Research on Highly Suspected True Alarm Model for Fire Alarm Data Based on Deep Learning Method

In this study, we studied smoke and CO alarm concentrations to develop a combination detector for both smoke and CO. The derivation of smoke and CO alarm concentrations involved examining CO concentrations by smoke concentration using a sensitivity tester, typically employed for the type approval of photoelectric smoke detectors. Fire alarm and false alarm tests were performed to analyze smoke and CO concentrations further. In the fire alarm test conducted with polyurethane foam, generating black smoke with a reduced light scattering effect in the smoke chamber, no fire alarm was triggered, although the fire alarm concentration of the existing photoelectric smoke detector of sensitivity type 2 was set at 15 %/m. The measured smoke concentration reached only up to 10.2 %/m. Conversely, in the false alarm test with sliced pork belly and mackerel, false alarms occurred as the smoke concentration exceeded 15 %/m. The determined smoke and CO alarm concentrations for the combination detectors were established at 7.5 %/m for smoke (continuously for 10 s) and 15 ppm for CO (continuously for 10 s). Applying these concentrations in subsequent tests revealed an improved performance. In the polyurethane foam fire, where the existing smoke detector failed to operate, a fire alarm was issued at 328 s with the newly derived concentrations. For smoldering wood, the combination detector issued a fire alarm 305 s faster than the existing smoke detector. Furthermore, in the false alarm test with hamburger patty, sliced pork belly, and mackerel, the adaptability of false alarms was enhanced.

Emergency evacuation, mock drills, and response collected by detectors and sensors play an important role in averting the risks and disasters. Recent scenario gives the detailing of the false alarm predictions which do affect the general time line of the organization but also create the panic situation which somewhere results in loss of life and property. The use of fire alarms which are also connected with sensors plays an important role in emergency preparedness. Fire alarm work in coordination with detectors and sensors play an effective role as a best response. False Alarm is the result of poor installation, neglect in maintenance, environment, and anthropogenic factors. According to one of the data, Great Britain noticed somewhat 312,000 false alarms in the year 2011–12. The review paper analyzes the trends and latest developments in fire alarms both in industries and residential areas. The paper discusses and gives a critical view of the current problems of fire alarms/false alarms and its future implications in dealing with emergency response and evacuation.

In the occasion of outdoor warehouse and large scale indoor warehouse, the currently widely used infrared and smoke sensors for fire alarm may can not work properly for various kinds of interferences and noises, which is prone to error fire prediction or omission prediction and other problems. Based on the design of reliable flame recognition algorithm, the paper constructed practical fire alarm system based on multi-source information fusion. The system used time-division multiplexing technology to combine temperature sensor, photoelectric smoke detector and infrared CCD image detector into multi-source fire disaster detector group. It also utilized multi-source information fusion decision-making method to complete task of accurate prediction of fire level and alarm, which improves the reliability of fire alarm and decrease the false alarm rate.

Comprehensive smokefree policies in health care settings can have a positive impact on patients’ smoking behaviour, but implementation is impeded by concern that surreptitious smoking may increase fire incidents. We investigated the incidence of routinely reported fire and false alarm incidents in a large mental health organisation in England over an 81-month period when different elements of a smokefree policy were implemented. We used negative binomial regression models to test associations between rates of fire and false alarm incidents and three hospital smokefree policy periods with mutual adjustment for occupied bed days: (1) an indoor policy which allowed disposable e-cigarettes; (2) a comprehensive policy which allowed disposable e-cigarettes; and (3) a comprehensive policy with all e-cigarette types allowed. We identified 90 fires and 200 false alarms. Fires decreased (incidence rate ratio (IRR): 0.35, 95% CI: 0.17–0.72, p = 0.004) and false alarms increased (IRR: 1.67, 95% CI: 1.02–2.76, p = 0.043), each by approximately two-thirds, when all e-cigarette types were allowed, after adjusting for bed occupancy and the comprehensive smokefree policy. Implementation of smokefree policies in mental health care settings that support use of all types of e-cigarettes may reduce fire risks, though measures to minimise effects of e-cigarette vapour on smoke detector systems may be needed to reduce false alarm incidents.

In the era of smart home technology, early warning systems and emergency services are inevitable. To make smart homes safer, early fire alarm systems can play a significant role. Smart homes usually utilize communication, sensors, actuators, and other technologies to provide a safe and smart environment. This research work introduced a model for the fire alarm system and designed a fire alarm detection (FAD) simulator to produce a synthetic dataset. The designed simulator utilizes a variety of sensors (temperature, gas, and humidity) to simulate fire alarm scenarios based on real-world data. The produced data is investigated and analyzed to classify the possible fire behaviors based on key assumptions taken from real-world scenarios. Different classification models are used to determine an optimal classifier for fire detection. The proposed technique can identify the false alarms based on parameters like temperature, smoke, and gas values of different sensors embedded in a fire alarm detection simulator.

