Machinery Condition Monitoring Using Wireless Self ... - CiteSeerX

Machinery Condition Monitoring Using Wireless Self-Powered Sensor Nodes Fred M. Discenzo, Ph.D. Kenneth A. Loparo Harry Cassar Dukki Chung, Ph.D. Rockwell Automation, Case Western Reserve BP p.l.c. Rockwell Automation, Inc. University 1 St. James’ Square Inc. 1 Allen-Bradley Dr 10900 Euclid Ave. London, SW1Y 4PD 1 Allen-Bradley Dr Mayfield Hts, OH Cleveland, OH 90807 United Kingdom Mayfield Hts, OH 44124 44124 ABSTRACT There is rapid growth in the development of ultra-low power processors, radios, memory, and architectures that form the basis for wireless sensor nodes. Smart distributed wireless sensor nodes, originally developed for military applications over 10 years ago, are finding increased use for wireless machinery condition monitoring for industrial and commercial applications. The major drawback to low-cost wireless sensor nodes is the need to regularly replace batteries. This paper describes a prototype self-powered wireless sensor system deployed in a shipboard application. The sensor node scavenges energy from machinery vibration and uses this energy to power an embedded processor, sensors, and radio. Vibration data and energy harvesting efficiency data is periodically transmitted to a data collection server on the ship. Important issues are presented including the efficient design of the energy harvesting module, power conversion and power storage, communications, power management, embedded diagnostic algorithms, site survey and wireless sensor node installation, and the need for a systems-level approach for self-powered wireless sensor nodes. Extensions to this core work are presented that include adaptive power scavenging and self-powered sensor nodes embedded in machinery. Introduction There is a growing proliferation of wireless devices for many industrial and commercial applications such as for security, safety, surveillance, and machinery condition monitoring. Although the wiring costs for communications and power cables are eliminated with wireless devices, these costs are replaced with the ongoing cost of battery maintenance (e.g. battery cost, manpower, logistics, environmentally conscious disposal, protection from potential leakage, and potential equipment downtime due to unexpected battery failure). A critical technology is the ability to scavenge energy from the environment and use this energy to power remote, distributed sensors, radios, actuators, processors, memories, and display elements. This technology is enabling and permits power-scavenging sensor nodes to never require servicing and to be deployed in inaccessible locations or embedded into machinery. Rockwell Automation, in collaboration with BP’s Chief Technology Office in the U.K., has developed and deployed two self-powered wireless sensor nodes on the BP tanker Loch Rannoch. These sensor nodes were deployed as part of a technology evaluation program to establish the viability of wireless sensor nodes operating in a harsh shipboard environment for machinery condition monitoring. The shipboard trial of self-powered sensor nodes was done in parallel with a larger scope test of wireless sensor nodes on the tanker. Background There are important technology changes occurring that promise to change the character of machinery monitoring. These changes will affect future machinery condition monitoring, safety, control, re-configuration, and security. New developments in diagnostic and prognostics algorithms, emerging CBM architectures and standards (e.g. OSA-CBM, MIMOSA Open O&M, and OPC/ISA), and new, low-power wireless communications standards and components (e.g. Bluetooth and IEEE 802.15.4) enable the deployment of many low-cost distributed sensors. An array of distributed sensors can provide superior capabilities for mission-critical applications and can directly reduce maintenance cost and total life-cycle cost. Wireless sensor nodes often employ energy efficient processors, memory, radios, and energy-management logic. These systems have been under development for over 10 years and are targeted for applications such as surveillance, target acquisition, and machinery monitoring

