Path: bloom-picayune.mit.edu!snorkelwacker.mit.edu!americast.com!americast.com\!americast-post Newsgroups: americast.mech From: americast-post@AmeriCast.Com Organization: American Cybercasting Approved: americast-post@AmeriCast.com Subject: Software packages for acquiring and analyzing data and for Date: Wed, 4 Nov 92 10:28:41 EST Message-ID: Software packages for acquiring and analyzing data and for controlling experiments are converting desktop computers into laboratory tools that are so important that some tests could not be performed without them. Many instrument makers and software vendors provide off-the- shelf packages for programming and running laboratory experiments on desktop computers. Others provide systems that allow users to display and present test results. Quick and Easy Testing Philip Gong, component test engineer at the Isuzu Technical Center of America (Plymouth, Mich.), said routine tests can be configured quickly and easily on an IBM-compatible PC fitted with a 16-channel data-acquisition board. Certain tests on parts destined for Isuzu automobiles are programmed using the Labtech Notebook package from Laboratory Technologies Corp. (Wilmington, Mass.). Gong said that Labtech Notebook is useful for tests with data-collection rates up to 1000 samples per second. One test performed at the center is designed to make sure that car doors will not fail during the life of the automobile. Gong said that the test, while basic in composition, exemplifies how laboratory software is useful in devising and controlling experiments. Reflectors are affixed to the front edges of the doors to be tested. A pair of spaced photoelectric sensors receive input from door movement as they register reflected light. The sensors register either a high or low pulse, indicating the presence or absence of reflected light. Using Labtech Notebook, Gong programs a routine that triggers a timer when the first sensor registers a 10-volt high pulse. The timer stops when the second sensor is triggered. The time between the pulses indicates the speed of the door. Gong said the experiment collects 900 samples per second, yielding accurate speed measurement. For each set of tests on a particular model, the doors are opened and closed many times by many people. The purpose is to get a sample of customers and determine an average for how fast they will open and close the door. ÒWe donÕt want the door to break before the end of the vehicleÕs life,Ó Gong said. ÒIf we ever find a failure, our engineers would have to redesign the door.Ó In addition to measuring the speed of door slams, software is necessary for calibrating the sensors and programming the data- collection routine. Another test performed by Isuzu measures the rotational angle of a hand brake versus the amount of force required to move it. The experiment is a two-channel test, one for force and the other for angle. Angle transducers attached to the brake are calibrated using Labtech Notebook so that every 10 degrees of arc produces 1 volt. Load cells to measure force are calibrated so that 10 kilograms of force is equal to 1 volt. The channels are connected to the data-acquisition board and the voltage is recorded by Labtech Notebook. Results are sent to an Excel spreadsheet from Microsoft Corp. (Bellevue, Wash.), where they are plotted. Curves representing the amount of force required to pull the brake and the consistency of brake motion are generated. The curves are studied for peaks and valleys, which betray inconsistencies in brake operation. ÒThis test lets us see if the brake can be pulled by regular people,Ó Gong said. ÒIf not, or if itÕs too loose, the designers can correct the length of the brake cables.Ó Software at the Controls A component of a complicated product such as an automobile requires extensive testing in order for it to be integrated successfully into the whole. Even a component so seemingly simple as a driverÕs seat has complexities that arise from its interaction with the rest of the automobile. Engineers at another Plymouth, Mich., lab, the Johnson Controls Automotive Systems Group, a division of Johnson Controls Inc. (Milwaukee), that manufactures automobile seats, test products that use hydraulics and pneumatics to simulate life-like conditions. Johnson engineers recently began testing for seat Òchuck,Ó or unwanted movement of the seat in response to vehicle inertia. They devised an experiment in which a servo-hydraulically controlled seat was placed under loads that simulated those resulting from vehicular motion. John Terrich, the Johnson test engineer that directed the seat chuck experiment, said software was essential in controlling the hydraulics and collecting the resulting data. In the experiment, six linear variable differential transducers (LVDT) were placed at control points on the seat frame where the seat would be attached to the track on the floor of an actual vehicle. Loads that simulated a back-and-forth rocking motion were applied. The sensors were calibrated so that every pound of force was equal to 0.01 volt from the transducers. In the first 10 seconds of the test the hydraulic servos applied from zero to 100 pounds of force in one