Search

• In previous temperature field reconstructions, the inherent delays introduced by acoustic sensors were often overlooked in the actual measurement of acoustic signals.In this paper, we propose a Spatio-Temporal Feature Extraction (STFE) method that takes into account this delay information to achieve more accurate temperature field reconstruction. • No one has done time-series prediction in the highly dynamic and nonlinear background of temperature field before, especially the location of each path is different, it is more difficult to do time-series prediction with time in different dimensions, and this paper contributes to the time-series prediction in this aspect. • The problem of time resolution of the reconstructed image of the temperature field has been solved after reconstructing the measured time-of -flight (TOF) and, at the same time, a great contribution has been made to both the accuracy and the overall error of the reconstruction of the position of the temperature field. The temperature distribution within a furnace chamber is a key parameter reflecting the combustion condition of the boiler, and acoustic tomography provides an accurate quantitative reconstruction of the temperature distribution within the chamber's overlay. However, during the measurement process of acoustic tomography, the inherent latency in ultrasonic sensor data acquisition is a challenge for the real-time detection of the temperature field. When using ultrasonic sensors to measure the time-of-flight (TOF) of acoustic waves in the furnace chamber, the temperature properties of the chamber are recorded at different time points due to the sequential nature of signal transmission and reception, and the propagation paths of the recordings are different at different moments due to the different installation locations of the sensors, so utilizing this temporal information to reconstruct the temperature field is a complex challenge. The aim of this study is to accurately reconstruct the temperature field at each moment by utilizing the TOF of different attributes. To this end, a sophisticated multi-dimensional time series analytical framework is proposed to elucidate the temperature field dynamics informed by TOF variations by extracting its temporal and spatial features, respectively, to more accurately simulate the evolution process in real engineering. Firstly, the graph structure of each acoustic wave propagation path is established. Spatial correlations are derived from Pearson correlation analysis to form the "equivalent path" matrix. Subsequently, non-negative matrix decomposition is employed to determine the weight of each position. Concurrently, the TOF matrix is utilized to extract temporal features and calculate the time weights corresponding to different temperature field conditions. These two coefficient matrices are reorganized to reconstruct the real-time temperature field for a specific frame. The reconstruction not only precisely delineates the temperature field's variations during the measurement process but also significantly improves the fidelity of reconstruction. Specifically, the position reconstruction accuracy improves by over 4%, with an overall accuracy enhancement exceeding 2%. [ABSTRACT FROM AUTHOR]