[1][2]. In spite of their energy efficiency, the need for a reliable, long-term energy source remains a roadblock to the broadscale deployment of thousands of distributed sensor nodes. For many distributed sensor applications it is not practical to provide wire-line power due to factors such as cost, weight, reliability, safety, or environmental hazards (e.g. explosive environments). For example, in many industrial applications the cost of wiring may be $40 USD per foot or more and often exceeds the cost of the remote sensor. The need for costly wiring can be eliminated if the distributed sensor node is self-contained and is self-powered. Options for self-powering sensor nodes include batteries, micro-fuel cells, micro-generators, and power-scavenging technologies. A self-powered sensor node utilizing micro-generators or fuel cells provide limited benefit and is currently not considered a viable solution. Utilizing storage batteries for remote sensors is also not considered a generally effective solution since the batteries required to power each sensor adds significant cost, weight, reliability and maintenance problems, particularly when the sensors are located in difficult to reach areas. These problems are multiplied when hundreds of thousands of battery-powered sensor nodes are deployed. Furthermore, batteries have an uncertain life and it is difficult to accurately predict the battery charge depletion profile. Batteries encapsulate caustic chemicals and heavy metals that may, in time, leak causing damage to equipment, injury to workers, and may affect other nearby equipment and the environment. Batteries also suffer from low power density and rapid deterioration and aging especially in hostile environments. For these reasons batteries are not considered a generally viable method to power remote wireless sensor nodes. Batteries may be effective for targeted applications requiring low duty cycle and low power requirements in readily accessible locations. Harvesting power from stray energy for remote sensors is well-suited for CBM sensor applications. CBM systems often require periodic sampling from sensors which may be distributed across a machine, vessel, or facility. Remote sensor processing can typically be performed periodically and at a frequency consistent with the rate of local power generation. Newer materials such as certain piezo-electric materials exhibit high coupling efficiencies and can operate at elevated temperatures (>550°F). These materials are already being used for vibration and ultrasonic sensing and are well-suited for power scavenging devices for CBM applications. A remote self-powered sensing device may be composed of four core elements not including the machine or communications (Figure 1). The first element is a sensing element such as temperature, pressure, vibration, or magnetic field. The second element is a parasitic energy transducer capable of transforming stray energy into an electrical potential. The third element is a power storage device such as a rechargeable battery or a capacitor bank. And the fourth element is a processor and algorithms for interpreting the sensor data. There are often additional elements such as local memory, digital output, or communications such as a wireless data link. The communications link may not be required if the self-powered sensor node is used to locally store information and operate as a “black-box” system or if only a local display such as blinking LED is used to signal a machine fault.

Figure 1 – Components of a Self-powered Sensor Node A more complete configuration may include multiple power storage elements with different electrical characteristics, multiple sensor elements, multiple energy harvesting modules with different electrical and mechanical characteristics, actuator elements, and adaptive hardware and software elements to optimize system performance.

The core elements described above can provide valuable machinery condition monitoring capabilities. The ability to perform machinery condition monitoring without the cost of wiring and the burden of periodic battery replacement provides attractive benefits for many industrial, commercial, and marine monitoring applications. Established power scavenging techniques such as piezo-electric devices and photovoltaic cells can provide the power needed for machinery condition monitoring for reasonable environmental conditions and provide periodic machinery monitoring. Regular machinery monitoring is particularly valuable for shipboard systems. Routing power and signal cables on ships is frequently very difficult due the presence of thick compartment walls, limited free space for cable trays and conduits, and watertight compartment requirements. Similarly, manually capturing data at machines below deck is time consuming for the ship’s crew and can be dangerous during high seas or in the presence of water, power cables, or petrochemical fluids or gases. A program was defined to evaluate the capability of existing energy harvesting technologies and to establish the deployment issues and operational benefits afforded by wireless, self-powered machinery condition monitoring. A set of preliminary specifications were developed and a shipboard target application was established. BP Shipping has supported this project by providing technical information and access to the FPSO tanker Loch Rannoch (Figure 2 BP Ship Loch Rannoch) [3][4]. Figure 2 BP Ship Loch Rannoch Shipboard Trial i. Objectives The objective of this program is to establish the viability of an energy harvesting system comprised of sensors, power generator, power conversion electronics, and power storage components. These components are assembled into a system to be evaluated in the context of machinery condition monitoring in a harsh environment (e.g. a ship machine room). ii. Device Design Several integrated self-powered sensor nodes were designed and constructed for this program. The self-powered sensor nodes continuously scavenge energy from the environment, convert and store the scavenged energy, and use the stored energy to periodically power sensors, operate analog to digital converters (ADC), operate a microprocessor, illuminate local LED’s, and perform radio communications. A picture of the self-power sensor node is shown in Figure 3 Self-powered Sensor Node. The connector shown at the top is used to temporarily provide DC power to initially charge the array of storage capacitors. After the capacitor bank is charged, the energy scavenged need only be enough to replace the depleted power from the operation of the sensor node.