direction. In the next 10 seconds the force dropped back to zero. Then the process was repeated in the other direction. The LVDTs registered the amount of force resulting at the control points during test runs, each collecting 10 samples per second. The experiment was programmed and controlled using the Viewdac data-acquisition and control package from Keithley Instruments Inc. (Cleveland) running on an IBM-compatible PC. Viewdac was used to generate the waveforms that controlled the loads applied to the chair. The advantage of using the software, Terrich said, was that complex waveforms simulating multiple changes of direction could be programmed easily. Without the software, only one waveform could be generated at a time. The actual data were collected using a Keithley Model 575 data-acquisition and control system. Terrich said that once the sensors were in place and calibrated and the experiment programmed, the amount of manual intervention required was minimal. The programming requires about a week of effort. ÒAll we had to do was set the test position of the seat and push the START TEST button,Ó Terrich said. The results of the test were plotted using the Quattro Pro spreadsheet from Borland International (Scotts Valley, Calif.). Widespread familiarity with Quattro Pro and its ability to handle the text requirements of the experiment were cited as reasons for its use instead of Viewdac. Terrich said tests are much more useful when programmed and controlled using laboratory software rather than manual methods. The tests themselves are more repeatable as the sensor calibration and test process information are stored in the computer. Terrich has written Viewdac routines allowing Johnson test engineers to view data in real time before the data are transferred to disk to ensure integrity. More importantly, accumulated results are collected in a data base that is allowing engineers to quantify millimeter-scale seat chuck movement. ÒPrior to our use of data- acquisition software, test results were often open to interpretation,Ó Terrich said. Bill Turnbull, another Johnson test engineer, said software- based laboratory systems are vital when experiments are testing for multiple factors or when extremely precise control is required. Turnbull has devised a fatigue test for automobile seats that relies on air rather than hydraulics to apply loads. Differing loads are applied in sequence according to a script programmed in Viewdac. As the test cycles through, LVDTs collect data, which are used to determine fatigue levels. ÒAir poses a problem for control,Ó Turnbull said. ÒThere are tubes and valves everywhere and it all has to be precisely regulated. There would be no way to run the test without computerized data-acquisition and control systems.Ó Adaptive Testing One of the main advantages of software-based instrumentation is the flexibility it allows in configuring tests. The Measurement Technology Center (MTC) of NASAÕs Jet Propulsion Laboratory (JPL) in Pasadena, Calif., which functions as an internal VAR, providing instrumentation and data-acquisition systems to other departments, depends on software to configure a wide variety of systems to meet the needs of the JPLÕs research community. These systems are configured from an Òinstrument loan poolÓ of sensors, signal conditioners, and data-acquisition boards. MTC technicians put the components together to form data-acquisition systems for any of the major computing environments found at the JPL, including Macintosh, IBM-compatible PC, Hewlett-Packard Series 700, and Sun. The LabView graphical programming language from National Instruments (Austin, Tex.) is used to develop systems for the Macintosh. LabWindows, also from National Instruments, is the programming environment for IBM-based systems. The Virtual Engineering Environment from Hewlett-Packard (Palo Alto, Calif.) is used for HP systems. The N-Power package from Signal Technology Inc. (Santa Barbara, Calif.) is used to configure test systems on Sun Sparcstations. Edmund Baroth, technical manager of the MTC, said it would not be practical to acquire and maintain enough data-acquisition systems in the hardware to cover the variety of testing work carried out at the JPL. Instead, the center provides generic pressure sensors, strain gauges, accelerometers, thermocouples, spectrum analyzers, and other sensors that can be used for any application. Communication with these devices is generally possible through standard interfaces such as RS-232 and IEEE-488. Data-acquisition systems can be configured for a test using the appropriate components in a modular way. These components can then be reconfigured into data-acquisition systems for entirely different tests. Two to eight weeks are required to configure a test from scratch. ÒOur center demonstrates how data-acquisition and analysis systems are easier to configure and control than people think,Ó Baroth said. He added that computers and programming software also give testers the ability to analyze and visualize results on the same platform on which they collect the data. This concept of Òdata fusionÓ is desirable because it leads to more focused results. Baroth contends that