A dataset, collected under an industrial setting, often contains a significant portion of noises. In many cases, using trivial filters is not enough to retrieve useful information i.e., accurate value without the noise. One such data is time-series sensor readings collected from moving vehicles containing fuel information. Due to the noisy dynamics and mobile environment, the sensor readings can be very noisy. Denoising such a dataset is a prerequisite for any useful application and security issues. Security is a primitive concern in present vehicular schemes. The server side for retrieving the fuel information can be easily hacked. Providing the accurate and noise free fuel information via vehicular networks become crutial. Therefore, it has led us to develop a system that can remove noise and keep the original value. The system is also helpful for vehicle industry, fuel station, and power-plant station that require fuel. In this work, we have only considered the value of fuel level, and we have come up with a unique solution to filter out the noise of high magnitudes using several algorithms such as interpolation, extrapolation, spectral clustering, agglomerative clustering, wavelet analysis, and median filtering. We have also employed peak detection and peak validation algorithms to detect fuel refill and consumption in charge-discharge cycles. We have used the R-squared metric to evaluate our model, and it is 98 percent In most cases, the difference between detected value and real value remains within the range of 1L.

翻译：在工业环境下收集的数据集通常包含大量噪声。在许多情况下，使用简单的滤波器不足以提取有用信息（即无噪声的精确值）。例如，从移动车辆收集的包含燃油信息的时间序列传感器读数。由于动态噪声和移动环境，传感器读数可能非常嘈杂。为此类数据集去噪是任何实际应用及安全问题的前提。安全问题是当前车辆方案中的首要关注点。用于检索燃油信息的服务器端可能轻易被黑客攻击。通过车辆网络提供精确且无噪声的燃油信息变得至关重要。因此，我们开发了一个能够去除噪声并保留原始数值的系统。该系统对需要燃油的车辆工业、加油站及发电厂同样有帮助。在本工作中，我们仅考虑燃油液位值，并提出了一种独特解决方案，通过插值、外推、谱聚类、凝聚聚类、小波分析及中值滤波等多种算法滤除高幅值噪声。我们还采用峰值检测与峰值验证算法来识别充放电循环中的燃油加注与消耗过程。我们使用R平方指标评估模型，其精度达到98%。在大多数情况下，检测值与真实值的差异保持在1升范围内。