Method of Fault Data Analysis of Sensor Node based on Wireless Data Communication Protocols Seung-Ki Ryu * Research Fellow, Korea Institute of Civil Engineering and Building Technology, Korea (*corresponding author) skryu@kict.re.kr ABSTRACT In general, sensor node system of wireless communication is configured by measuring the status of the facilities. For example, due to pressure, loads and stresses acting on road facilities upon the external environment, structural deformation of the facility may be generated. Structural deformation leads to the accident of the ground collapse, and it can result in numerous fatalities. The accident can be prevented by sensor installation for detecting changes in the external environment in advance. A traffic detection system that measures the flow of vehicles on road shoulders is installed in order to analyze the distribution of the traffic volume of road network. Traffic observation point has a certain node and link system, and the link consists of points encountered each other or by distance. The collected data at the measuring point is transmitted to the center by using a wired or wireless communication network mostly. At this point, the center must distinguish whether the received data is normal or not. To do this, the center of the server system sends a type of fault code of site equipment with the traffic information. The center determines the operational status and fault type of site equipment installed extensively and analyzes the failure type codes as schematically. However, as the failure code is too complicated, a lot of types of combinations are generated, and in most cases we do not know the exact cause of failure. Although each manufacturer has its own fault code system, this method is not convenient because it requires transcoding. This study proposes a method for estimating the failure primarily based on data communication protocols. Keywords: Fault Code, Fault Diagnosis, Fault Data Clustering, Data Communication Protocols 1. INTRODUCTION The measurement process is carried out for the effective management after opening the road. The research technique of the national road traffic is one of a series of methods in traffic volume survey. It uses realtime traffic measurement equipment, and the data collected from the equipment is stored in the center server every day. The road traffic DB (database) is the national statistics used in road planning and construction, road pavement, road safety improvements, and road traffic research including a variety of fields. piezoelectric sensor in many ways in order to increase the accuracy of the information acquisition. Typically, there are two types of sensor arrangement. One way is to place an induction line between two piezoelectric, and the other is to place a piezoelectric sensor between two inductions lines. In recent years, there is the trend to unify the first method to have good signal characteristics of the piezoelectric sensor. Unlike the piezoelectric sensor induced ray sensor is good and has the impulse signal characteristics. It is possible to clearly distinguish the pulse cycle characteristics when a vehicle passes. The field equipment continuously transmits the traffic volume data to the center sever by using wired and wireless networks. As the field equipment is exposed to the outside for a long time, performance degradation occurs with climate change, smoke, traffic accidents, corrosive, and aging. Traffic measurement equipment is a roadside apparatus for measuring the point-based traffic information such as traffic volume, vehicle type, and speed. This device is embedded in the road by combining the loop sensor and the piezoelectric sensor, and it detects electrical signal when a vehicle gets through the embedded sensor. Since loop sensor is installed in the depth of 30mm~50mm below the line leading to the road surface, a certain amount of electromagnetic field in the induction line is generated. As this method utilizes the variation of the electromagnetic field generated in the induction line, the unusual vehicles and environmental factors causes a measurement error. The field equipment has been used in combination with the induction line sensor and Fig 1: The concept of traffic volume survey system The induction line sensor provides a signal that can be detected to know the start and the end of the vehicle and the piezoelectric sensor between inductions generates signals that can be analyzed for the number of vehicle axes, axis-to-axis distance, and passing speed. The data collected by the field devices are transmitted to the central server via a wired or wireless communication terminal provided in the field device, and the center system consists of a firewall, a linked server, a 54
web server, a data acquisition server, a statistics server, the local server, and router communication. The center hardware system is composed of a physically common server device. The center software system stores data for 1 hour, 24-hour, and weekly in the field device and transmits the data by connecting the one-to-one field equipment. Fig 1 is a center hardware configuration and represents the physical structure of the field devices and the center system. The center system stores the data every five minutes, also stores the data again in 1 hour. At that time, the 5 minutes data from center DB is combined with the data from field equipment to store the 1 hour data. However, two different 1 hour data is generated and stored differently because of the data deficient during transmitting. In other words, the conditions such as communication period and delay cause the time difference between the field and the center, and as a result, it generates a big error. Thus, the testing of the integrity that compares the field equipment and center data is conducted in this study. This process is a way to directly compare the data from the data center and field equipment. After searching the accumulated past DB, it analyzes the data pattern. If the graph is to show an unusual form which is mismatched daily and weekly patterns, the administrators intuitively determines that there is a problem in the field of data acquisition equipment. This approach takes a considerable time to analyze the active state of the field equipment, and there actually is a limit to determine the type of failure to collect device-specific details. Also, since the field device has a separate lookup failure for each product, the center is inconvenient to have to drive each of the matching for different products. This study proposed a method for estimating the failure of field equipment based on communication protocol rather than traditional approach that combine a lot of fault codes and types that occur in the field equipment. 