TMCnet News
The Improvement on Agricultural Machinery Safety Using Wireless Sensing System and Stability Control Algorithm [Sensors & Transducers (Canada)](Sensors & Transducers (Canada) Via Acquire Media NewsEdge) Abstract: In this paper we presented an autonomous wireless system for automatic identification of farm implement that can significantly enhance safety of agricultural machinery. The prototype system we have designed and implemented uses the vibrations of the farm implement to generate the energy required for system operation. Using vibration energy, our system is able to communicate to the farm tractor the implement information, which allows improving the stability control system of the agricultural machinery, comprised of both the tractor and the implement. This system acts as pre-crash safety device, because it is designed to jointly work with a commercial control system solution, which is dedicated to warn the farmer of potentially hazardous working condition of the farm tractor. Copyright © 2014 IFSA Publishing, S. L. Keywords: Agricultural Machinery, Wireless Sensing System, Stability Control Algorithm. (ProQuest: ... denotes formulae omitted.) 1. Introduction During last couple of decades several electronic systems have been pervasively introduced in the automotive market. In particular, huge efforts have been spent to integrate advanced embedded systems into cars, e.g. increasing the active and passive passengers' safety. Similarly, the integration of such safety systems is of course highly desirable also in agricultural machineries. Unfortunately the adoption of advanced control system in the agricultural market is pursued with a much slower pace [1-2]. Up to now, most of the efforts have been devoted to bring passive protection systems into agricultural machineries, like roll-bars and seat belt, whereas active systems for crash or accident prevention have attracted less interest [3]. National and international statistics confirm that tractor accident is one of the highest causes of fatality. During the period May 2009-May 2010, 296 accidents have been reported in Italy leading to 174 fatalities, 140 of whom were farm tractor drivers. Similarly, in the United States 1412 farms workers died from tractor overturns between 1992 and 2005 [4], Tractor overturns represent the main cause of agricultural machinery operator deaths. For instance, a large percentage of these accidents refer to off-road environment, especially in fruit and wood related agricultural operations. Moreover, statistics also show that most of the accidents involving machineries occur due to human errors, e.g. typically related to operators ignoring the machinery safety warnings or taking shortcuts to save time. In this scenario, there is an evident need for on-board control systems capable to enhance both machinery crash prevention and machinery operator safety. Only few safety systems have been introduced since 90's [5], and mainly devoted to vehicle orientation and inclination control monitoring. These devices generate warning alarms that inform the driver if the tractor is operating in a dangerous condition. They embed accelerometers and gyroscopes to collect real-time information about farm tractor stability, allowing more complex and automatic dynamic vehicle stability control. The main limitation of such system is that no information about connected implements are provided to the tractor control unit. This causes the occurrence of dangerous working condition for the whole agricultural machinery because the vehicle dynamics varies significantly when an implement is connected to the tractor. At the best of our knowledge, none of the current control systems takes into account any change in the dynamics of the agricultural machinery. This is mainly due to two reasons: i) Most of the implements available on the market do not have an ECU on board, and consequently they cannot exchange any data with the tractor; ii) In the agricultural environment the use of wireless systems is limited by their battery lifetime, which is not comparable with the typical lifetime of the implements. For this purpose, we propose a wireless system designed to be directly installed on the implement that communicates with the tractor to provide the implement characteristics (e.g. weight, sizes, the center of mass, etc.). Therefore, the parameters describing the agricultural machinery can be adjusted in the tractor control unit to consider the farm tractor with the connected implement. With our system, the ECU can instantaneously detect hazardous working condition of the tractor with the connected implement. This increases the vehicle stability control algorithm effectiveness and enhances the active safety of the agricultural machinery. Moreover, the proposed wireless device can harvest energy from natural vibrations occurring on the farm implement during its normal operation. The collected energy is stored in a capacitor acting as short term energy reservoir, which preserves the secondary back-up battery from performing several charge and discharge cycles. In this way, the vibration harvester sustains the regular operation of our system, while the battery is used only after long period of implement inactivity (e.g. during winter period). Thus, the combination of low-power wireless architecture with the energy harvesting technology enables our system to overcome both of the limitation associated with the usage of wireless devices in agricultural environment. The paper is organized as follows. Section II describes the architecture of the system. In Section III we describe the vibration energy harvester we have designed and realized, while Section IV presents the experimental results derived from the system characterization. Section V concludes the paper. 