University of Connecticut Department of Electrical and Computer Engineering

Transcription

1 University of Connecticut Department of Electrical and Computer Engineering ECE 4901: Fall 2016 Spring 2017 Team 1707 (Trans-Tek): Automated Test System and Fixture for Angular Displacement Transducers Final Report Michael Caiafa (CE) Andrew DiGiugno (EE) Shane Keney (CSE) Michal Zielinski (EE) Sponsor Contact Information: Nancy Hamilton, CEO Jeff Gladu, Chief Engineer May 5,

2 Abstract This document will encompass the design of an automated test system and fixture for angular displacement transducers, or ADTs. This design was requested by Trans-Tek Incorporated, who has worked in conjunction with our team through the duration of its development. This report includes various design ideas that have been used to create a successful design. Background Angular displacement transducers (ADTs) are position sensors based upon a differential capacitance such that the output of the ADT will be proportional to the angular displacement of the shaft. Because it is primarily used as a sensor, it is imperative that it is not only calibrated, but that its calibration is also verified. Currently, Trans-Tek accomplishes the latter with an outdated test fixture (see Figure 1). To use it, the user must manually set the ADT into the test fixture and continue with the verification process with the aid of in-house developed software. This software is used in conjunction with the test fixture (Figure 1) and pre-existing electrical equipment needed for the verification process. This procedure continues with the user being prompted by the software to align the null position of the ADT with the 0 index mark on the test fixture gear wheel. The user is then responsible for fastening the ADT in this location so that the rest of the verification procedure can continue. After the null position has been located and the ADT is fastened to the test fixture the user, prompted by the software, moves out to the maximum stroke of the ADT being tested to record output via button press. The user then is instructed to take the remaining angular degree points of the ADT passing back through the null position and reaching the maximum angular stroke on the opposite side of null. These points are all taken by disengaging a gear lock mechanism, rotating the gear wheel to the defined angular degree specified by the software program and then recording the specific output voltage associated by button press or mouse click in software. The verification procedure is then completed with a linearity calculation of the output voltages recorded at the specific angular degree points taken through the test of the specific ADT unit. Problems with this test fixture are rooted primarily in the fact that it requires constant user intervention. Because the fixture is hand-operated, it necessitates the use of excessive employee time. ADT units can have up to 101 angular degree points to record and as such can be a very time consuming testing procedure. This time consumption is increased when user errors such as improper setup of the ADT or skipping over a specified step requires user to go back and correct or redo an entire verification test. This issue is exacerbated even further by the tediousness of the testing procedure. Furthermore, the current test fixture shown does not have a definitive accuracy model and has even additional error in mechanical construct with nondefinitive stability when locked into a specific gear location. Finally, the fixture only allows the ADT to be tested in three-degree increments, which is not a sufficient resolution for Trans-Tek s needs. 2

3 Figure 1. Current ADT test fixture. Courtesy of Trans-Tek Inc. 3

4 Objective The ultimate goal of this design was to redesign this test fixture in a manner that resolves the aforementioned issues. Specifically, the design reduces the required user intervention with an overall objective of having a fully automated test fixture, reaching a definitive accuracy model, and lastly increasing the overall resolution of the test fixture. Solution General To accomplish the objectives, our solution was to use a highly accurate motor drive to rotate the ADTs. This motor is computer controlled, which therefore limits the need for human interaction. There are three main options on the market evaluated for suitability as the motor type used in the design: stepper, servo, and encoded stepper. All three allow for accurate position control; the stepper and encoded stepper can accomplish this via microstepping, and the servo can accomplish this with its fine resolution encoder. Ultimately, the encoded stepper was chosen. The ADTs to be tested exist in two different possible dimensions. As a result, the mechanical test fixture must be adaptable to each. Finally, with the aim of this new fixture being full automation, has been fully integrated into Trans-Tek s current software suite. This software suite is written in Visual Basic, and has existing methods to communicate with the ADTs via GPIB between the power supply and the digital multimeter. The software setup does not conflict with this design; therefore, the existing components have not been modified. The following block diagram (Figure 2) illustrates the changes made to Trans-Tek s current ADT test infrastructure (illustrated in red). The current PC software has been upgraded to allow for communication to the motor controller. This communication is done via RS232. From the motor controller, power signals are sent to the motor and encoder feedback is sent back to notify the controller of the motor s position. Finally, the motor rotates the ADT accordingly. 4

