Harish Ramamurthy, B. S. Prabhu and Rajit Gadh
Wireless Internet for the Mobile Enterprise Consortium
University of California, Los Angeles
Los Angeles, California, USA.
Asad M. Madni
BEI Technologies, Inc.
Sylmar, California, USA
Abstract— a wireless smart sensor platform (based on patent
pending technologies “A Generic Wireless Transducer
Interface” [1] and “Application of generic reconfigurable
wireless interface for industrial automation scenarios” [2])
targeted for instrumentation and predictive maintenance
systems is presented. The generic smart sensor platform, with
‘plug-and-play’ capability, supports hardware interface,
payload and communications needs of multiple inertial and
position sensors and actuators, using a RF link (Wi-Fi,
Bluetooth, or RFID) for communications, in a point-to-point
topology. The design also provides means to update operating
and monitoring parameters as well as sensor/RF link specific
firmware modules ‘over-the-air’. Sample implementations for
industrial applications and system performance are discussed.
I. INTRODUCTION
Intelligent wireless sensor-based controls [2] have drawn
industry attention on account of reduced costs, better power
management, ease in maintenance, and effortless deployment
in remote and hard-to-reach areas. They have been
successfully deployed in many industrial applications such as
maintenance, monitoring, control, security, etc [3]. In this
research, the focus is to address the issues faced by
instrumentation systems and predictive maintenance
industrial applications and to design a solution to cater to the
issues faced by these applications.
Instrumentation systems are open/closed loop control
systems like motor control. They are formed using sensors
and actuators and the objective is to control certain
parameters, or state of the system. All the system elements
are always in communication with each other, typically,
requiring real-time performance. They also require in-built
fault-tolerance for communication/node failure – to return to
a safe-state in a deterministic amount of time.
Predictive-maintenance involves tracking physical state
of equipment or machine, and to take action, if an acceptable
or allowed state(s) is violated. Predictive-maintenance
applications are not active all the time in order to conserve
energy. The sensors are either periodic or event-based; they
wake up, check status and go back to sleep. In case of any
violation, they raise an alarm or record the digression. They
are very useful in keeping machine down-times low and help
locate the problem before the machine breaks down.
Both these systems employ different types of sensors
(e.g., position, accelerometers, gyros, etc.) and actuators
(e.g., motors) often deployed within the same network,
having different capabilities, interfaces, and supporting
different protocols for data and communications. Formation
of systems from such diverse distributed sensor elements
entails versatile control modules, which understand different
sensor protocols and utilize them. In addition, the operational
challenges are exacerbated when different RF links have to
be used to satisfy the requirements of bandwidth, payload,
delay, jitter, range, noise immunity and others (including
cost) for communication.
The proposed Smart Sensor Platform is an attempt to
develop a generic platform with ‘plug-and-play’ capability to
support hardware interface, payload and communications
needs of multiple inertial and position sensors, and
actuators/motors used in instrumentation systems and
predictive maintenance applications. Communication is
carried out using a RF link (Wi-Fi, Bluetooth, Mote or
RFID), in a point-to-point topology. The design also
provides means to update operating, monitoring parameters
and thresholds as well as sensor and RF link specific
firmware modules ‘over-the-air’. It is composed of two
main components – a sensor-wireless hardware interface and
system integration framework, which facilitates the defining
of interaction between sensors/actuators based on process
needs. The intelligence necessary to process the sensor
signals, monitor the functions against defined operational
templates, and enable swapping of sensor and RF link
resides on the microcontroller of the hardware interface. A
variety of industrial motion sensors like gyro, Inertial
Measurement Unit (IMU), linear position, absolute and
incremental encoders and actuators like motor, have been
interfaced and successfully tested with the platform.
The organization of this paper is as follows. Section 2
covers related work on sensor networks, and specific
initiatives for industrial automation. Section 3 describes the
intelligent wireless sensor architecture. Implementation
details, snapshots, simulation and experimental results of the
current implementation are presented in Sections 4 and 5.
Finally Section 6 reports the conclusions on the research.
II. RELATED WORK
The field of automation has continuously evolvedstarting
from early days of register level programming for
data acquisition and point-to-point wired links for
communication, to the current virtual instruments and
Ethernet, a wired communication paradigm for networking
industrial systems. Developmental efforts in this area can be
broadly classified as:
A. Industrial Initiatives
Includes the design of industrial open protocols for wired
communication also known as field buses like CAN,
DeviceNet and ControlNet; proprietary system formation
tools - Virtual Instruments from National Instruments,
Factory solutions from ABB, etc. Further development
involved open data exchange or messaging framework, for
e.g. OPC foundation which is trying to establish a standard
data exchange standard so that interoperability among
products (hardware and software) from different
manufacturers is achieved [4, 5]. Strong potential for
wireless is envisaged in enterprise-wide asset monitoring and
maintenance on an open protocol for communication like
ZigBee. Industrial initiatives have focused on the system
formation issues, but have been unable to exploit the
advantages of wireless technology.