2. System Architecture The proposed wireless system allows enhancing the farm tractor safety by delivering the implement characteristics to the tractor control unit. The system is made of two devices: the on-implement End device (ED) and the on-vehicle control unit, i.e. Master device (MD). Fig. 1 shows the system architecture. The Master device is designed to be located on the tractor and connected to the Multisensory control system manufactured by COBO Group [6]. Thanks to our system, the tractor control unit can automatically identify the connected implement and adjust the parameters of the tractor stability control algorithm accordingly. The identification is performed by the MD, which receives the implement characteristics (e.g. weight, sizes, the center of mass, etc.) sent by the on-implement ED using the wireless channel. Then, the MD forwards the implement characteristics to the tractor ECU that uses such information to update the vehicle parameters, allowing a proper execution of the tractor stability control algorithm. 2.1. On-Vehicle Master Device The Master device acts as Communication Bridge between the end device installed on the implement and the tractor ECU, i.e. the Multisensory control system. The communication between MD and ED uses the IEEE 802.15.4 protocol, while the link towards the multisensory platform uses a standard UART serial link. The MD device is based on a Free scale MCI3224 System-on-Chip (SoC) [7], which architecture allows minimizing the power consumption while obtaining the required performance. 2.2. On-Implement End-Device The end device (ED) represents the heart of the control system. Its role is to send the characteristics of the connected implement to the MD. The ED architecture adopts the same free scale MC 13224 SoC device used for the MD node [7]. This device internally runs both our software and the IEEE 802.15.4 stack required for the wireless communication. The main characteristic of the adopted CPU & RF chip is its ultra-low power consumption in both active and sleep mode of operation. Nonetheless, the key aspect in order to achieve ED's effective operation is to select the proper duty cycle between active and sleep modes, which is a tradeoff between opposite needs. Indeed, increasing the du ration of the active time is desirable to frequently update the implement status, but this significantly increases the ED power consumption, shortening the battery lifetime of the device. In order to relax the power consumption constraints, we included an energy harvesting module in the ED that gathers energy from the natural vibrations occurring on the implement during its normal working conditions. The vibration energy harvester uses a piezoelectric transducer and allows the ED to autonomously recharge its energy reservoir during the implement usage, extending the battery lifetime. The vibration energy harvester is described in the next Section. 3. A the Vibration Harvester System 3.1. Data Obtaining The most innovative part of the ED is the vibration energy harvester that converts the kinetic energy associated to the natural vibrations of the implement into useful electrical energy. The architecture of the vibration energy harvester is sketched in Fig. 2. It is comprised of three blocks: 1) An off-the-shelf piezoelectric PZT transducer; 2) An AC-DC converter that is custom designed for this application; 3) A power management block that controls the energy distribution. While the PZT simply converts the mechanical vibrations into an AC electrical voltage, the AC-DC converter transforms the input AC voltage into a stable output DC voltage, which is stored in a 1 rnF capacitor that acts as an intermediate energy buffer. The AC-DC converter maximizes the extraction of charges from the piezoelectric transducer by exploiting voltage multiplier architecture [8]. The following block is the power supply management that to optimize the energy provision to the ED node regardless its energy consumption variations (i.e. changing from sleep to active changes the current consumption of several order of magnitude). Thus, in order to optimize the system energy usage, the power supply management circuit autonomously decides to drain the required energy from either the intermediate capacitor or the back-up energy storage, which is given by a Li-Ion battery. The selection of the more suitable energy source (e.g. either the intermediate capacitor or the battery) depends on the charge status of the intermediate capacitor. The power supply management circuit starts draining energy from the capacitor when its voltage reaches the battery voltage ( VBAT = 3.6V ), and keeps using the capacitor energy until its voltage drops down to 2.5 V, which is the minimum voltage required by the voltage regulator to properly operate. In this condition, the energy stored on the intermediate capacitor is E^ = 3AmJ. The effectiveness of the proposed approach is strictly related to the capability of the energy harvesting circuit of collecting energy from vibrations. To this aim, the design of the harvesting circuit is crucial, and depends on the strength and frequencies of the vibrations occurring in the considered target environment. We investigated the natural vibrations of some typical implements by collecting experimental data during standard operation conditions. Such data were registered by our industrial partner using 3 axial accelerometers in field tests of their systems, which was comprised of different tractors and different implements. This provided a statistical description of the available frequencies and acceleration of vibration present on different implements connected to different farm tractors. Collected data suggested that the most useful vibrations have common