To begin we’d like to give some background on the state of the problem. Firstly, fire smoke and cigarette smoke are much different. Particles from cigarette smoke are much smaller and have a much smaller particle concentration. Second, detectors used commercially are designed for fires, and are therefore insensitive to smoke from cigarettes.
We were approached with a problem from hotel management. Non-smoking rooms were being damaged because of use of cigarettes and possibly other substances within the designated no smoking facilities. This causes future guests discomfort and possible health issues. It also causes hotel repair costs and their reputation to be affected. The customer needs a way to help enforce no smoking policies as well as catch those who break it before damage has been done. Specify as customer problem statement
This is a block diagram of the cigarette monitoring system, it consists of smoke detection units installed in hotel rooms and a computer at the hotel’s front desk which monitors the rooms. The smoke detection unit will contain a pair of sensors, a microprocessor, and a transmitter receiver for wireless communication with monitoring station. We plan on building 1 smoke detection unit, 1 monitoring station, and 4-5 relay nodes. The relay nodes will be capable of wireless communication but not cigarette detection. They will be used to test the network functionality more rigorously.
By April, we plan to be able to deliver a bundle of our work. This will include documentation of our preliminary research and design ideas, our working smoke sensor, descriptions of our testing, and resulting information to show our model works under sponsor/user specifications.
Here are the requirements that were given to us by our customer along with the methods that we will use to validate that we met that requirement. First is a cost requirement for the system. Each sensor node in each room is required to be less than $50. In the lobby, the monitoring station is to be less than $1500. Next our customer required us to design a sensor that is less than twice the size of a commercial smoke sensor. This value comes out to be about 8.5x8.5x3 inches. We are required to develop a system with greater sensitivity over commercial smoke detectors. The network capacity for our project is required to support at least 100 rooms. We will insure that multiple devices can communicate with the lobby’s monitoring station. Our project will have to be functional under US regulations so it can be used in the US. Therefore, we will abide by the standards and regulations already established by the government.
To determine the effectiveness of our system, we will set up a variety of controlled tests. These tests will determine the capability of our system in terms of smoke detection. We want to limit the amount of false positives to less than 5 per day per room. Again, we will set up controlled test for verification.A latency of 20 seconds is the max latency for the network, which is the time it takes a sensor to communicate with the lobby’s monitoring station. To test this we will establish tests to measure the time it takes for a message to be sent from the sensor. The detection latency is the time it takes the sensor to send an alarm message after a cigarette has been lit. This is required to be less than 5 minutes. For validation, we will set up controlled tests in rooms that have similar characteristics as hotel rooms to determine how long it takes for an alarm to be set off.
So what exactly is cigarette smoke? As you can see, it consists of three different groups: gases, chemicals, and particulates. For our purposes, we are attempting to sense the presence of particulates, as they are the most prevalent and detectable in the smoke.
Here is a breakdown of our project. We realized that there were 3 major components to the system, each of which are more or less mutually exclusive from each other, so one component is not necessarily dependent on another component until the final product is to be assembled. The first component is cigarette detection, which includes both smoke detection and flame detection, both of which we will go into greater detail soon. The second component is the wireless network, which includes how the sensor nodes in the rooms are to communicate with the monitoring station at the front desk. The last component is the monitoring station, which consists of the device at the front desk of the hotel that will signal the managers and workers of a smoking alarm.
We will now compare the performance of existing smoke detection techniques. On the left most column of the chart are the two most common methods of smoke detection, Ionization and Photoelectric, and Laser which is a special case of photoelectric. We will be proposing to use a laser type detector in our design. The center column shows their sensitivity in regards to obscuration level. Obscuration measures smoke concentration, so the higher the obscuration, the more smoke there is. So in general, Ionization detectors are better than photoelectric when small amounts of smoke are present. However, several companies have implemented photoelectric sensors using lasers and these have been very sensitive to small amounts of smoke. The next column shows the range of particle sizes which the detectors are most reactive to. The range for ionization detectors fit well for the average size of cigarette smoke particles while the lower range of photoelectric detectors is very near cigarette particles. Unfortunately we could not find data on Laser type detectors but we hope to show that the range for our proposed design will be lower than photoelectric. Overall, our design aims to optimize the laser technique to increase sensitivity in terms of detecting small amounts of smoke as well as the size of smoke particles. Minimizing false positives will also be important to our design.Ionization and Photoelectric are the most common fire smoke detectorsObscuration: measures concentration of smoke (higher obscuration more smoke)These are the size of the particles which the detection method is most sensitive toLaser is a special case of photoelectricwe are going to choose laserLaser has smaller particle range than photoelectricOptimize for lowest obscuration possible and particle size of cigarette smoke
Images are of a basic ionization chamber. Radioactive material, in this case Americium, is used to create alpha particles. These particles react with the oxygen and other gases in the air to create ions. Two charged plates are closely separated. The charged ions create a current between the two plates, on the order of picoamps. The introduction of smoke particles de-ionizes the air particles and causes the current seen between the plates to decrease. By monitoring the change in current, the sensor is able to detect particles in the air.