B. Academic Initiatives
Though wireless sensor networks have become mature
[6, 7, 8, and 10], the focus has been on environment
monitoring, military and homeland security applications.
Though viability studies have been conducted for using
wireless communication in these industrial applications, not
much impetus has been given over full system deployment.
Application-specific wireless implementations have been
proposed but a generic system building approach has not
been investigated.
Deployment of wireless infrastructure in industries will
occur incrementally and interoperability (between different
systems) and extendibility (different application needs) will
form the requirements of prospective solutions. The smart
sensor platform research initiative is an attempt to develop
such an end-to-end solution with support for incremental
deployment, extendibility and scalability.
III. DESIGN
The motive of the smart sensor project is to create 1) a
general purpose hardware interface for diverse sensors and
actuators, which can be customized for any application
through over-the-air firmware downloads and 2) create a data
processing infrastructure at the backend to implement
applications. The proposed solution consists of a network of
sensors, and actuators communicating with the central
control unit using standard RF-links. The basic scenario is
shown in Figure 1. The sensors are directly connected to the
central control unit (workstation here) through a RF link,
which can be Bluetooth or WiFi.
Each sensor or actuator is equipped with a reconfigurable
generic wireless interface or smart sensor interface. The
interface extracts data from the sensors and commands the
actuator and provides a data communication interface to
the central control unit. A sensor/actuator
Figure 1. General Application Scenario
coupled with smart sensor interface is termed as a smart
sensor node.
A. Smart Sensor Node: Hardware Design
The sensors/actuators found in industrial applications can
be classified by analog, digital, serial (or combination of
these) signals used for data communication. The smart
sensor interface interprets sensors/actuators’ signals, and
converts it into digital data/commands. For this 14-bit
200ksps ADC, 8 channel 10-bit 9.6ksps ADC, DAC, 16
GPIO, and USARTs are used. The hardware design is
shown in Figure 2.
B. Smart Sensor Node: Software Design
The digital data extracted by the hardware interface has
to be bound by a context and processed to convert it into
useful information. This intelligence is provided by the
software that resides on the smart sensor interface.
The software design of the smart sensor interface is
shown in Figure 2. The software module stack on the smart
sensor interface consists of three layers. The bottom layer is
the device driver which directly interfaces with the hardware
interface and extracts digital data. The device manager
interfaces with the device drivers and exposes a multipledata
channel interface to the firmware layer. In the software
framework, each sensor/actuator is composed of a
combination of digital, analog or serial channels.
Establishment of context to the extracted channel data is
done at the firmware layer. The firmware layer “composes”
the sensor by combining data from multiple data channels. It
also implements the application specific functionalities like
real-time performance, data communication protocol with
central control unit, smart sensor node management, etc.
This separation of data acquisition tasks across three layers
in the smart sensor interfaces helps support functionalities
like over-the-air update of parameters, plug-n-play of
sensors, multiple sensor support, multiple wireless
technology support, universal data interface etc.
C. Application Integration Software
The application integration software resides on the
central control unit and handles application-specific
customization of the smart sensor nodes. Based on the Java
Beans framework, the software enables formation of systems
from discrete smart sensor nodes. Specific description for
real-time and predictive maintenance applications is
provided in the implementations section.
IV. IMPLEMENTATION
A. Real-Time Control
The objective of this implementation is to demonstrate
the non-deterministic real-time performance of the smart
sensor node. Deterministic real-time performance cannot be
achieved with the smart sensor node as wireless
communication is used, which is prone to errors. In order to
achieve near real-time performance the smart sensor node
tracks the traffic of wireless channels and uses a simple TCPlike
congestion control scheme to regulate the traffic. Once
the node senses congestion, high traffic, or connection loss, it
brings the node into a “safe-state”. The node then simply
waits for the central control unit to reconnect or signal
degradation to abate.
The system built for demonstration was a proportional
gyro-motor-encoder system (Figure 3). In this proportional
gyro-motor-encoder system, each sensor/actuator pair is
connected to a smart sensor interface and uses Bluetooth to
communicate with the central control unit. The gyro senses
the angular tilt and communicates it to the central control
unit, which in turn sends appropriate command to the motor.
Further, the encoder attached to the motor tracks the position
of the motor. In this application the safe state of the system is
to bring the motor to a halt.
B. Predictive Maintenance:
A typical factory environment is considered, where the
health of the machinery/equipment is regularly monitored
and any digressions/violations from the tolerable behavior
during operation are recorded. The recorded information of a
machine typically consists of information like threshold
violations, time of the event, extent of the event, etc. The
status of machines is typically checked by a qualified
machinist who inspects the machine when the main power
has been switched off. Any proposed solution should thus
operate passively and data should be stored locally.
In the current implementation, smart sensor nodes,
equipped with sensors to monitor the status of a machine,
store the health information in a RFID tag. RFID tag is used
as a plain wireless non-line-of-sight data storage [9]. In this
mode, the maintenance personnel can retrieve the required
health information by querying the tag even when the central
computer has been switched off.
