Predicting Low-Cost Gas Sensor Readings From Transients Using Long Short-Term Memory Neural Networks

Abstract

With the everyday growth of the Internet of Things (IoT), the number of connected sensor devices increases as well, where each sensor consumes energy while being constantly online. During that time, they collect large amounts of data in short intervals leading to the collection of redundant and perhaps irrelevant data. Moreover, being commonly battery powered, sensor batteries need to be frequently replaced or recharged. The former requires smarter and less frequent data collection, while the latter being complementary to the former requires putting them to sleep while not being used in order to save energy. The focus of this article is low-cost gas sensors as they need to preheat for several minutes to reliably collect gas concentration. However, instead of waiting for a sensor to heat up, a transient, i.e., a data trend that the sensor collects while heating up is analyzed. It is shown that long short-term memory (LSTM) neural network can be used to learn and later predict the actual gas level from a part of the transient. This way, instead of being constantly online or fully preheating, the sensor needs to be turned on for only 20 s and then sleep for 120 s. With high accuracy, our approach decreases energy consumption by up to 85% compared to a system where sensors are constantly online, and more than 50% compared to a system where a sensor collects actual values instead of a part of the transient.

Using Machine Learning to Overcome Interfering Oxygen Effects in a Graphene Volatile Organic Compound Sensor

“…In nearly all classes across the 4 models, the HSS indicates that models were performing at least slightly better than random chance and, in many cases, much better than random chance. Time series methods including LSTM have previously been shown to perform better than traditional methods in various applications, including gas sensors. − Because transients are easier to capture with time series analysis, the amount of required gas exposure time can be reduced because a long sensor warm-up period is not needed. , This work shows that this is a promising improvement in analysis for graphene sensors.…”

“…30−33 Because transients are easier to capture with time series analysis, the amount of required gas exposure time can be reduced because a long sensor warm-up period is not needed. 34,35 This work shows that this is a promising improvement in analysis for graphene sensors.…”

The Spatiotemporal Dynamic Impact Mechanism of Soil Greenhouse Gases under Precipitation Based on Environmental Law

“…The kernel DM models were first used in applications such as tracing the sources of gases. Later, with the introduction of variance maps, they enabled the prediction of gas trends [36,37]. The principle of the model is essentially a density estimation.…”

Reduction of the Measurement Time of a Chemiresistive Gas Sensor Using Transient Analysis and the Cantor Pairing Function

“…Also, they are independent of materials and device structures and can be used in conjunction with other methods mentioned above. A few techniques have been proposed to reduce a sensor’s measurement time using the transient response based on the curve fitting method, low pass derivative filter method, and neural networks. − The curve fitting can reduce the measurement time by only 45%, whereas neural networks are computationally intensive algorithms and require massive training data. The response derivative method and temporal response methods are simple and effective in measuring the concentration of the gases by correlating the signal peak with the gas concentration.…”