frequencies of ~ 1 kHz with typical accelerations up to 2 g for all the observed implements and tractors under test. We used such data to perform an extensive characterization of the piezoelectric transducer (PZT), which converts the kinetic energy into electrical energy in the form of an AC voltage [9]. Fig. 3 shows the frequency response of the adopted PZT transducer. As shown, independently of the stimulus accelerations (e.g. 2 g, 1 g and 0.5 g) the resonance frequency of the transducer peaks around 1 kHz. There is a slight shift of the resonance frequency when changing the acceleration, which however does not affect significantly the capability of the PZT transducer to effectively convert the vibration energy at frequencies around 1 kHz. Fig. 4 shows the characterization of the output power generated by the PZT transducer [9] when subject to three different vibration stimuli: 2.0 g at 980 Hz, 1.0 g at 1010 Hz and 0.5 g at 1060 Hz. Measurement results show that the adopted PZT transducer generates up to 724 pW at 2.0 g, while the output power is equal to 167 pW and 32 pW at 1.0 g or 0.5 g, respectively. Such data represent the maximum power the PZT can provide at different accelerations, and constitutes the upper limit for an ideal energy harvester system with 100 % efficiency). This information is crucial also for the calibration of the power management system. Indeed, in order to maximize the system lifetime, the energy consumption should be lower or equal than the energy harvesting rate [10]. To obtain an autonomous behavior, the duty cycle of the ED device must be adjusted accordingly to the amount of energy collected by the vibration harvester. As typically occur in wireless sensor network (WSN) applications, the ED spends most of its time in a deep sleep state, allowing the system to have ultra-lowpower consumption and a very long battery lifetime. When the ED device is active (its main task is to send the implement information to the MD) the power consumption can't be drastically scaled down without losing performance. Thus, the power and execution time required by the ED to perform its tasks have to be carefully characterized. Fig. 5 shows the measurements of the ED current consumption when supplied with either the maximum (3.6 V) or the minimum (2.5 V) allowed voltage. From these measurements we can observe that the required execution time, Tac, is constant and close to about 5 ms. The peak current does not significantly depend on the supply voltage (26.5 mA at 3.6 V, 28.2 mA at 2.5 V), leading the total energy consumed at the two voltage supply level to be comparable (192 pJ at 3.6 V and 185 pJ at 2.5 V). Thus, we assumed that the energy consumption for the ED in the active mode, Eact, remains the same, -200 pJ (it has been slightly overestimated to have some margins). On the other hand, the current consumption of the ED running in sleep-mode is around 30 pA with the supply voltage is either 2.5 V or 3.6 V, and the power consumption is Psleep = 75 pW. Thus, the average power consumption of the ED, Ped, depends on the time interval between two subsequent communications, Tsleep. ... (1) For example, assuming the system will wake-up, and transmit data every 10 s, the resulting energy and power consumptions will be (2): ... (2) In this kind of application the maximum time interval allowed between two subsequent data transmissions is fixed for safety reasons to 30 s. The resulting duty cycle is very low, as typically occurs in WSN applications. Since Tsleep » Tac, Eq. (1) can be simplified as follows. ... (3) Of course, reducing the sleep time period, i.e. the time interval between subsequent communications raises the power consumption, reducing the period where the capacitor alone can sustain the ED activity. Considering the maximum time interval of 30 s between subsequent communications, we can calculate the maximum duration of the charge of the intermediate 1 mF capacitor, Dcap. ...(4) If the ED activity increases D cap reduces. This gives the opportunity to include additional power management policies: for example, we could decide either to have more frequent implement data transmission, thus increasing the ED power consumption if the battery is full, or reduce the transmission frequency to allow the charge of the battery while the ED node continues its activity. 4. Experimental Results 4.1. This is a Subtitle Example Fig. 6 shows a picture of the implemented prototype of the wireless ED. The performance of the vibration harvester embedded in the ED has been evaluated monitoring the voltage across the intermediate capacitor, which is used as energy reservoir. The measurements confirm that battery lifetime is increased by exploiting the vibration energy harvester we realized. Fig. 7 shows the voltage measured across the intermediate capacitor during the typical implement working conditions. Vibrations are generated using an electro-dynamic shaker capable to control both the vibration frequency and acceleration [11]. As shown in this figure, the capacitor is charged by using the energy generated from vibrations of 2 g at the frequency of 980 Hz. In addition, the capacitor discharges when the ED wakes up and becomes active. In this test the ED was sending its identifier every 10 s and staying idle in the rest of the time, thus emulating system power consumption larger than usual. This test demonstrates the capability of the harvester to sustain the node activity and data transmissions alternating the usage of both the intermediate capacitor and the back-up battery. The ED is battery-powered when the energy generated by the piezoelectric transducer is stored into the capacitor. On the other hand, when the voltage across the capacitor reaches 3.6 V the power management circuit disconnects