5 Figure 2. Block diagram for Trans-Tek ADT testing infrastructure. Black, darker boxes represent current elements; red, lighter boxes represent proposed additions. 5

6 Motor Choice Stepper Motor A stepper motor is an electric motor that moves in discrete steps. As shown in Figure 2, the magnetized rotor has a set number of teeth that encompass its perimeter. When one of the electromagnets is activated, the rotor shifts so that the teeth become aligned with the poles of the electromagnet. If the electromagnets are activated in sequence, the rotor will undergo a certain amount of motion. Likewise, if the currently activated electromagnet stays energized, the rotor becomes locked into place. This allows for position control. Figure 3. Diagram of a stepper motor. Numbers 1, 2, 3, and 4 are labels for each of the electromagnets, and the red, center portion is the gear-shaped iron rotor. 1 To implement a stepper motor in the proposed design fixture, there exist a few considerations. First, stepper motors with a total number of steps outside of either 200 or 400 are not widely available. With those two step options, it is impossible to rotate the ADT shaft by a whole number degree value. This is not satisfactory for Trans-Tek s needs. To compensate for this, microstepping can be used. Whereas activating the electromagnets in sequence results in full-stepping, activating more than one electromagnet at a time allows for the rotor to be moved in partial steps. Therefore, the resolution can be increased well past 200 or 400 steps, and, with an appropriate stepper motor controller, will allow for a high enough microstep count to adjust the ADT shaft in whole degree increments. Another consideration is that the stepper motor system is open loop. For the purposes of this solution, an open loop system means that it is not possible to be given feedback concerning the current position of the stepper motor. This is problematic because the test fixture must rely on position control, and if the position of the stepper is not where it is expected to be, the fixture will produce inaccurate results without any way of notifying the user. It is also important to note that microstepping is known to decrease accuracy, but data sheets do not specify by how much. Thus, without position feedback, it will be difficult to determine if the required accuracy is being met

7 Servo Motor A servo motor is a closed loop motor system that allows for precise position feedback and control. Unlike a stepper motor, it rotates smoothly; it does not undergo a rapid sequence of steps to induce rotation. The position is detected by a device called an encoder. Figure 4. High-level design of a servo motor system. 2 Because the motor does not have steps, position control is done by continuously inducing minute rotations of the shaft. These rotations correspond to the location determined by the encoder, and the rotations are the result of the motor system attempting to align itself properly. This leads to a non-negligible dither. The specific servo motor chosen for this project solution is both brushless and DC. It also has 10,000 different positions defined by its controller, allowing for an accuracy of ±0.036 degrees. Therefore, it is capable of providing a per degree resolution while simultaneously meeting the required accuracy. Stepper vs. Servo The servo motor both generates less vibration and is noticeably quieter than a stepper motor. However, this all comes at a cost; a complete servo solution will be about $ more than a complete stepper solution. Furthermore, as previously mentioned, the servo experiences a non-negligible amount of dither when holding at a specific point. This dither is not present in stepper motors, and is thus a drawback to the servo option. Based upon these details, it is clear that the benefit of the servo is that it has an encoder, and is therefore a closed-loop system. Meanwhile, the stepper motor is beneficial mainly because it doesn t have the dither of the servo, which upon further research was determined to be degrees. This value is too large for the servo motor to meet the project specifications. Ideally, a stepper with an encoder, or a closedloop stepper system, was the best design choice

