Abstract: This paper introduces the design scheme of a fire alarm system based on ARM9 multi-sensor data fusion. The system mainly consists of a sensor module, an A/D conversion module, an S3C2440 control module, an alarm module, and an execution module. The fire detection primarily employs a fuzzy inference data fusion algorithm, with the controller handling both fire data processing and algorithm implementation. Operational results show that this method can improve the system's accuracy and reliability.
Keywords: ARM9; data fusion; fuzzy inference
Abstract: This paper introduces design of the fire alarm system based on multi-sensor data fusion of ARM9 controller,This system is composed of sensors,A/D converter,S3C2440 controller,alarm and executing modules.
Keywords: ARM9; datafusion; fuzzyreasoning;
Given that current single, ordinary type of fire detectors and alarms can no longer meet the needs, using multiple sensors to comprehensively collect various abnormal information before a fire occurs, and using multi-sensor information fusion technology to process the fire information provided by the sensors, can greatly improve the reliability of the entire alarm monitoring system.
1 System Hardware Design
The core controller of this hardware system is the Samsung S3C2440, a 16/32-bit multi-functional, low-power embedded processor with an ARM920T core. The S3C2440 is a high-end embedded microprocessor from Samsung, suitable for industrial control, smart home appliances, and other portable product development. Its 400MHz clock speed fully meets the real-time processing requirements of fire monitoring and alarm systems. Its main control chip and rich peripheral interface circuitry can be used to connect to various digital devices for data exchange. The fire early warning system based on multi-sensor data fusion adopts a modular structure, mainly composed of a sensor module, an A/D conversion module, an S3C2440 controller, an alarm module, an execution module, and the necessary power supply and memory modules for the controller. Figure 1 shows the system structure diagram.
The multi-sensor module consists of multiple sensor groups, each comprising a temperature sensor, a combustible gas detector, and a smoke detector. In this system, the temperature sensor is the HM500 from HuMIREL (France), known for its low cost, small size, long lifespan, high selectivity, and good stability. The combustible gas detector uses the latest infrared gas sensor developed by Shenzhen Jianda Technology Co., Ltd. These detectors are installed in hazardous indoor and outdoor locations where the detected gas is prone to leakage. They are highly sensitive to low concentrations of pollutants in the air, exhibiting high sensitivity to odors, CO, H₂, and O₂, and can even detect concentrations of a few ppm. The smoke detector uses the GE 514C smoke detector from the United States, featuring self-diagnostic capabilities and drift compensation to resist interference caused by dust. These sensors monitor multiple parameters of the fire process. The data is transmitted to the S3C2440 controller via an A/D conversion module, and with intelligent discrimination technology, early warning, reduced false alarms and missed alarms, and improved reliability can be achieved. The A/D conversion module used in this design is the TI ADS805 12-bit high-speed parallel converter, which features high sampling speed and good stability.
The S3C2440 controller uses a 16/32-bit RISC microprocessor with a 6-layer PCB process, featuring low power consumption and high-speed processing capabilities. Its simple and stable design is ideal for products with high power requirements. It employs a new bus architecture (AMBA) with a 32-bit advanced processor core. Its maximum clock speed reaches 533MHz, ensuring real-time performance even when processing large amounts of sensor data. Its power management module can provide multiple voltage options, including 1.8V for the chip core, 3.3V for the chip's I/O sections, and 5V for some external integrated circuits. This intelligent power management module effectively addresses the different power requirements of various system components, reducing power consumption, minimizing interference noise between different power supplies, and improving system integration. Its memory module includes two SDRAM chips totaling 64MB and one 64MB NAND flash (K9F1208), with other capacities of NAND flash memory available as needed. The memory module stores the system's operating program and data collected by sensors during fire monitoring.
The alarm module primarily activates audible and visual alarm signals to notify on-duty personnel when a fire hazard is detected, allowing them to take appropriate measures. The execution module activates nearby fire extinguishing devices when a fire occurs, minimizing the damage caused by the fire.
2 System Software Design
The software portion of this system mainly includes power-on initialization, system self-test, clock initialization, interrupt settings, and peripheral initialization. Then, the main program (main()) function is run. After task creation, the extended port controls the ADC to switch channels for data acquisition. After preprocessing the data such as smoothing filtering and calibration compensation, a data fusion algorithm is used to analyze the fire situation, determine whether an alarm should be triggered, and perform cyclic detection. The system software flowchart is shown in Figure 2.
Unlike typical data acquisition and processing systems, this system's data acquisition and processing software both run on the processor. Besides continuously acquiring the latest fire scene data, the system software also performs real-time data processing. The system is based on an 8/16-bit microcontroller. Programs are typically written in a foreground/background manner, with a large infinite loop running in the background and multiple interrupts in the foreground. This approach becomes inadequate as the program size increases, system functions become more complex, and especially with a large number of concurrent modules, making it difficult to guarantee the real-time performance of measurement and control. Furthermore, programming is difficult and it's inconvenient to add functionality. Considering software complexity, computational load, and real-time requirements, the system adopts the μCOS-II operating system.