Hardware Design Software Design
Figure 2. Smart Sensor Node Design
Figure 3. Implementation Snapshots
To demonstrate we consider as an example, an
application scenario where every threshold violation of the
linear sensor has to be recorded with timestamp. The
threshold parameters are set through the application software
module during deployment. We use the ISO15693 (13.56
MHz) tags for storing data. These tags have memory ranging
from 256 bytes to 2KB. A handheld reader connected to a
PDA is used to read tag data. On the PDA, the records are
presented in a tabular format. Snapshots of the current
implementation are shown in Figure 3.
V. EXPERIMENTAL RESULTS
Experiments were conducted to study the relevant
performance metrics such as link delays, bandwidth with
varying distance, traffic and packet bursts. For lack of space
we only provide the delay experiments here. For detailed
description of experiments please refer [11].
Round trip delay is an important characteristic of a
control system. It places an upper limit on the responsiveness
of a system. Delay performance of different wireless
technologies was tested and is presented below (refer Fig 4).
To summarize, the effects of the following parameters on
delay of Bluetooth are (refer Figure 4):
• Distance: With increasing distance, the delays
become larger and jittery.
• Traffic: No considerable effect
• Packet Bursts: Mild effect with performance
degrading with more packets per burst
For Wi-Fi
• Distance: Performance degrades with distance, delay
increases and becomes jittery.
• Traffic: Performance worsens with increasing traffic.
The effect is more pronounced at larger distances.
• Packet Bursts: As time to access channel is constant,
bigger payloads experience less per-byte delays
Thus, Bluetooth seems to fit better in industrial
application scenarios where limited bursts of data need to be
delivered in real-time in a noisy environment. Wi-Fi seems
to fit better in scenarios where huge amount of data need to
be transmitted in a less noisy environment.
VI. CONCLUSIONS
A wireless smart sensor platform targeted for
instrumentation systems and predictive maintenance was
presented. Sample implementations for instrumentation
systems and predictive maintenance applications were
discussed and presented. Tests were carried out to determine
system performance and were presented. The experimental
results show that a sustained near-real-time system can be set
up with the smart sensor nodes.
REFERENCES
[1] H Ramamurthy, XY Su, BS Prabhu and R Gadh, “A generic wireless
transducer interface”, US Patent Disclosure, UCLA Case No. 2005-
245-1 (Pending).
[2] H Ramamurthy, B. S. Prabhu and R Gadh, “Application of generic
reconfigurbale wireless interface for industrial automation scenarios”,
US Patent Disclosure, UCLA Case No. 2006-002-1 (Pending).
[3] U.S DOE, “Industrial Wireless Technology for the 21st Century” –
Report, Technology Foresight, Winter 2004/ TF-2004-1.
[4] Ryan Wynn, “Plug-and-play sensors”, Machine Design,
75(10), May 2003, pp. S10.
[5] OPC HDA Specifications, Version 1.20.1.00 edition, 2003, Available
for download at: www.opcfoundation.org.
[6] IEEE Standard for a Smart Transducer Interface for Sensors
andActuators- IEEE 1451.3 edition, 2004.
[7] J. M. Hellerstein, W. Hong, S. Madden, and K. Stanek, “Beyond
Average: Towards Sophisticated Sensing with Queries,” in
International Workshop on Information Processing in Sensor
Networks, March 2003.
[8] Y. Yao and J. E. Gehrke, “Query Processing in Sensor Networks,” in
First Biennial Conference on Innovative Data Systems Research
(CIDR), January 2003.
[9] B. S. Prabhu, Xiaoyong Su, Harish Ramamurthy, Chi-Cheng Chu,
and Rajit Gadh, “WinRFID – A Middleware for the enablement of
Radio Frequency Identification (RFID) based Applications”, Invited
chapter in Mobile, Wireless and Sensor Networks: Technology,
Applications and Future Directions, Rajeev Shorey, Chan Mun
Choon, Ooi Wei Tsang, A. Ananda (eds.), John Wiley, December
2005.
[10] W. Heinzelman, A. P. Chandrakasan, and H. Balakrishnan, “An
Application-Specific Protocol Architecture for Wireless Microsensor
Networks,” IEEE Transactions on Wireless Communications, vol. 1,
pp. 660–670, October 2002.
[11] Harish Ramamurthy, B. S. Prabhu, Rajit Gadh, “Experimental studies
on Bluetooth and Wi-Fi radio links for real-time control”, Tech
Report, WINMEC-REWINS-006, Dec 2004.
Round Trip Delay with Heavy Traffic at 20ft
0
50
100
150
200
250
300
350
1 11 21 31 41 51
Iterations
Round Trip Delay (ms)
Wifi
Bluetooth
Round Trip Delay with Heavy Traffic at 5ft
0
10
20
30
40
50
60
1 11 21 31 41
Iterations
Round Trip Delay (ms)
Wifi
Bluetooth
Figure 4. Experimental Results: Round-Trip Delay
ليست هناك تعليقات:
إرسال تعليق