Distributed optical fiber sensors (DOFSs) have attracted significant attention in recent years due to their capabilities for real-time, distributed, and long-distance sensing. They can be applied to pipelines, cables, tunnels, and other scenarios. DOFSs detect Rayleigh, Raman, and Brillouin scattered light to sense various parameters such as vibration, temperature, and strain. In practical event recognition applications, multiple parameters are often affected simultaneously. The signals collected by different DOFS systems have different features in the time and frequency domains. Vibration signals are dynamic, change rapidly, and require high sampling rates, while temperature signals are relatively static and change slowly over longer time. Therefore, appropriate signal processing methods are essential for accurate event recognition. Conventional signal processing methods typically rely on a single sensing parameter. When the signal-to-noise ratio (SNR) decreases or event features change, such methods may confuse events with interference signals of similar characteristics, leading to misidentification and reduced recognition accuracy. To address this, combining multiple DOFS systems and multi-parametric signals is an effective method. However, there are some technical challenges, such as time scale mismatches between multi-parametric signal types and the extraction of appropriate features. To overcome these issues, we propose a one-dimensional convolutional neural network (1d-CNN) that takes both extracted features and dynamic/static multi-parametric signals as input. This approach enables effective signal fusion and improves event recognition accuracy compared to traditional methods.

We propose a multi-parametric event recognition method based on 1d-CNN, integrating both extracted features and multi-parametric signals. The vibration signals are pre-processed and used as dynamic inputs, while the temperature signals are pre-processed and used as static inputs. In addition, envelope features are extracted from the vibration signals to serve as static vibration features. The proposed method extracts features from both the time and frequency domains. A specially designed multi-input 1d-CNN model combines the features and dynamic/static signals. The model consists of a convolution module, a fusion module, and a fully connected module. The convolution module extracts features from input signals through convolution layers and reduces the size of the extracted features using pooling layers. The fusion module unfolds the feature into one-dimensional vectors and combines multiple vectors into a single vector. The fully connected module classifies events based on the fused vectors and outputs the event types. To verify the effectiveness of multi-parametric signal fusion, four typical pipeline leakage events are recognized, and the accuracy, precision, recall, and F-score are compared.

The vibration and temperature signals for four typical events, including leakage events and interference types, are collected using DOFS systems. A dataset is constructed using dynamic vibration signals, static vibration signals, temperature signals, and extracted features (Table 3). The recognition results are visualized with a confusion matrix, and performance is evaluated using accuracy, precision, recall, and F-score. The effectiveness of feature fusion and multi-parametric signal fusion is compared and analyzed. Compared with only vibration signals, the fusion of features and vibration signals improves the accuracy by over 5% (Fig. 5). Compared with the single-parameter method, the fusion of dynamic vibration signals and static temperature signals improves the accuracy by over 6% (Fig. 7). The input of static vibration signal features further improve the model performance (Fig. 7). The proposed method achieves 97.25% recognition accuracy and an F-score over 0.95 for every event type (Table 4), which is superior to single-parameter methods (Fig. 9).

To achieve accurate event recognition in DOFS systems under conditions of high noise, interference, and feature changes, we propose a signal processing and event recognition method that integrates extracted features and multi-parametric signals into a 1d-CNN model, based on multiple DOFS systems. The proposed method extracts features from the sensing signals, obtaining both dynamic and static features, and then trains the 1d-CNN model with the features and multi-parametric signals to recognize four types of typical events and interferences in pipeline leakage recognition. Experimental results show that the fusion input of features and multi-parametric signals improves the model performance in multiple evaluation metrics, achieving over 97% accuracy and an F-score above 0.95 despite high levels of noise and interference. This method outperforms approaches that rely only on features or single-parametric signals. Therefore, the signal processing and event recognition method based on feature and multi-parametric signal fusion can effectively improve event recognition performance under challenging conditions, and enhance the practical application of DOFS systems.