Figure 3 Self-powered Sensor Node

There are seven core elements that comprise the self-powered sensor node. These elements are: 1) 2) 3) 4) 5) 6) 7)

Processor Radio Sensors Power Generator Power Storage Electronic power circuit Software

Two self-powered sensor nodes with embedded processor, radio, sensor, and generator have been constructed. Each device is programmed to sample three analog inputs. The first analog input is the sampled data from an accelerometer. The accelerometer is a single-axis sensor mounted on the inside of the sensor node enclosure. Vibration data is sampled for one second at 1 kHz. The second analog input is the voltage generated from the

piezo-electric generator. The third analog input is the state of charge of the capacitor bank. The generator voltage and the capacitor state of charge are sampled to provide an RMS or DC voltage level. Each of these three inputs are stored in the local processor memory and also transmitted to a third processor. The third module receives the data transmitted from the two self-powered sensor nodes and sends the data through the serial port to a PDA with a memory card. Software on the PDA displays the time waveform data received from the two remote sensor nodes on an LCD and also archives the data to a memory stick in the PDA. The configuration of the sensor nodes is shown in Figure 4 The objective is to scavenge enough energy from the environment to replace the power needed to support the machinery monitoring and data transmission functions. Previous work conducted by the authors and published results from others indicates that piezo-electric material is the most appropriate generator technology for this application. Studies have indicated that while photovoltaic cells may provide the greatest power-generation density, the most versatile, power-dense generators are based on piezo-electric materials[5][6][7].

Figure 4 Self-Powered Sensor Node System

Piezo-electric materials convert mechanical strain such as induced by machinery vibration to a voltage. When implemented as a cantilever design with one end fixed and the other end free to bend, a sinusoidal voltage is generated that corresponds to the harmonic motion of the beam. The power generated is a function of the amplitude and frequency of the motion of the piezo-electric beam. Maximum strain is obtained, and maximum power obtained when the resonant frequency of the piezo-electric beam matches the vibration frequency with the most energy (i.e. highest amplitude) on the target machine. An oil pump was identified as the candidate machine for monitoring since it operates continuously with a relatively high rotational speed. The target frequency where most of the vibrational energy exists for this oil pump is 7800 cps (130 Hz). A cantilever beam design was used for the piezo-electric generator element. The constraints for the generator is that it must fit within a three inch footprint and have a resonant frequency of roughly 130 cps. Figure 5 shows the piezo-electric generator developed for this program. The small threaded bolt at the narrow end of the generator can be extended or retracted as needed to provide fine tuning adjustment of the resonant frequency of the beam. Commercially available piezo-electric bi-morph material was used to construct the generator element1. This material consists of two layers of the piezo-electric material surrounding a brass metallic inner layer and a series bender connection is used. A narrow strip of composite material is mounted just above the timing screw to limit the Figure 5 deflection of the beam and prevent over-straining the piezo-electric Piezo-electric Generator element. Power conversion electronics were developed to rectify the sinusoidal voltage from the piezo-electric generator, scale this voltage for charging the super-capacitor bank, convert power from batteries for pre-charging the capacitor bank, provide a voltage signal for monitoring the generated power and the capacitor state of charge, and regulate the power going to the processor and radio. An array of nine super-capacitors was used to store the generated power. The super-capacitors provide excellent power density and negligible leakage current. The sensor node components described above are housed in a small plastic enclosure attached to a stiff mounting bracket. Figure 6 shows the arrangement of the components in the enclosure. The box has a removable cover secured with four screws to permit mounting devices on the inside of the box and on the box cover. The power conversion circuitry is assembled on one side of a small perf-board and the array of supercapacitors are mounted on the other side. The assembled perfboard with electronics and capacitors is mounted 1

Material used is T220-A4 from Piezo Systems Inc.