2. RELATED WORKS 2.1 Reviews on Existing Research Most equipment that is consisted of induction lines, incoming lines, and controllers uses inductance signal. This device is used for national traffic survey system and a part of it is applied to the intelligent transportation system (ITS). The field equipment is that a buried inductance coil generates electromagnetic fields, and when supplying the micro current, electromagnetic fields are generated. The sensor generates a frequency of 10 khz to 200 khz in order to form a uniform induced magnetic field. The vehicle and speed are detected when a vehicle passes over an induced magnetic field through observing slight magnetic variation. The inductance coil forms an induced magnetic by sending energy having specific frequency to the inductance coil through an incoming line, and the magnetic flux changes depending on a vehicle s passing, and then the magnetic flux variation changes inductance. In addition, the piezoelectric sensor is a type of a ceramic-based and it generates a signal of an impulse pattern when the vehicle is passing and measures the number of axes, axis-to-axis distance, and speed of vehicle. Ryu and Lim had studied clustering techniques for data error of traffic information systems, and they presented a failure criterion by analyzing the data types in their paper [1]. The field equipment to measure the traffic volume has been applied to the installation guidelines that modified the US standard guidelines [2]. US Department of Transportation has circulated a production and installation manual for the traffic information field equipment, and it presents a data aggregation methods such as the placement of induction lines sensor and piezoelectric sensors, the time-averaged data, the average daily data, and annual data [2]. Lee et al. studied splash over which is malfunction symptoms because of the sensitive sensor [3]. In addition to these research papers, there are papers on power surges among failure types. Ryu et al. proposed a method for stabilizing the output voltage for the input power instability of the electromagnetic sensor [4]. 2.2 Fault types of Existing Field Equipment The field equipment has failures frequently since it is exposed to the weather conditions. As a result of failure type analysis, a malfunction of power and communication occupies about 80%. The power supply issues account for most and the next is a communication error problem. Unlike the before study of the power problem [4], this study proposes a method for determining a communication problem. The first step is to identify the fault code in the field equipment because the field equipment problems occur in the communication with the center. The following table is a fault code of the field equipment. The cause of the failure is not easy to interpret because it is common to be multiple codes occurring in combination. In other words, there is a limit to interpret the code for all the fault situations as fault codes are often complex error than single errors. 55
Table 1: The Classification of error code Code Meaning Features 1 Power supply error 2 Low battery warning 4 Modem connection error 8 Laptop connection error 16 CPU reset 32 Detecting board error by lane 64 Front door open Single sensor failure 128 Rear door open 256 Type A sensor 1 lane #1 error 512 Type A sensor 1 lane #2 error 1,024 Type A sensor 2 lane #1 error Designated lane specific sensors(omitted below) 0001 External memory and recall errors 0002 Back-Panel parallel communication error 0004 Serial communication error b/w remote spot and lane 0010 Number of control data error by lane 0020 Type A sensor #1 error 0040 Type A sensor #2 error Single controller failure 0100 Type B sensor #1 error 0200 Type B sensor #2 error 1000 Speed detection error 2000 Particular vehicle, No data for 10 cycles(1 cycle=5 minutes) 4000 Normal vehicle, No data for 10 cycles(1 cycle=5 minutes) 0080 b4, b6 0192 b6, b7 52224 b24, b25, b14, b15 Multiple sensor failure Number of different cases occurring According to the table, the number 1 to 1024 of the controller code may be determined by the failure of single sensor unit, and the rest of the codes are omitted to fill out the table because those code are generated by the composite error. 56
3. FAULT DETERMINATION METHOD BASED ON DATA PROTOCOL 3.1 Data Communication Protocol The roadside controller transmits the data through a predetermined communication protocol to communicate with the center. Data transfer protocol between the roadside controller and the center system transmits the data in accordance with predetermined rules as the data transmission and reception procedure of each other. Fig 2 shows the data transfer process that is conducted among the controller, the roadside communication device, and the center server systems. In other words, the data communication is attempted by the protocol in 5 minutes. At this time, if the connection fails, 12 data sets are created for one hour and the accumulated data sets for 24 hours are 288, which is a process for determining a data transmission success. 3.2 Clustering method of Data Communication Error The clustering method of error data sets uses a transient error data or a long-term failure data based on 288 data sets considering error frequency, error distribution, and error persists. The failure data can be analyzed according to the failure data access rate of the communication protocol, and in this study, the frequency distribution were analyzed by testing the connection data of 288 cycles. Fig 2: Data communication protocol It is to correspond to a random value generated in the mutual authentication process by the server and to transmit the data encrypted or not. The data structure is composed of DATALEN (packet data length), CMD (command), DATA (data), and the respective sizes are 2bytes, 2bytes, and N bytes. The collection period is 5 minutes for the error code and the accumulated data sets for one day are 288. This study proposes a method of diagnosing the operational state of the equipment on the basis of the communication result of the data connection in 5 minutes. Fig 3: Data access rates of communication protocol Fig 3 shows the results of protocol access rates by lane and by car. The collected data indicates the data communication connection state between the roadside controller and the central server of the three months from January to March. The AVC (AC Voltage Control) equipment was installed to collect sensor targets for each lane, and the traffic environment and a communication state was different. The connection state for data is poor in January and March, and the connection state is stable in February. The blue line on the graph is the reference value of the data connection stage, 288, and the red line 57