graphical programming also gives users a high degree of interactivity with data and their experiments. An example of how the MTC uses the modular approach to building data-acquisition systems was detailed in a report Baroth and his colleagues presented to the American Institute of Aeronautics and Astronautics earlier this year. A test program intended to validate the concept of a segmented parabolic mirror was built around a Macintosh IIfx using LabView. A total of 15 locations on the 45-node adaptive structure were selected as measurement points. LabView was used to program the data-acquisition and instrumentation functions of the test. The data-acquisition and control system was configured to accept 64 channels of data including 30 LVDTs, 15 load cells, and 17 thermocouples. The two other channels were used for calibration. Data were acquired at the rate of 20 samples per second per active channel. The test-bed structure was equipped with six electric actuator motors, which permitted changes in the mirrorÕs shape. LabView was used to create a motor control operatorÕs panel. According to Baroth, graphical programming languages allow tests to be configured much more easily and rapidly than hard wiring physical instrumentation and using nongraphical language routines. Changes in the test requirements can be implemented quickly. The various hardware components remain useful for other tests with only changes in software required to make them work in new capacities. FordÕs Answer to Knocking Hardware-specific data-acquisition systems are generally capable of handling higher rates of data collection than microcomputer- driven systems running data-acquisition and control software. However, the speed of current PC, Macintosh, and workstation platforms is allowing them be used for more intensive data- collection applications. Researchers at Ford Motor Co. (Dearborn, Mich.) are finding that flexible software-programmable tests can be used to collect and process many samples per millisecond. The goal of one such test is to study emissions from automobiles with an eye toward making the product more environmentally responsible. One major source of uncontrolled emissions is engine knock, which is caused by abnormal combustion in the piston cylinders. A research project at Ford is developing test strategies to determine the characteristics of engine knock events, which are as damaging to the engine as they are to the environment. In addition to increasing emissions, secondary combustion causes acoustic waves. These waves disrupt the surface quench layer of cooling gas in the cylinder. Energy from the pressure waves and additional heat are transferred to the engine block. Unchecked, engine knock can lead to a meltdown of the piston heads and damage to the engine block itself. ÒThe step gradient of engine temperature increases by an order of magnitude if gas layers are upset by acoustic waves,Ó said Gottfried Hogh, a computer applications engineer at Ford who has devised a test for acquiring and analyzing engine knock data. ÒThe effects are especially noticeable in high-speed operations.Ó When placed inside test cylinders, piezoelectric sensors register pressure waves occurring during the 2-millisecond period when the piston is near the center of the power cycle. This is when abnormal combustion generally occurs. The sensors pick up pressure changes and produce electric charges commensurate with the intensity of the pressure waves. The signals from the in-cylinder sensors are amplified by a signal conditioner and converted from analog format to digital. The digital readings are, in turn, sent to a digital-signal processing board from National Instruments occupying a slot on an Apple Macintosh Quadra 900 computer. Fast Fourier transform analysis is used in conjunction with a proprietary algorithm developed by Hogh to translate the waveform into a knock-intensity level. A threshold level was set to sort out pressure waves produced by abnormal combustion events from harmless background noise. ÒThere are all sorts of acoustic events that occur during piston cycles,Ó Hogh said. ÒSignal analysis is needed so that you can ignore the other noise.Ó Controls for the experiments were created using National InstrumentsÕ LabView graphical programming software. Hogh said the main constraint in the experiment is processing the data from one firing before the next cycle. Samples were collected every 10 microseconds. The number of samples returned per engine firing varied in accordance with engine speed. Experiments run at 4000 rpm returned 512 samples per firing for fast Fourier transform analysis. The waveform analysis results for a particular engine firing were stored as a single data point. Statistical analysis of the data points from multiple waveform results would then be performed on a sample of events. Short tests would contain between 200 and 600 samples while longer analyses involved 6000 to 10,000 samples. Hogh said very good statistical results could be obtained using a sample of 4000 engine firing events. Hogh stressed that the engine knock analysis tests have the status of a research project at Ford and are not yet influencing automobile