to the inside of the cover. The processor and radio are components in a commercially available device called a mote1. The mote is mounted to the side of the enclosure and is shown next to the cantilever generator. The piezo-electric generator is shown mounted to the bottom of the enclosure. The battery holder is mounted to the side of the enclosure and is shown with two lithium batteries installed in Figure 6. Capacitor The batteries are used to charge the pre-charge Batteries for capacitor bank prior to deploying the connector pre-charging system. It is not essential to pre-charge circuit the storage capacitors however starting with a charged capacitor bank eliminates Accelerometer the need to wait days or weeks for the generator to build up a sufficient stored Piezo-electric charge to power the processor, sensors, cantilever generator and radio. Battery power may be provided using the battery holder as Telos Mote shown in Figure 6 or provided by a DC supply through the capacitor pre-charge connector on the side of the case. Figure 6 Assembled Self-Powered Sensor Node iii. Installation & Operation There were many uncertainties in this experiment including the reliability and efficiency of the energy harvesting device during prolonged unattended operation at sea. Other uncertainties include not knowing the expected duty cycle or operating characteristics of the machinery providing our source of power. Even if the machine operates continuously it must operate with an appropriate level of vibration near our targeted, tuned frequency. Other shipboard uncertainties include the weather and ambient vibration caused by other shipboard machines, sea state, climate, and ship operation. Research has been done in each of these areas and it is expected that harsh weather conditions and high seas are likely during the sea trial [8]. Movement of ships crew, closing of compartment hatch doors, and radio interference are also considerations for this trial. Work was done prior to the installation of the self-powered sensor nodes to minimize these risks and insure a successful sea trial. Background work included laboratory testing of components and system level testing of the self-powered sensor nodes. A site survey was conducted aboard the ship prior to installing the equipment that included monitoring radio transmission performance at the planned radio frequencies while operating various shipboard machines such as motors and radar. Detailed documents were developed defining the procedures for the site survey, equipment testing, shipboard installation, and equipment decommissioning. To reduce the risk of failure due to a device fault or due to inadequate power generation capability, two identical self-powered wireless sensor nodes were constructed. One node was deployed using batteries plus power scavenging. This permits prolonged monitoring of power generation and energy utilization even if insufficient energy for operation is generated. The second self-powered sensor node was deployed without batteries and relies solely on power scavenging for continued operation. Both sensor nodes were installed on the same oil pump in the machine room. Each sensor node is programmed to hibernate for one hour. Each hour the sensor node wakes up, turns on the accelerometer and after an appropriate settling time samples vibration. The accelerometer is turned off and voltages are then sampled from the piezo-electric generator and from the capacitor bank. The radio is then cycled on. After a suitable delay to permit the radio to power up and synchronize the sampled data values are transmitted to a third mote connected to a PDA as shown in Figure 4 above. The self-powered sensor node then turns off the radio and returns to hibernate mode. The third mote receives the sampled data and send the values to a PDA through the serial port. The PDA displays the time waveform data from both sensor nodes and logs the sampled data to a memory stick. The third mote and PDA are powered from the ship’s power supply.

1

The Telos mote from Moteiv Corporation was used for this prototype system

Experimental Results The equipment was installed on the oil tanker and was left operating and unattended for four months from midAugust 2004 to mid-December 2004. Periodic readings provided by the ships crew indicated the system was continuing to operate. Following the sea trial all equipment was removed from the ship and the data files captured from the sensor nodes were copied to a server for analysis. Over 8,000 data files were captured. Most of the captured data looks valid however some corrupt file names and data values were observed. The cause for the data faults is currently being investigated. Vibration data was captured at 1 kHz with 12bit resolution. This information can be used to compute the input power to the energy harvesting device. This output voltage form the piezo-electric generator was also captured to permit computing the output power. The ratio of the input power to the output generated Watts power provides a measure of the conversion efficiency. One of the self-powered sensor 4.50E-02 nodes consistently generated very low power 4.00E-02 levels with the maximum power observed on 3.50E-02 the order of tenths of watts. The other self3.00E-02 2.50E-02 powered sensor node was able to reliably Watts 2.00E-02 generate hundreds of microwatts. Laboratory 1.50E-02 and analytical results suggest this device is 1.00E-02 capable of generating perhaps ten times this 5.00E-03 amount of power under the expected 0.00E+00 operating conditions. The lower level of power 1 241 481 721 961 1201 1441 1681 1921 2161 generation observed is likely due to the Time variation in rotational speed of the motor-pump system and lower amplitude of vibration. Figure 7 Figure 7 shows the power generated for the Plot of Average Power Versus Time for Mote 7 more efficient self-powered sensor node. The above diagrams suggests the oil purifier runs periodically. It is instructive to view the ratio of the power generated to the input power. The power generated is recorded at each sample interval by the mote as the rectified voltage from the piezoelectric generator. The input power to the device may be approximated by the vibration sensed from the accelerometer attached to the inside of the power scavenging enclosure. At each sample interval, acceleration data is recorded. The power conversion efficiency may be calculated as the ratio of output power to input power. The following equation is used to compute this ratio:

