A deterministic memory management architecture for real-time brake test bench based on virtual instrumentation framework

Abstract

The long-term stability of industrial measurement and control systems depends on effective software memory management. This paper presents a memory management method for LabVIEW-based measurement and control systems to address long-term stability issues caused by the implicit memory mechanisms of the dataflow paradigm. The method combines behavior feature classification with hierarchical resource control, and is demonstrated on a data-intensive train brake test bench. First, the formation of implicit data copying, reference-based garbage collection and memory fragmentation in the LabVIEW Runtime System is analyzed, and the key factors affecting long-term stability are identified. A two-dimensional classification model based on execution frequency and memory occupancy is then built to quantitatively describe the resource requirements of functional modules. Based on this model, a three-layer management framework comprising a communication layer, an innovation layer and a data layer is designed. The communication layer uses a queue mechanism with reference semantics to eliminate unnecessary data copies; the innovation layer implements deterministic classification management that combines dynamic and static strategies; and the data layer performs periodic buffer cleaning to suppress memory fragmentation. Experimental results show that the proposed method reduces memory usage from 460 to 119 MB (a 74% decrease) and stabilizes CPU utilization at 7 % ± 2 % , achieving stable long-term operation. The proposed method provides a universal memory management approach for LabVIEW-based measurement and control software and offers theoretical guidance for the reliability design of similar systems.