Dual Chamber sensor: helps to eliminate effects from humidity and atmospheric pressure on the sensor’s readings. One chamber is the sensing chamber (open to outside world – affected by particles in the air, humidity, and atmospheric pressure), while the other chamber is the reference chamber (partially closed to the outside world – affected by humidity and atmospheric pressure). Electronics measure difference between two chambers. Any change will be only due to particulate matter and not humidity or atmospheric pressureLarger voltage on plates: maximize current -> easier to detect differenceMinimize distance between pates -> maximize current
Mention that because of these reasons we will not be doing IonizationMeasuringpicoamps is difficult to do accurately. Amplifier circuits may be used to step up the current, but it would also step up any noise induced in the system, such as EMI from the network radio. Gathering and using radioactive material is not only dangerous, but also difficult to get proper certification. The plates need to close together, causing difficulties in the construction of the chamber.
This shows the principle of operation of a photoelectric type detector. It typically consists of an infrared LED and a photosensor which are placed off axis. In the absence of smoke, the light from the LED will not reach the photosensor. When smoke particles enter the path of the light, they will randomly scatter the light and the photosensor detects this scattered light.The light source and photosensor are off-axis, without smoke no light reaches the photosensorWhen smoke enters the light’s path, light will scatter randomly and some light reaches the photosensor
Now we’ll discuss how we will improve on the existing technique. In order to increase sensitivity we will use a laser diode rather than an LED. Laser diodes are generally thousands of times brighter than LED’s. So, by increasing the amount of light we emit, more light will be available to be scattered. In addition, while fire smoke detectors use infrared LED’s with a wavelength of 850 nm, we plan on using a red laser with a wavelength of 635 nm. Recall that cigarette particles are 100-300 nm on average. By using a wavelength closer to this range, more light will be scattered. We also plan to use an array of 4-5 photosensors rather than one so that the sensor can capture more of the scattered light. However, the ambient light within the room could be sensed by the photosensors. So, the housing of the whole unit will need to minimize ambient light reaching the photosensors in order to avoid false positives.Laser Diode – commercial detectors use IR (850 nm) LED’s ; mention pinnacle’s claim of making a LASER detector better than ionization - laser = more light to be scattered (even small amount of smoke results in significant scattered light) - lower wavelength = more scattered light - 635, 650 nmPhoto-sensor array: “catch” more scattered light -use large active area sensors Minimize ambient light – with housing design
Although the photosensor array captures more light, much of the light would be scattered away from the sensors. So, our design has a mirror in that region to redirect the scattered light towards the sensors. Also, if we place a lens in front of the laser the volume of the light beam will grow, increasing the probability of smoke particles entering the beam path. Finally, the design will have an optical filter placed in front of the photosensors which only passes light near 635 nm. This will minimize the ambient light reaching the sensor and therefore minimize false positives. Mirrors: redirects scattered light toward photosensor arrayFilter: remove all spectrum components except laser’s (minimizes false positives)Lens: increases effective area of laser beam (i.e. more likely to hit smoke)
The final design tweak is a Lock-in Amplifier which is a general technique used to extract signals from a noisy environment when the expected frequency of the signal is known. Rather than using a DC signal to drive the light source as fire smoke detectors do, a sine wave is used. So, in the presence of smoke, the photosensor would also have a sinusoidal signal. The lock-in amplifier then compares these two sine waves and will amplify only the portion of the photosensors signal which matches the frequency of the laser’s driving signal. So for example, the ambient light in the room will essentially be a DC signal and because it’s frequency does not match the frequency of the laser’s sine wave, the lock-in amplifier will block this part of the signal. Another example would be a light being turned on, which would be a very high frequency event and so it too would not affect the sensor. Laser: sine wave instead of DC, photosensors signal will also be sinusoidallock-in amplifier attenuates any part of the signal with a frequency other than the sine wave which drives the laserAmbient light is essentially DCTurning a light on is very high frequency and random noise will likely be some other frequency
There are several ways in which we can optimize this sensor to work effectively sensing cigarette smoke and to eliminate false positives:Design the sensor’s housing to allow more airflow to the sensing chamberConcentrate the airflow directly to the laser beamUse software to reject anomalies (dust detection, etc.)This will go a long way to reject false positives from our primary sensor.
