EDGE COMPUTING ВА CLOUD COMPUTING: DIFFERENCES AND INTEGRATION

Abstract

Closed-loop quality control in metal cutting is reported as achieved whenever a system pairs a monitoring signal with an adjustable process parameter. The monitoring component and the control component are, in the great majority of published systems, validated on two separate benchmarks: a detector is scored on how accurately it identifies a defect or wear state, and a controller is scored on how closely it holds a target under a specified disturbance. This article establishes that the industrially relevant sense of closing the loop, holding tolerance as operating conditions drift away from the conditions a system was validated under, depends on the point at which the optical signal and the process-state signal enter one shared representation inside the control architecture. Recent adaptive-control and machine-vision studies were collected and classified by validated signal class and by the architectural point at which any two signals are combined; this classification shows that each field has measured its own limit for single-modality validation, using different terms for a property with the same architectural source. A three-step reading procedure applied directly to one patented laser-cutting control system's claim language, checked against its own accompanying monograph, locates this property inside an actual claim structure: an independent claim defines a control loop from three process-state channels, and a separate, dependent claim adds a visual channel to that already-defined loop. This distinction supports a specific, checkable test for whether a described closed-loop system fuses vision into its control state by definition or appends it afterward.