Equation 1: Power Ratio

1 50

50

∑ i =1

Vi2 2 x 34 . 8 k Ω

1 500

500

∑ i =1

Ai

The average power generated and the power conversion efficiency is shown for several representative data samples in Table I below.

Table 1 Average Power Generated and Ratio of Input Power to Generated Power for Mote 7 MOTE 7 1 50 Data Collection Data and Time 10_10_04__10_05_19_AM 10_10_04__10_47_05_PM 10_10_04__11_03_55_AM 10_10_04__11_45_41_PM 10_10_04__12_02_31_PM 10_10_04__12_19_21_AM 10_10_04__1_01_07_PM 10_10_04__1_17_56_AM 10_10_04__1_59_43_PM 10_10_04__2_16_32_AM . . . . 9_9_04__8_38_33_PM 9_9_04__8_55_22_AM 9_9_04__9_37_09_PM 9_9_04__9_53_58_AM

50

∑

i =1

1 50

Vi2 2 x 34 . 8 k Ω

Vi2

50

∑ 2 x 34 .8 kΩ i =1

1 500

500

∑A i =1

2 i

1 50

Vi2

50

∑ 2 x 34 .8 k Ω i =1

1 500

500

∑ i =1

Ai

Average Genreated Power (Watt) 2.61E-04 2.57E-04 2.61E-04 2.58E-04 2.60E-04 2.66E-04 2.58E-04 2.63E-04 2.57E-04 2.65E-04

Ratio 3.77E-05 3.56E-05 3.73E-05 3.51E-05 3.68E-05 3.90E-05 3.64E-05 3.84E-05 3.61E-05 3.94E-05

Ratio 9.96E-05 9.63E-05 9.91E-05 9.57E-05 9.83E-05 1.02E-04 9.77E-05 1.01E-04 9.70E-05 1.03E-04

4.24E-03 4.24E-03 4.24E-03 4.24E-03

1.04E-07 1.04E-07 1.04E-07 1.04E-07

2.94E-05 2.94E-05 2.94E-05 2.94E-05

The power ratio captured roughly every hour is shown in Figure 8. During operation, the machinery vibration is likely at an adequate amplitude level for the energy harvesting device but most likely at a frequency distant from the resonant frequency of the energy harvesting cantilever beam. The cantilever beam was tuned for a nominal rotational frequency of 7800 cps or 130 Hz. The data plots obtained from the onboard machinery database indicate a frequency of 7968 cps. This results in a stimulus relatively far from the 7800 cps target resonant frequency of the cantilever beam. This will result in reduced amplitude of vibration of the cantilever beam and a corresponding significant reduction in the amount of energy generated.

Figure 8 Power Conversion Ratio for Mote 7

Conclusions The functional elements comprising the self-powered wireless sensor nodes are tightly coupled. Successful integration of the component elements is crucial to successful operation of the energy harvesting system. Furthermore, the design of the integrated energy harvesting module must be compatible with the expected operating environment. The generator design employed was based on extracting maximum energy from the relatively low levels of vibration observed. This caused to device to provide reduced power levels when straying from the targeted resonant frequency. Efforts are currently in progress to dynamically change the frequency response of the energy harvesting device in response to changes in the environment and operating equipment energy spectrum. A systems approach is essential for defining the requirements of each of the system elements and how they integrate into an operational system. A systems approach will insure that the needed power is stored and used in an efficient manner to accomplish the sensor node functions. The power budget includes requirements for sensors, radio, processing duration and frequency of operation, power conversion efficiencies and power losses and leakage.