design. However, the technique is being taken seriously as a way to identify and rectify problems that relate to performance and reliability. A Bridge Too Far A research project devised to study how dynamic loading contributes to train trestle fatigue over time demonstrated the usefulness of microcomputer-based software data-acquisition systems for collecting data from remote locations under harsh field conditions. The Association of American Railroads and the National Science Foundation funded a project in 1988 to analyze fatigue experienced by common railroad bridges. Researchers at the University of Illinois (Urbana-Champaign) who carried out the studies found many problems with a hardware-based solution. The first phase of the project involved collecting data from a trestle that was a three-hour drive from the university. Researchers attached strain gauges at various points on the open- deck-plate bridge. The gauges were wired to a self-contained data- acquisition system capable of collecting and conditioning 26 channels of data. The unit was programmed in high-level Basic. The amount of data collected made downloading by modem to the laboratory prohibitively expensive so analysis had to be performed unattended on the spot. After the data were collected, the unit performed a rainflow histogram analysis on the loading experienced by the bridge. Due to variations in train composition that affect loading, data on 100 train passings were required to develop an accurate model. ÒThe first bridge was a huge learning curve,Ó recalls Scott Schiff, a former graduate student at the University of Illinois who was involved in the project and who is now an assistant professor at Clemson University (Clemson, S.C.). ÒWe had problems with the reliability of the equipment, vibrations on the bridge, and the time required to process data.Ó The unit used at the first bridge required several hours to run the analysis routines and could not collect additional data while analysis was in progress. Thus, some trains were missed, which lengthened the data-collection process more than was necessary. Also, the unit did not have the storage capacity to keep all of the raw data collected so it was erased after the histograms were generated. Unfortunately, since the unit did not have a monitoring system, some analysis was performed on incomplete, corrupted, or otherwise useless data. During the six months required to collect good data on 100 train crossings, the researchers decided to change their approach. Without altering the instrumentation significantly, they replaced the self-contained data-acquisition unit with an IBM-compatible 80386 PC equipped with a 32-channel data-acquisition card and a 105-MB hard drive. The SnapShot Storage Scope package from HEM Data Corp. (Southfield, Mich.) was installed to acquire and store the data. The Snap-Stream software module, also from HEM Data, provided high-speed data streaming to improve the data- acquisition-triggering process. The system was programmed to collect three minutes of data when tripped by the weight of a passing train. The PC is set up at the data-collection site and must be enclosed in a ventilated box to protect it from inclement weather. The PC-based system collects and stores data collected from strain gauges but does not run the analysis routines. Researchers drive out to the data-collection site every two weeks to download data onto floppy disks for analysis back at the lab. ClemsonÕs Schiff, who is continuing the work at his university, said the PC- based approach allows data on 100 crossings to be collected in about one-third of the time required previously. The easily programmable software also allows him to deal with unforeseen problems without major alterations in the test. ÒIn addition to vibrations, we found that the cabling we strung on the bridges was acting like a huge antenna, picking up signals,Ó Schiff said. ÒThe software let us program the system to filter out extraneous noise.Ó Virtual Instrumentation Current personal computers and workstations have the memory and processing power to handle many low-end and intermediate laboratory testing applications. Add-ons like memory extension and graphics accelerators provide even more capacity. Equipped with data-acquisition boards and running a proper software package, desktop systems can tackle tests once reserved for difficult-to- program and inflexible dedicated instruments. Graphical user interfaces provided by many laboratory software packages allow test engineers to build a virtual instrument and its attendant control panel. Components such as sensors and signal conditioners are standard enough to be addressable by most commercially available data-acquisition and analysis systems. ÒSoftware is giving engineers the ability to design and control tests involving multiple sources of data and tens of thousands of data points with custom interfaces,Ó noted Richard Walter, president of HEM Data. ÒAll of this is possible using common components and conventional computers with little modification.Ó m Copyright 1992, Mechanical Engineering. For more information, send-email to American Cybercasting Corporation (usa@AmeriCast.COM)