The high rate of false arrhythmia alarms in Intensive Care Units (ICUs) can lead to disruption of care, negatively impacting patients’ health through noise disturbances, and slow staff response time due to alarm fatigue. Prior false-alarm reduction approaches are often rule-based and require hand-crafted features from physiological waveforms as inputs to machine learning classifiers. Despite considerable prior efforts to address the problem, false alarms are a continuing problem in the ICUs. In this work, we present a deep learning framework to automatically learn feature representations of physiological waveforms using convolutional neural networks (CNNs) to discriminate between true vs. false arrhythmia alarms. We use Contrastive Learning to simultaneously minimize a binary cross entropy classification loss and a proposed similarity loss from pair-wise comparisons of waveform segments over time as a discriminative constraint. Furthermore, we augment our deep models with learned embeddings from a rule-based method to leverage prior domain knowledge for each alarm type. We evaluate our method using the dataset from the 2015 PhysioNet Computing in Cardiology Challenge. Ablation analysis demonstrates that Contrastive Learning significantly improves the performance of a combined deep learning and rule-based-embedding approach. Our results indicate that the final proposed deep learning framework achieves superior performance in comparison to the winning entries of the Challenge.

This study is a small residential factory with a lightweight steel structure for fire characteristics analysis and alarm equipment analysis. This kind of factory fire grows very quickly, and there is no perfect fire protection zone, the temperature rises rapidly when the fire occurs, and it is difficult for the fire brigade to rescue. Use FDS for fire simulation and discuss the activation time of fire alarms with different functions, when the people in the building are awake, the thermocouple distribution position is combined with the heating curve to analyze the critical value of the human body’s dangerous temperature, and the building is partitioned and the allowable evacuation time includes: 13 minutes in the factory area, 11 minutes in the corridor, and office area there are 12 minutes and the room has 15 minutes. In addition, the difference in the action time of the two detectors, the smoke detection type and the fixed temperature type, is discussed. In the simulation results, the smoke alarm device can detect the fire 137 seconds earlier.

Contemporary fire smoke detectors predominantly utilise the intensity of scattered light to identify fire smoke and initiate fire alarms. However, particles from non-fire aerosols may trigger false fire alarms, and response times for black and white smoke may be inconsistent due to variations in the refractive index. To address the limitations of existing fire smoke detection techniques in terms of reducing the false alarm rate and improving the detection sensitivity, this study aims to optimise the performance of fire smoke detectors by introducing a three-wavelength Sauter mean diameter (SMD) detection method based on Mie scattering theory and a three-stage model. Specifically, a third-wavelength correction channel is added on the basis of the dual-wavelength design, and the effects of the aerosol particle size distribution, refractive index and scattering angle on the detection performance are evaluated via simulation analysis. The results show that the ability of particle detection can be improved by increasing the intermediate wavelength, and the attenuation of large-angle scattering signals can be compensated by the weighting function. Compared with that of the traditional dual-wavelength method, the average relative error of detection can be reduced from 41% to 8.9% when the three-wavelength Sauter mean diameter methods of 450 nm, 550 nm and 950 nm are used, which effectively balances the detection sensitivity of black smoke and white smoke.

Chemical oxygen demand (COD) is an important indicator of organic pollution in water. It plays a crucial role in assessing water quality and protecting the environment. The rapid and accurate detection of COD is essential for continuous water quality monitoring. Traditional methods are often time-consuming and require chemical reagents. To address these challenges, a novel approach based on a one-dimensional convolutional neural network (1D-CNN) combined with UV-vis spectroscopy is proposed. This method is efficient, fast, and reagent-free, offering significant advantages over conventional techniques. The proposed 1D-CNN method incorporates one-dimensional multi-scale feature fusion to enhance the accuracy of COD detection. The method improves spectral feature extraction by fusing features extracted from three parallel sub-convolutional and pooling layers within the same channel. Experimental results show that the fusion network performs well in COD detection. The proposed method demonstrates superior performance compared to traditional and deep learning methods such as partial least squares regression (PLSR), support vector machines (SVM), artificial neural network (ANN), and 1D-CNNs. It significantly improves the accuracy of UV-vis spectroscopy for COD detection, achieving higher precision in real-time water quality monitoring.

The Cargo Fire Verification System was developed to address the problem of frequent false smoke alarms that are of particular concern in long range flights of passenger aircraft. The system uses low-cost CCD cameras operating in the near infrared range to directly detect fire and hotspots. In addition, LED illumination units are appropriately switched on and off, and the obtained images are analysed to detect smoke. Fusion of image processing results with temperature and humidity readings allows reliable detection of true fires and elimination of false alarms due to condensed moisture (fog) and dust. It was necessary to create a suite of fire sensitivity and false alarm immunity tests applicable to these vision based fire detection systems. This paper will concentrate on the design of the system, testing aspects, and test cell modifications.