The following summarizes the conclusions from this study. 1. There is a need for adaptive self-powered systems, 2 Energy Harvesting is a viable technology for wireless sensor nodes – the benefits are significant for applications such as shipboard machinery monitoring 3. It is important to establish hardened devices that operate reliably, and require minimum set-up and installation effort. The remote operation of sensor nodes in particularly harsh ship engine rooms requires an additional level of integrated design, testing, and validation before deployment. 4. Data analysis requires more than sampled data from the sensor node. It is important to interpret sampled data in context with ship operation, other machinery operation, sea state, and with historical data. 5. Collaborative development and in-field technology evaluation can accelerate development – the complexity of highly distributed, remote technology development and deployment has been addressed by the significant up-front laboratory testing, data analysis, documented field procedures, site surveys, and trained staff members. The shipboard trial has demonstrated the significant benefits provided by wireless sensor nodes for shipboard machinery monitoring. The potential benefits that include increased machinery reliability, reduced maintenance cost, reduced maintenance effort, and enhanced safety have been recognized by a recent trade publication award [9]. These benefits are further expanded by implementing self-powered systems designed to never require maintenance. This is an enabling technology that not only improves the economics of deploying sensor nodes but also permits devices to be located in inaccessible locations and embedded inside un-powered rotating machinery. The fundamental technologies demonstrated in this limited scope shipboard trial promise to change the way machinery will be monitored in the future. Acknowledgement The BP CTO organization is to be commended for providing the vision, the leadership, and the commitment to this innovative program. Special recognition goes to BP Shipping including the captain and crew of the Loch Rannoch for supporting this important field evaluation program. Rockwell Automation Marine Management and GMS staff under the leadership of Mr. John Zubak provided invaluable support to this program through weekly communications and installation services provided throughout the sea trials. References [1] "Wireless Integrated Microsensors", presentation given at The Seventh IEEE Solid State Sensor and Actuator Workshop on June 2-6, 1996 at Hilton Head, South Carolina. [2] Discenzo, F.M., Loparo, K.A., Cheng, D., Twarowski, A., “Intelligent Sensor Nodes Enable a New Generation of Machinery Diagnostics and Prognostics”, Machinery Failure Prevention Technologies, MFPT 2001, Virginia Beach, VA, 2001. [3] “In Dust We Trust”, The Economist, Technology Quarterly, Vol.371, No.8379, pp.10-12 [4] “Mighty Motes”, Plant Services, November 2004 [5] Ghandi, K., “Compact Piezoelectric Based Power Generation”, Power Point presentation by Continuum Control Corporation to DARPA/USAAMC, 13-14 April 2000, Billerica, MA, USA, www.darpa.mil/dso/trans/energy/briefings/4Ghandi.pdf. [6] Roundy, S. J., “Energy Scavenging for Wireless Sensor Nodes with a Focus on Vibration to Electricity Conversion”, Ph.D. Thesis, University of California at Berkeley, Berkeley CA, USA, May 2003, http://engnet.anu.edu.au/DEpeople/Shad.Roundy/paper/ShadThesis.pdf [7] Roundy, S. J., ““Energy Scavenging for Wireless Sensor Nodes with a Focus on Vibration to Electricity Conversion”, Power Point presentation, University of California at Berkeley, Berkeley CA, USA, 19 February 2003, http://engnet.anu.edu.au/DEpeople/Shad.Roundy/EnergyScavenging.pdf. [8] Economic Development Unit, Shetland Islands Council, “Shetland in Statistics”, Shetland Litho, ISBN 0 904562 40 9, 2003, Green Head, Lerwick, Shetland, ZE1 0P, UK, pp. 7, http://www.shetland.gov.uk/council/documents/18170-Shet-in-Statistics.pdf [9] “100 Best IT projects in 2004”, Infoworld, November 2004