the battery, and uses the capacitor to supply the energy required the data transmission to the system, discharging again the capacitor. Once the voltage across the capacitor drops to 2.5 V, which is the minimum supply voltage allowed for the ED, the power manager switches back to the battery-power, allowing the capacitor to charge again. Obviously the power consumption of the ED during the idle state is lower than the consumption associated with data transmission, as highlighted by the waveform [12]. Nonetheless it is crucial to reduce the average power consumption as low as possible using both sleep mode and longer period between transmissions to maximize ED performances. Furthermore, the system performances have been evaluated under different working conditions, emulated by generating different vibration stimulus by using the shaker. These measurements have been performed both at the start-up of the system and in stationary conditions. At the start-up, the intermediate capacitor is fully discharged, e.g. as after a long period of inactivity, and it needs to be charged to 3.6 V before it can be used as energy source. In stationary conditions, the voltage across the capacitor varies in the range 2.5 V-3.6 V. The results are summarized in Table 1, where íchrg corresponds to charging time of the capacitor, Pharv is the average power provided by the harvester module, and Epzt is the energy generated by the piezoelectric transducer. Measurement results confirm that increasing the strength of the vibrations reduces the charging time and increases the power provided by our harvester to the load. This is true both when charging the capacitor for the first time and when the system works in stationary condition. It is also worth noting that the total amount of energy extracted from the PZT, Epzt, to charge the capacitor slightly increases for vibration stimuli with lower strength. This is due to the lower conduction losses associated with smaller generated current occurring at smaller vibrations [13]. It is still important to observe that the variation of the energy required to charge the capacitor is smaller than the variation of the charging times, meaning that the efficiency reduction is negligible at smaller stimulus intensities. 5. Conclusions Noticeably, this system has the capability to harvest energy from the natural vibrations occurring during regular operation of farm implements. This feature overcomes the restriction of traditional wireless solutions due to limited battery lifetime, allowing the system to autonomously operate when the implement is on duty. Thus, the energy harvester module we realized allows powering an innovative system that enhances hazardous operation prevention, combining greenness and safety for agricultural machinery applications. Acknowledgements The paper is Supported by Key Program of Agriculture Research of Science and Technology Department of Fujian (No. 2012N0023), Foundation of Educational Research of Young Teachers Fu'jian Educational Committee. (JA. 13293), Foundation of Educational Research of Young Teachers Fu'jian Educational Committee. (JB. 13183), Foundation of Teaching Reform of Higher Education of SanMing University. (J1309/Q). Use the Acknowledgements section if it is necessary. References [1] . Sarghini F., D'Urso G., An Early Warning Device for Identification of Tractor Accidents, Rapid Alert and Assistance, in Proceedings of the International Conference on Work Safety and Risk Prevention in Agro-food and Forest Systems (SHWA'10), Ragusa - Italy, September 16-18,2010, pp. 494-500. [2] . Associazione Sostenitorie Amici della Polizia Stradale, http://www.asaps.it [3] . Myers J. R., Hendricks, K. J., Agricultural Tractor Overturn Deaths: Assessment of Trends and Risk Factors, American Journal of Industrial Medicine, 2010, pp. 1-11. [4] . National Agricultural Statistics Service, 2006 farm and ranch safety survey, Department of Agriculture, National Agricultural Statistics Service, Washington, DC: U. S., Report No. Sp Cr 3-1 (1-08), 2010. [5] . Hamrita T. K., Tollner E. W., Schafer R. L., Towards a robotic farming vision: advances in sensors and controllers for agricultural system applications, in Proceedings of the Industry Applications Conference, Thirty-First IAS Annual Meeting, (IAS'96), 1996, Vol. 3, 3 pp. 1678-1686. [6] . COBO Group, COBO Spa, http://www.cobospa.it [7] . Freescale MC1322x Platform in a Package, http://www.freescale.com [8] . J. F. Dickson, On-chip high-voltage generation in MNOS integrated circuits using an improved voltage multiplier technique, IEEE Journal of Solid-State Circuits, Vol. 11, June 1976, pp. 374-378. [9] . Piezo Bending Generator, Piezo Systems Inc., http://www.piezo.com/ [10] . A. Kansal, J. Hsu, S. Zahedi, M. B. Srivastava, Power management in energy harvesting sensor networks, ACM Transaction on Embedded Computing Systems, TECS, Vol. 6, No. 4, September 2007, pp.1-38. [11] . Data Physics, Electrodynamic Air Cooled Shakers, 2 lbf (9 N) to 12000 lbf (53.4 kN), http://www.dataphysics.com/ [12] . Zhen Wang, Sanyang Liu, Xiangyu Kong, Artificial Bee Colony Algorithm for Portfolio Optimization Problems, IJACT: International Journal of Advancements in Computing Technology, Vol. 4, No. 4, 2012, pp. 8-16. [13] . Gang-Ling Zhao, Li-Qun Chen, Jing-Li Fu, First Integrals of Discrete Birkhoffian System, IJACT: International Journal of Advancements in Computing Technology, Vol. 4, No. 4, 2012, pp. 17-23. 1 Liu Jian Jun School of Mechanical & Electronic Eng., Sanming University, Jing Dong RD 25#, San Yuan District, San Ming City, Fujian Province, 365004, China Tel: 15280567367, fax: 0598-8397252 E-mail: [email protected] Received: 26 May 2014 /Accepted: 29 August 2014 /Published: 30 September 2014 (c) 2014 IFSA Publishing, S.L. |