This paper proposes a design and experimentally implements of a multimodal machine learning (ML)-based indoor fire detection and alarm system which combines image processing and Internet-of-Things (IoT)-based environmental sensing for mitigating false alarm. The proposed system consists of two fire ML-based detecting models: the image processing model used for detecting a fire based on fire images captured by a camera, a sensing data processing model used for predicting a fire based on room environmental data. Additionally, a communication module (i.e. 4G/Wi-Fi) is equipped to use for interacting with the home-owner and the fire station to get a confirmation before activating any further actions. In the proposed model, detected results of both abovementioned detecting models are combined to enhance the reliability of the final decision (i.e. whether it is a true fire or false fire). A true fire is recognized when two detecting modules both detected that it is a fire. Otherwise, when only one detecting module recognized that it is a false fire, it is recognized as uncertain. In this case, it needs to be confirmed by the home-owner based on fire images captured by the camera and sent over the communication module. If it is recognized as a true fire, a true alarm is raised and an alarm message containing house’s location information is sent to the fire station for activating the emergency case of fire. Otherwise, the fire is recognized as a false alarm and then no action is taken. The numerical and experimental results show that the proposed system achieves a high accuracy of fire detection and avoids false alarms.

In many real applications, the ground truths of class labels from voltage dip sequences used for training a voltage dip classification system are unknown, and require manual labelling by human experts. This paper proposes a novel deep active learning method for automatic labelling of voltage dip sequences used for the training process. We propose a novel deep active learning method, guided by a generative adversarial network (GAN), where the generator is formed by modelling data with a Gaussian mixture model and provides the estimated probability distribution function (pdf) where the query criterion of the deep active learning method is built upon. Furthermore, the discriminator is formed by a support vector machine (SVM). The proposed method has been tested on a voltage dip dataset (containing 916 dips) measured in a European country. The experiments have resulted in good performance (classification rate 83% and false alarm 3.2%), which have demonstrated the effectiveness of the proposed method.

The purpose of this study was to determine student responses after using learning media that have been developed, namely Mabarung picture books (various types of Building Space). The test was conducted in class 5 SDN 02 Bugangan Semarang and students in grade 5 SDN 01 Bugangan Semarang. The research method used is development research. the final stage in this study was the researcher used a student questionnaire in collecting data. The questionnaire was in the form of 20 statements regarding learning that had been done using MABARUNG media. The results of the questionnaire analysis of student responses at SDN 02 Bugangan showed a percentage of 96.25% and at SDN 01 Bugangan showed a percentage of 97.50%, meaning that MABARUNG media could foster interest and enthusiasm of students in learning mathematics and could facilitate students in understanding building material space is packed into a story. Keywords: Development, Mabarung Picture Book (Various Types of Space Building).

Deep learning methods have become popular among researchers in the field of fault detection. However, their performance depends on the availability of big datasets. To overcome this problem researchers started applying transfer learning to achieve good performance from small available datasets, by leveraging multiple prediction models over similar machines and working conditions. However, the influence of negative transfer limits their application. Negative transfer among prediction models increases when the environment and working conditions are changing continuously. To overcome the effect of negative transfer, we propose a novel deep transfer learning method, coined deep boosted transfer learning, for wind turbine gearbox fault detection that prevents negative transfer and only focuses on relevant information from the source machine. The proposed method is an instance-based deep transfer learning method that updates the weights of the source and the target machine training samples separately. The weights of different source training samples are gradually decreased to reduce the impact on the final model. The proposed method is verified by the Case Western Reserve University bearing and real field wind farm datasets. The results show that the proposed method ignores negative transfer and achieves higher accuracy compared to standard deep learning and deep transfer learning methods.

Short-term traffic flow prediction can provide a basis for traffic management and support for travelers to make decisions. Accurate short-term traffic flow prediction also provides necessary conditions for the sustainable development of the traffic environment. Although the application of deep learning methods for traffic flow prediction has achieved good accuracy, the problem of combining multiple deep learning methods to improve the prediction accuracy of a single method still has a margin for in-depth research. In this article, a combined deep learning prediction (CDLP) model including two paralleled single deep learning models, CNN-LSTM-attention model and CNN-GRU-attention model, is established. In the model, a one-dimensional convolutional neural network (1DCNN) is used to extract traffic flow local trend features and RNN variants (LSTM and GRU) with attention mechanism are used to extract long temporal dependencies trend features. Moreover, a dynamic optimal weighted coefficient algorithm (DOWCA) is proposed to calculate the dynamic weights of CNN-LSTM-attention and CNN-GRU-attention with the goal of minimizing the sum of squared errors of the CDLP model. Then, the neuron number, loss function, optimization algorithm, and other parameters of the CDLP model are discussed and set through experiments. Finally, the training set and test set for the CDLP model are established through the processing of traffic flow data collected from the field. The CDLP model is trained and tested, and the prediction results of traffic flow are obtained and analyzed. It indicates that the CDLP model can fit the change trend of traffic flow very well and has better performance. Furthermore, under the same dataset, the results from the CDLP model are compared with baseline models. It is found that the CDLP model has higher prediction accuracy than baseline models.