI summary
Various embodiments of the present technology generally relate to industrial automation environments. More specifically, embodiments include systems and methods to train machine learning systems to perform autonomous control in an industrial automation environment. In some examples, a data aggregation component receives operational data from Original Equipment Manufacturer (OEM) devices, identifies a device type for the operational data, and transfers the operational data for the device type to a machine learning interface component. The operational data characterizes the operations of the OEM devices. The interface component receives the operational data for the device type and generates feature vectors based on the operational data configured for ingestion by a machine learning model. The interface component transfers the feature vectors to a machine learning model. The interface component receives a training indication from the machine learning model that indicates an autonomous control output for the device type of the OEM devices.