8 StepSERVO The term StepSERVO is used by Applied Motion to refer to a closed loop stepper motor. This motor has the capability to hold a constant position without any dither with the added benefit of active position feedback via a built in encoder. In addition, it is capable of 20,000 positional counts which gives it a higher resolution than any other option discussed. This option also fits well within the specified budget and accuracy constraint as the positions are only separated by degrees. RS232 vs. Ethernet Another consideration that was made is the communication protocol used between the motor controller and the PC. The options available were RS232 and Ethernet. RS232 would require no additional cost but has an impending obsolescence. The biggest reason to use ethernet is to make the design more future proof. However, it should also be noted that ethernet is less susceptible to electronic noise and possesses a higher bandwidth. For the StepSERVO option, ethernet is only available on motor models that are much larger than what is required. This means that, if ethernet were to be chosen, the accompanying motor would not only be more expensive, but also require the use of a larger test fixture. Based upon the comparison between the two protocols, the marginal benefits of ethernet do not outweigh the price savings and design considerations of the larger motor. 8

9 Software Development One necessary specification given by Trans-Tek was integrating the automated test fixture into their current software suite of VB.NET. The software was designed and functions as closely to the current testing procedure as possible for ease of use for QC Department of Trans-Tek who will be the ones in direct use of the automated test fixture. The below information regards the development of the software as well as the way in which it functions that is not already explicitly described in the user manual attached to the end of this report (see Appendix). The first important feature added to the software was creating a button present on the Main Window screen for the user to distinguish between an automated calibration run and a manual calibration run. This is designed so that in the case of an automated testing calibration not being possible the user can manually calibrate the ADT unit. Next it was necessary to creating the serial communication between the PC, software, and motor. The software will be using the serial connection to move the motor as necessary to the necessary points during the calibration run. In setting up the serial communication between the software and motor we found that it was necessary to append a carriage return to each command as the stop bit for the motor. The following below is the basic structure that will be used to send commands to the motor via the software: Command Code + (Additional parameter) + vbcr The following commands will be used in the software: EP, FL. The first command (EP) is used in the software for zeroing the encoder position of the motor as well as requesting the current encoder position for encoder feedback as will be described later. The second command (FL) will be used specifically for moving the motor to the necessary points during the calibration run. The additional parameter field as shown above will be populated with 0 when calling EP command code to zero the encoder position and will also be populated with any number between and when used with the FL command to represent a relative location count for the motor to move to. Once the serial communication was established the development of the calibration test with the motor and integration into the software began. The 600 and 605 series that the automated test system needed to be designed to calibrate were not going to be any different when sampling points. However, different parameters needed to be measured based on the different units and would be reused from sections of the programming code supplied by Trans-Tek. These sections were used to measure things like input voltage, input current, etc. based on the unit and would be taken as soon as the calibration began before sampling points of the ADT unit. The measured parameters would be updated into the Take Data window which is the window that will be used when calibrating the unit. Upon conclusion of taking these parameter measurements and updating the Take Data window as necessary, the software will automatically move the ADT to the null location. The null location for each unit will represent the point at the middle of the stroke of the ADT. The software first determines what voltage it needs to look for the specific unit being tested based on an XML file containing all the necessary unit details. Next, the software determines the general 9

10 location of the motor based on the voltage that is currently being outputted by the unit. The software can almost instantaneously then move to the null location if the user aligned the unit close enough to null. To find the null location the unit must be aligned ±45 degrees to the actual null. This alignment step was not possible to automate as it is only based on aligning a visual mark on the shaft with a mark on the housing. However, it is very easy to attain alignment manually when placing it into the test fixture. After finding the null location a window will appear called the Pre-Adjust window which will allow the user to move the motor as necessary to adjust any mechanical or electrical constraints of the unit being tested. Once the user continues past this screen the main portion of the calibration run will commence. For symmetric ADT units, the null location will be at zero volts. For asymmetric units, the null location will be at whatever voltage is the at the middle of the stroke. From the null location, the software will move the motor to the minimum stroke location and slowly increment through the stroke of the unit taking the output voltage at each point. The Take Data window will be updated correspondingly as the calibration run proceeds. On conclusion of the calibration run the software will calculate the linearity of the data that was collected. This is calculated by first creating a best fit line to find the optimal voltages at each degree. This is then subtracted from the measured voltage and then divided over the stroke of the unit to calculate a percent error. For a practical description of the software, please consult the Automated ADT Test Fixture User Manual, provided as a supplement to this document. Figure 5. Main test results display of software 10