Most tasks in this system are scheduled to run by calling OSTimeDly(), and each task can be assigned a different timing interval through system functions. The ADC data acquisition program, data processing program, and data fusion algorithm program are mainly written in embedded C language.
3 Data Fusion Algorithm
The application of multi-sensor information fusion technology in industrial process monitoring systems has yielded some engineering results. In these systems, after sensors collect data from objects and the environment, the data undergoes fusion processing before being used in control strategy calculations. Currently, commonly used information fusion methods can be broadly categorized as follows: First, classical methods based on estimation and statistics, including weighted average, least squares, and DS evidence theory; second, information theory-based fusion, including template methods and entropy theory from cluster analysis; and third, artificial intelligence-based fusion methods, including fuzzy logic, production rules, neural networks, genetic algorithms, fuzzy integral theory, and expert systems.
When applied to multi-sensor information fusion, we consider A as the set of possible system decisions and B as the set of sensors. The element μi in the relation matrix R A+B between A and B represents the probability of inferring decision i from sensor i. X represents the credibility of each sensor's judgment. Y, obtained after fuzzy transformation, is the probability of each decision.
Specifically, we assume there are m sensors observing the system, and the system may make n decisions. Then: A: { y1 /decision, y2 /decision, ..., yn /decision n} B: { x1 /sensor, x2 /sensor, ..., xn /sensor m} The sensor's judgment on each decision is represented by a membership function defined on A. Let sensor i's judgment on the system be: μi1 /decision, μi2 /decision, ..., μin /decision n, 0 ≤ μy ≤ 1. That is, the probability of the result being decision j is μij, denoted as vectors μi1 , μi2 , μi3 , ..., μin . Then the relationship matrix A×B formed by the m sensors is:
The confidence level of each sensor's judgment is represented by its membership degree on B: X = { x1 /sensor1 × x2 /sensor2 …, xn /sensorn}. Then, by performing a fuzzy transformation according to Y = X * RA * B , we can obtain: y = ( y1 , y2 , y3 , …, yn ). That is, the probability of each decision after comprehensive judgment is y. Finally, the possible decisions are selected according to certain criteria (such as the maximum membership degree method, the center method, etc.) to obtain the optimal result. When making decisions based on the calculated y value, the following rules should be noted: ① The decision result should have the maximum membership degree. ② The membership degree of the decision result must be greater than a certain threshold (generally 0.5). ③ The difference between the membership degree of the decision result and the membership degree values of other decisions must be greater than a certain threshold (e.g., 0.1).
4. Data fusion experiment for fire monitoring
This design uses temperature sensors, combustible gas detectors, and smoke detectors for fire monitoring. The data fusion method is shown in Figure 3.
Figure 3 illustrates a general method for data fusion based on fuzzy reasoning. In a fire fault monitoring system, the weights of each sensor are first determined. In this design, the weights for the temperature sensor and the combustible gas detector are set to W1 = 0.5, W2 = 0.3, and W3 = 0.2, respectively. The final decision is divided into two types: fire present (Y1 ) and no fire (Y2 ) . Based on the current operating state, the membership function of each sensor X for each decision Y is determined. Then, a linear transformation is performed to determine the final result. For example, at a certain moment, the membership degrees for fire presence or absence are determined based on the temperature sensor data as μ11 = 0.45 and μ12 = 0.55, and based on the smoke sensor data as μ31 = 0.9 and μ32 = 0.1. A linear transformation is then used to obtain Y.
Based on the results, there is a fire hazard, and an aerosol fire extinguisher should be activated. Table 1 shows the experimental data of fuzzy fusion in the fire fault monitoring system.
5. Conclusion
Applying fuzzy inference data fusion to a multi-sensor cable fire fault monitoring system demonstrates higher accuracy and reliability compared to single-sensor systems. Operational results show that this method is practical and effective in improving the reliability of fire fault detection and reducing false alarm rates. However, this method also has some drawbacks. For example, there is no unified theory regarding sensor weights and the allocation of membership values for each sensor to the decision, requiring empirical adjustments.
References:
[1] Han Junfeng, Shi Yuqiu, Fuzzy comfort sensor system based on multi-sensor information fusion technology [J]. Sensors & Microsystems, 2007, 26(1): 60-61
[2] Lu Yanhe, et al. Information fusion in fire early warning systems [J]. Network Information Technology, 2006, 25(3)
[3]Labrosse J J.Micro uC/OS-II.The Real-time-Kernel (Second Edition)[M],USA:CMP Book,2002
For details, please click: Implementation of a Fire Alarm System Based on ARM9 Multi-Sensor Data Fusion
