Building Information Model Generation From Unstructured Telemetry

The patent replaces the manual mechanical mapping process with an automated machine learning system. The ML model learns from example mappings between telemetry data and information models, then automatically performs the mapping without human intervention, thus reducing time consumption while maintaining accuracy through learned patterns.

The patent introduces an intermediate machine learning model that acts as a mediator between raw telemetry data and the target information model. This intermediary learns the mapping relationships from training data and applies them automatically, resolving the contradiction between speed and accuracy.

The patent copies expert knowledge implicitly by training the ML model on examples of correct mappings created by experts. Instead of transferring knowledge through time-consuming instruction and communication, the system captures expert patterns in the training data and reproduces them automatically during inference.

The system enables self-service by allowing the ML model to learn and apply mapping knowledge independently without requiring ongoing expert intervention. Once trained, the model autonomously performs mapping tasks that previously required expert knowledge transfer.

The patent creates a universal ML model that can handle multiple building types and telemetry data formats. The model is trained on diverse examples from different buildings, enabling it to adapt to various data structures while outputting a consistent information model format, thus achieving compatibility without requiring building-specific adaptation logic.

The system handles building-specific variations by learning parameter transformations during training. The ML model adapts to different building configurations by adjusting its internal parameters based on the input data characteristics, while maintaining a standardized output structure for compatibility.