11 Mechanical Fixture For the redesigned test fixture to be effective, it must also have a stable mechanical structure that holds the motor, motor controller, and ADT during testing. A preliminary design for the fixture is shown below. Figure 6. Straight-on and isometric views of the prototype CAD model mechanical fixture Figure 7. CAD model of final mechanical fixture revision(left) and final test fixture machined from aluminum(middle and right) The base of the fixture is a rigid, four sided rectangular structure. The motor is mounted to the lower level of the structure, while the ADT is attached to the upper level. This vertical orientation eliminates radial forces on the shafts of the motor and ADT, preventing any measurement artifacts due to, for example, the heavier section of the motor rotor being pulled down by gravity. In our first two revisions, to allow the testing of multiple sizes of ADTs, a 11

12 replaceable mounting ring is used. The ring itself sits on a shelf in the top surface of the structure, and has an inner shelf that is machined to the outer diameter of the ADT. A different size mounting ring is used for each size of ADT. This scheme allows the fixture to support any size ADT with a diameter less than 5 in., simply through the creation of a new mounting ring. The mounting ring is retained by two set screws (represented by rods in the diagram) and corresponding threaded holes in the upper lip of the ring. When a batch of ADTs is to be tested, the appropriate mounting ring is installed, and the set screws are screwed in part way, so that they do not enter into the inner area of the ring. The ADT is then inserted into the ring, and the set screws are tightened until the ADT is firmly retained in the fixture. Another benefit of this design is that the ADT can be removed from the fixture by just a quarter turn on each set screw. The motor is mounted on the lower level of the structure in a bracket, which interfaces with the machined top surface of the motor, on the same side as the shaft. This machined surface also has integral threaded holes for attaching to the bracket. The bracket attaches to the lower level of the structure with 4 screws to be screwed into the lower level itself. The holes present in the bracket will be oversized to allow fine lateral adjustment of the location of the bracket on the lower level, allowing the shaft of the motor to be precisely aligned with the shaft of the ADT. The two shafts will be connected with a coupling with two threaded holes and set screws. Since the diameter of the ADT shaft varies according to the model, different couplings can be made for each model. Again, this allows the easy extension of this fixture for new ADT models. Since none of the team had significant CAD or machining experience, we worked closely with Trans-Tek on fine tuning this fixture for manufacturability and cost-effectiveness. After discussing our first two revisions with both engineers and employees who would be using our test fixture in the future, we created our third and final revision. Differences in this revision included a base plate for added stability and the fixture being attached to its power supply. This was suggested by the ADT testers at Trans-Tek because a single connected unit would be easier to move from place to place. Affixing the unit to the power supply also allows for more comfort when using the fixture as it is higher off of the desk and the tester will not have to bend down to place the ADT. In addition to this, the original concept of mounting rings with set screws to allow for different size ADT s was removed from the design and replaced with a new mounting system which was less cumbersome when changing ADT type. The original design required the set screws to be threaded all the way into and out of the ring every time it was to be changed. The final design incorporated two separate mounting plates, each with pinholes allowing them to be attached to the fixture. This allows for quick and easy switches between testing the 600 and 605 series. In addition to this, if a different size ADT ever needs to be tested in the future, a new mounting plate could easily be made to fit it. The final test fixture was built out of aluminum, which is inexpensive and easy to source. Aluminum was chosen over steel for its lighter weight and easier machining. 12