I say primary sensor, because we will be using our photoelectric detection alongside another secondary sensor. As a secondary sensor, we will use UV and IR detection in combination to detect the lighting of a match or lighter. This will not directly show that smoking is going on, but will serve as a flag to the monitoring station. When this occurs, the sampling rate of our primary sensor will increase. By using this we hope to further reject the occurrence of false positives. (Note that sampling of the primary sensor INCREASES; this sensor is used as a support system, and both primary and secondary sensors run simultaneously.)
The second component of our project is the wireless network. The data that will need to be communicated over the network includes the smoke status of the room, the flame status of the room, the room number associated with the sensor, and a time stamp. We will focus on using a low data rate in order to ease the capabilities on our network.
Say why we didn’t choose ZigBee here (interoperability, cost, lack of ease)Data bandwidth implies we will meet network latency requirementFrequencies meets sell-able in US requirementDon’t say these:2.4 GHz – smaller antenna915 MHz – larger antennaMultiplexing: DSSS, FHSS, OFDMModulation: FSK, OOK
MiWi is a proprietary wireless network protocol from Microchip which was designed to have the same functionality as Zigbee and yet be easier to develop and use. This is beneficial for us since we’ll be able to devote more time to working on smoke detection. One way they’ve made the development process easier is by providing software to assist in the network’s configuration as well as debugging and analysis. They also sell a product called ZENA, shown here, which allows a computer to monitor data transmission within the network. You can also enable a computer to be an actual node in the network. So, we will use ZENA to enable the computer at the front desk to receive data from the smoke detection units. In general, there are two ways Microchip defines its network protocol. First, it provides a software package which is combined with the application firmware we will write. This provides the microprocessor the ability to perform network functions. Microchip also specially designs transmitter receivers to be compatible with MiWi. In addition, MiWi networks have a maximum of a little over 8,000 network nodes, which easily meets our requirement of a 100 room hotel. Additionally, MiWi provides automatic mesh routing. This means that data will always take the shortest route to its destination and if that path is broken, it will automatically find a new path. This feature is great since it will minimize network latency.- Shorter develpoment time: We can focus on smoke detectionMiWi Development Studio: software suite enables easier developmentZena: network analysis tool monitors data traveling on the network, can also connect a laptop to the network (monitoring station)Software package: loaded onto microprocessor, automates networking operationsTransceivers: hardware RF transceiversRoom capacity meets customer requirementsMesh routing: finds shortest path to destination, reconfigures path if path is brokenMeets network latency requirement
Capable of 100 room hotelMemory footprint – coordinator (25 KB) end device (16 KB) - ZigBee: coordinator (100 KB) end device (40 KB)
Special Functionality – noise detection scan (across channels), transceiver sleep/wake,
Note that the laptop solution will easily be < $1,500The final component is the monitoring station, which will be located at the front desk of the hotel. It is here where the staff will be alerted to any smoking violations in any of the rooms. We did some research and the computers at the hotels have their own specific hotel operating system and it would be very difficult installing other programs on these machines. Therefore we decided to buy a cheap laptop that will specifically run our program. As Tim said earlier, ZENA will be used to help monitor the wireless communication packets that are sent through the system. As for the application software itself, there are two options. First is LabVIEW. Most of the details are already programmed into the software and that would help with implementation. The LabVIEW software comes with a fee and also would limit us to National Instruments hardware, which would also increase cost. However the most important downfall of LabVIEW is not many people are familiar with the program and how it functions. We find that user friendliness is extremely important. The worker at the desk should be able to easily navigate and operate the application.
The next interface that is possible is simply developing a Java application. Java provides the ability to control every aspect of our application and is very straightforward in developing GUIs. Netbeans for example has a tool that allows you to develop a GUI without writing any code. All it requires is you to drag buttons and other components where you want them. Although this would require more programming, it would provide much more control both under the application’s “hood” and visually. Because Java provides so much specific control, we could insure that the user’s interface is pleasant and easy to understand and operate.