shows the daily data values by the connected state of equipment. Performing normal data communication protocol based on the roadside controller and the center has 288 data sets during a day. However, even if 288 data sets are not created because of temporary disability, it is not determined a malfunction since there is a case that the data is stored at roadside controller. Therefore, there is a limit to make the fault determination logic using only one day data. In this study, the entire data analysis period was extended from one day to three months because it is not possible to determine a failure of equipment by the data of just one day as failure diagnosis reference. As a result of analysis for data communication protocol during three months, only one lane has higher failure rate of protocol since mid-march. The percent error value (%Error) was calculated based on the above data protocol graph. value of 20% or greater indicates that the data state and the protocol are in poor condition. A data set of L1 to L4 means error data occurred a sensor by lane. Some of error values for L3 are more than 20%, which showed an abnormal distribution of data. Other data showed the distribution of data within the normal 20% below. Therefore, it means that the sensor equipment of lane #3(L3) is failed. Fig 5: Percent daily values based on the error distribution of the data communication protocol Fig 4: 5 minutes data connection results (% error) Fig 4 represents the data access results to produce 5 minutes percentage error value and it is the data for three months for four lanes and an AVC equipment. The x-axis represents time in three months on a daily basis, and y-axis is percent error value by lane. Fig 6 shows the percent compared to the reference value for re-distribution of the error value, and by this figure, the normal and the abnormal condition of the equipment could be clearly distinguished. Also, failure determination threshold could be set. For about 10 days from the data of the three months, we can see that a single lane connection is poor. As a result of three months cumulative data analysis, there was no case that is completely generated 288 data sets for a day and percent errors from the data sets were produced from a five minutes data value in a non-connection rate within 20%. Only for one lane, the deterioration of data communication protocol performance has occurred since mid-march. 4. ANALYSIS RESULTS OF FAULT DIAGNOSIS 4.1 Standards of Fault Judgment Data communication protocol between the roadside controller and the center system is based on the data connection value of 288 data sets and the failure determination reference of AVC equipment could be set as a result of data analysis for three months. Fig 5 represents the percent error values as a point distribution and the low density points could be distinguished through the point distribution. The error Fig 6: Communication connection failure criterion 4.2 Compatibility of the Results The error clustering criterion was set to 20% based on the access rate of the data communication protocol, and long-term data connection was not made after analysis of the actual system. As shown in Fig 6, the error percent value of the lane determined by a communication failure was distributed between 31.6% ~ 70.14%. The annual daily traffic volume and one hour 58
data beyond the reference value were analyzed in order to validate the proposed clustering method. However, for analysis of the data collected during this period, it did not measure 50% of the actual speeds of 100 passing vehicles. It proved that the vehicle speed value is not accurately collected from the traffic information. Thus, when the percent error value is over 20% based on the communication access rates, it is possible to determine that a problem has occurred and the collection performance of the field equipment is abnormal. 5. CONCLUSIONS This study proposed a method of determining the presence of failure in the remote roadside controller. Also, this method is effective and simpler than the conventional method. As mentioned before, it was not possible to correctly interpret the fault code because each manufacturer has a different fault codes. So, it was timeconsuming to determine the presence of equipment failure using complex fault codes, and also finding the cause was not only a simple process in reality. ACKNOWLEDGEMENTS This research was supported by a grant from Global Technology Research Project (Development of High-reliable USN Sensor Node and Data Analysis System for Ground Collapsing Prevention, 2014) funded by the KEIT. REFERENCES [1] S. Ryu and S. Lim, Clustering Scheme of Data Communication Error for Traffic Detection System, 61th KIEE (Korean Institute of Electrical Engineers) (2013), 1201-1202. [2] FHWA, Traffic Detector Handbook: Third Edition, FHWA-HRT-06-108 (2006). [3] H. Lee and B. Coifman, Identifying chronic splashover error at freeway loop detector, Transportation Research Part C: Emerging Technologies 24 (2012), 141-156. In this study, failures were determined using a communication protocol access rate value rather than interpreting the complex fault code. The access rate of the communication protocol is how to represent the connection status of the field devices and the center system. As a result of analyzing the distribution of access rate, we could set an optimum threshold value. Through this study, it is possible to predict collection performance of field equipment by using only the communication access rate. It can be said that in this method, it is possible to efficiently manage field equipment because the failure point can be estimated in advance. Furthermore, it can not only improve the utilization rate of AVC equipment but also support a predictable maintenance work schedule. [4] S. Lim, J. Yang, and S. Ryu, Improvement of detection method and power supply device of vehicle detector, Advanced materials research 787 (2013), 478-484. AUTHOR PROFILES Seung-Ki Ryu received the degree in electrical engineering at the Chung Buk National University in Korea. Currently, he is a research fellow at Korea Institute of Civil Engineering and Building Technology. His research interest covers intelligent transportation systems, information technology, ubiquitous city, construction IT convergence and logistics. 59