13 Accuracy Verification While the StepSERVO chosen for this design is specified to have an accuracy well within that which is required for this project, testing has been done to verify said accuracy. In order to do this a mount has been printed which allows a focusable laser pointer to be affixed to the motor shaft. The laser is projected onto a wall 40 feet away. By measuring the displacement of the laser point as the shaft rotates, it is possible to use simple trigonometry to determine the accuracy and repeatability of the motor. Figure 8. Showing the mathematics used to determine accuracy of the motor using a laser pointer In practice, a much higher focus was placed on the repeatability of consecutive movements, rather than absolute accuracy. It can be assumed that the encoder disk used in the StepSERVO is a precisely made part, and that any errors in the system are a result of the motor drive s inability to perfectly position itself relative to the encoder disk. As such, repeatability of the motor drive to move according to the encoder disk becomes the only concern. A set of results of this kind of repeatability verification are presented in Figure 6. These results are digitized versions of those collected manually by marking a piece of paper with the current laser position after every movement. 13

14 Figure 9. StepSERVO repeatability results. The first test scenario in Figure 9 represents repeated movements of a single step at a time, equal to degrees. It can be seen that after a few initial evenly spaced movements, the motor enters a regime of paired steps, where every other step is almost as large as two steps, but the steps in between are much shorter than normal steps. The most probably cause for this is that after a number of initial steps, error in the motor positioning accumulated such that the motor index is positioned at the very edge of an encoder slice. As the motor begins to move, it almost immediately enters the next encoder slice and starts decelerating before it was able to accelerate to full speed. This is the short step observed. At this point, the motor is at the far edge of an encoder slice. As it begins to move again, it goes through much more angular displacement before it enters the next encoder slice. Since the motor now has significant inertia, by the time it decelerates, it has rotated again to the very edge of an encoder slice. This is the long step. The cycle then repeats. However, the next test in Figure 9, which represents repeated movements of two steps at a time, does not exhibit this behavior. In fact, the step sizes are very regular. Evidently, error does not accumulate when the movement of the motor is greater than a single step. Since during ADT testing angular movements are almost always at least one degree (55 steps), error will not accumulate, and step sizes will always be as expected. Also, since the accuracy specification is +/ degrees, even a single missed step of degrees will not invalidate our desired accuracy. In cases where reliable single step movements are required (such as in the acquisition of the exact null position of an ADT, the third test in Figure 9 shows that it is possible by performing movements of, for example, 10 steps forward and 9 steps back. Error does not accumulate for these relatively large movements, and the resulting displacements are perfectly regular. Several methods of obtaining native single step support were attempted, but none proved fruitful. These included attaching a load mass onto the shaft of the motor (since ADTs themselves present such low torque loads), and adjusting the PID coefficients of the motor 14

15 controller. Even a mass large enough to cause steady state oscillation in the motor did not improve native single step reliability nor did the manufacturer recommended changed to various motor parameters. Ultimately, the manufacturer definitively stated that the motor is not reliable under single step movements. However, this does not compromise the design in the least. The workarounds discussed above preserve all desired functionality while compensating for the shortcomings of the StepSERVO. 15

16 Results Figure 10. Final design setup After the design was functionally complete, it was desired to obtain a quantitative measure of the superiority of the redesigned, automated fixture as compared to Trans-Tek s existing, manual fixture. To accomplish this, a statistical analysis of the repeatability of each fixture was performed. Using a particular 605 series unit as the device under test, 15 calibration runs were performed using each fixture. This particular 605 unit had a stroke of 300 degrees, with 51 points to be sampled at 3 degree increments along the stoke. After each calibration run, the voltage readings at each point were saved into a spreadsheet document. After all 30 trials were completed, the standard deviation in voltage at each angular position for both fixtures was calculated and plotted. The resulting plot is shown in Figure