To summarize, our design consists of three independent components. First, our primary sensor (photoelectric) and secondary sensor (UV/IR) serve to detect the presence of smoke. Second, our wireless network communicates this information quickly and efficiently. Finally, our monitoring station receives the information and informs the attendant through a user-friendly interface.
This is our proposed project schedule. As mentioned earlier, we see this project as having 3 independent stages. By implementing parallel engineering in those 3 stages, we hope to have a working prototype by mid march, giving us over a month to test and fine-tune our design.
Our budget is 2,500 dollars and throughout our research we’ve been considering what we will have to buy and checking prices on those items. Based on that this is how we estimate we will be spending our money with the total being well under budget at 1,400 dollars. This leaves some room for unexpected purchases.
Smoke sentry 2_pdr_presentation
Preliminary Design Review Senior Design Clinic Team 10
Team Members and Roles Core Team: Jeff Frank (EE) – Team Lead Kelvin Kosbab (EE) Tim Nguyen (EE) Mike Masek (ME) Advisor: Ramesh Rajagopalan Observers: Chris Engelmann (Sponsor) Kevin Nicholson (Engineering Coordinator) 2
Background Smoke from fires and cigarettes differs greatly Cigarette smoke particles: ○ Considerably smaller ○ Reduced particle concentration Commercial smoke detectors are designed for fires ○ Insensitive to cigarette smoke 3
Problem Statement Very challenging for hotels to enforce no smoking policies Some guests are sensitive to smoke residue ○ Discomfort ○ Health issues ○ Unpleasant odors Smoke damage can result in thousands of dollars in repair costs Customer needs system capable of notifying hotel staff of smoking violations 4
Statement of Work Deliverables: Research documentation Design documentation Prototype of a cigarette smoke monitoring system Test specifications Data supporting a prototype functioning within specifications 6
Customer Requirements Cost target: Detection Unit: < $50 Monitoring Station: < $1,500 Size target: < twice fire smoke detectors smaller than 8.5” x 8.5” x 3” Improved sensitivity over existing detectors Network capable of 100 room hotel Functional under US regulations 7
Requirements ValidationFunctional Requirement Validation Capable of detecting Establish controlled tests cigarette smoke Record conditions Minimize false positives Run in controlled setting < 2 per hotel Count total false positives Latency Measure using Network: < 20 seconds timestamp Detection: < 5 minutes Testing in similar room 8
Optimization Design Tweaks Lock-in Amplifier Laser sine wave Photosensor sine wave Noise rejection All frequencies components not equal to sine wave attenuated ○ Ambient light ○ High frequency transients 21
General Techniques Optimize airflow into smoke chamber Concentrate airflow to one area Laser beam path Algorithm rejects dust detection 23
Flame Detection UV and IR detection in combination Used as a secondary sensor Applied as an aid for primary sensor Increase sampling rate of Photoelectric sensor Send notification to front desk 24
Sensor Nodes:Microprocessor Requirements on microprocessor not demanding Choose a PIC from Microchip ○ Many microprocessors available Choose the smallest, cheapest one with enough I/O and peripherals Requires SPI, timer, sufficient I/O pins 25
Wireless Network Transmitted Data Smoke status Flame status Room # Time Stamp 26
Wireless Network Protocol IEEE 802.15.4 Low data rate, Personal Area Networks ○ Zigbee ○ MiWi 2.4 GHz – International ○ 250 Kbit/s ○ Shorter range 915 MHz – North America ○ 40 Kbit/s ○ Longer range 27
MiWi Less complex implementation of ZigBee Shorter development time ○ MiWi Development Studio ○ ZENA MiWi standardizes… Network protocol ○ Software package Transceivers Hotel room capability: 8,128 Mesh routing 28
Monitoring Station Application run on Windows 7 laptop USB connected MiWi transceiver (ZENA) Option 1: LabVIEW Pros: Easiest to implement, most of the programming already done Cons: License fee, not as flexible, limited hardware options (NI hardware only)http://labviewwiki.org/images/thumb/3/32/LabVIEW_Logo_Vertical_4c.jpg/200px-LabVIEW_Logo_Vertical_4c.jpg 31
Monitoring Station Option 2: Java application Communication with ZENA via C commands and system commands Pros: Free, flexible, easy to implement, more control, experience in Java programming Cons: more coding, more testing http://www.digitaltrends.com/computing/most-vulnerable-browser-plugin-think-java-not-flash/ 32