17 Figure 11. Repeatability of old and new test fixtures. Clearly, the new design is a marked improvement over Trans-Tek s old fixture. The large spikes is the middle of the ADT stroke (most likely due to particularly worn gear teeth) present on the old fixture are completely gone, and the standard deviation values are universally lower. In fact, when averaging standard deviation values over the entire stroke, the new design shows a standard deviation in voltage that is only 27.6% that of the old fixture: a very significant improvement. 17

18 Timeline Figure 11. Timeline for August 2016 through December Figure 12. Timeline for January 2017 through May Development for this design began in early September A variety of background research, such as the theory behind ADTs, the intricacies of the project, and minor planning, was done until the end of September. Coinciding with this was the development of our project statement, completed on September 22nd. Throughout much of early and mid October, the project progressed into developing solution proposals. This is where the stepper and servo motor options, the RS232 and ethernet options, and the test fixture ideas were considered. As these ideas were being contemplated, the presentation for the design review was created and based upon these ideas. The presentation was given October 28th to professors in the University of Connecticut s Electrical and Computer Engineering department. The presentation also served as the foundation of the project proposal, completed November 2nd. After creating these deliverables, project development was resumed with the modeling of the fixture in SOLIDWORKS. During the weeks of November 14th and November 21st, slight modifications to the test fixture were made, the StepSERVO motor option was chosen, and a bill of material was sent to Trans-Tek. During the winter intersession, the focus was to create a test fixture prototype. This was done at Trans-Tek s facilities, utilizing their three-dimensional printer. With the prototype created, the other components--motor and ADT--were affixed to it to verify its design. The spring semester began by outlining software development. The weeks of January 18

19 16th and January 23rd were used to verify serial communication in VB.NET to the motor. They were also used to explore the mathematics of motor step verification via laser testing. Continuing with software development, GUI design began during the week of January 30th. Software remained in development throughout the months of February and March, with the development of both the logic and GUI occurring in tandem. As development progressed, there was a meeting with Trans-Tek at the end of February. This meeting involved the discussion prototype feasibility and employee feedback about its ease-of-use. It was at this point that slight revisions were made to the test fixture model, and these revisions were approved by Trans-Tek to become the final fixture. While waiting for this final fixture to be machined, laser testing was done throughout the beginning of March. This time included a correspondence with Applied Motion about the laser test results. At the end of March, software development was concluded, and testing began. Testing was continued when the final fixture was machined, which occurred in mid-april. Testing was also the last major component of the project; therefore, the project was completed during the week of April 17th. The remainder of the month consisted of preparing for Demo Day presentations, and completing the deliverables. Budget Trans-Tek offered a budget ceiling of approximately $2000. This did not include the parts and material necessary to create the fixture itself, as those were supplied freely and inhouse by Trans-Tek and their machine shop, respectively. Figure 13. Final budget and bill of material 19

20 Conclusion The overall goal of this project was to improve ADT calibration testing. This objective was split into three major objectives: reach a definitive accuracy model, increase the overall resolution of the test fixture, and have the new test setup be fully automated. In terms of the definitive accuracy, the comparison of fixture output voltage standard deviation by angular position (Figure 11) shows that this has been accomplished. The overall resolution was inherently increased by the choice of motor. Whereas the old test fixture had 120 angular positions, the motor encoder has 20,000 counts. Through the use of microstepping, 20,000 different encoder positions allow for 167 times more angular positions. With the new fixture, once an ADT calibration test is initialized, there is no need for user intervention. The test runs to completion, and it outputs the calibration test results to the user. This satisfies the objective of full automation. Personnel and Collaborators Sponsor: Trans-Tek Inc. Project Advisor: Professor Ali Bazzi Team Members: Shane Keney Major in Computer Science Andrew DiGiugno Major in Electrical Engineering Michal Zielinski Major in Electrical Engineering & Computer Engineering Michael Caiafa Major in Computer Engineering Minor in Mathematics 20