DESIGNING A LOW-COST SENSOR SYSTEM CAPABLE OF DETECTING A V304 PLASMA CUTTING CNC MACHINE’S BRIDGE LATERAL DISPLACEMENT WITH AN ACCURACY OF 50 MICRONS
By
ALIN GEORGE GRIGORITA
GRADUATION REPORT
Submitted to
Hanze University of Applied Science Groningen
in partial fulfillment of the requirements for the degree of
Fulltime Honours Bachelor Advanced Sensor Applications
ABSTRACT
DESIGNING A LOW-COST SENSOR SYSTEM CAPABLE OF DETECTING A V304 PLASMA CUTTING CNC MACHINE’S BRIDGE LATERAL DISPLACEMENT WITH AN ACCURACY OF 50 MICRONS
by
ALIN GEORGE GRIGORITA
The present paper is to describe how the issue encountered by Voortman Steel Machinery, namely obtaining visible errors in symmetrical steel plate cutting due to bad bridge alignment, was handled by using a sensor system that can be built with minimum amount of resources and that can still take measurements with extreme accuracy (50µm).
When put at work, the V304 plasma cutting machine’s bridge is moving along two parallel rails back and forth, and every now and then, due to different reasons such as a desync at the rail motors, the bridge gets dislodged, therefore, it travels on the axis perpendicular to the rail axis. Although the value of interest is the difference in position of the two lateral extremities of the bridge with respect to the rail axis, the proposed solution is to measure with the aid of inductive proximity sensors the displacement on the perpendicular axis, which will be referred to as the x-axis, and calculate the desired original displacement. The system can be regarded as a maintenance system due to its added value to the machine’s output and to the avoidance of a long term destructive impact on the machine that a malfunction such as the one that is mentioned could inflict.
The lack of materials forced improvisation and reduced number of experiments but the accuracy of the system has been proven hence the concept can be regarded as a viable solution to Voortman’s issue.
DECLARATION
I hereby certify that this report constitutes my own product, that where the language of others is set forth, quotation marks so indicate, and that appropriate credit is given where I have used the language, ideas, expressions or writings of another.
I declare that the report describes original work that has not previously been presented for the award of any other degree of any institution.
Signed,
Alin George Grigorita
ACKNOWLEDGEMENTS
I would like to take advantage of this opportunity and thank Dr. Wilco Bonestroo for finding the project, assigning it to me and guiding me throughout the whole project course, while vouching for me as an engineer to other people and providing me a research environment at the department ‘Ambient Intelligence’ within Saxion University of Applied Sciences, Enschede. I would also like to express my gratitude towards Mr. Martin Alberink for being the link between Saxion and Voortman Steel and for guiding me and providing constructive feedback and setting me on the right path whenever things seemed to deviate. Finally, I would like to thank my supervisors from Hanze Institute of Engineering, Mrs. Fenna Feenstra and Mr. Bryan Williams who provided constructive feedback to my report submissions.
TABLE OF CONTENTS
Page 5
List of Figures...6
Chapter I. RATIONALE...8
II. SITUATIONAL & THEORETICAL ANALYSIS ...12
Principles...12
Sensors...18
III. CONCEPTIONAL MODEL ...26
Y-axis measurement...26
X-axis measurement...27
IV. RESEARCH DESIGN ...31
V. RESEARCH RESULTS………..…33
. VI. EXPERIMENTAL RESULTS……….……34
VII. CONCLUSIONS ………34
VIII. RECOMMENDATIONS………34
REFERENCES CITED...36
Appendix A. DISCARDED RESEARCH AND CONCEPTS...37
B. DATASHEET OF CHOSEN INDUCTIVE PROXIMITY SENSOR BAW M12MG2-UAC20B-BP03...48
C. EXPERIMENTAL PROCEDURE...51
D. TABLE OF DISTANCE READINGS………....53
Page 6
Figure 1 Voortman V304 plasma cutting machine [3] ... 10
Figure 2 Voortman V304 plasma cutting machine 3D SketchUp model ... 10
Figure 3 Voortman V304 plasma cutting machine bridge 3D SketchUp model ... 11
Figure 4 Sketch of the view from above of the Voortman V304 plasma cutting machine ... 11
Figure 5 ‘Illustration of a fibre optic intensity sensor for distance measurements. (a) Two-fibre sensor, showing the object and the operational principle (b) Various geometries for fibre bundle sensor heads’ [4] ... 13
Figure 6 ‘Principle of optical triangulation sensor. The unknown distance, D, is determined from the known distances 𝐸,𝐹 and the measured value of G—the distance to the pixel in the position sensitive detector (PSD) recording the image of the laser spot on the measured object’ [4] ... 14
Figure 7 ‘(a) Pulsed time-of-flight measurement showing three regimes where the time of flight is (i) much longer, (ii) similar to, and (iii) shorter than the pulse width. For nanosecond pulses these cases correspond to distances (in air) of >50 m, a few meters, and <1 m, respectively. (b) Intensity modulated time-of-flight is an alternative method suitable for distances in range (ii). The phase shift is indicated by the double-headed arrow.’ [4] ... 14
Figure 8 ‘Principle of fibre optic monochromatic and polychromatic confocal sensing. (i) Monochromatic confocal sensor with an object at the image plane. (ii) Monochromatic confocal sensor with an object displaced from the image plane. (iii) Polychromatic sensor where the image plane position varies with wavelength and a different wavelength will satisfy the confocal geometry for each object position’ [4] ... 15
Figure 9 ‘Basic setup for WLI and typical interference pattern as a function of the reference arm mirror movement. A low-coherence source is used, in this case a fibre-coupled 1.5 µm super luminescent diode with 60 nm spectral width. The interferogram shows an oscillation period of half the wavelength, and width corresponding to the coherence length of the source’ [4] ... 16
Figure 10 ‘Demonstration of fibre-optic-assisted Doppler velocimetry. (a) The experimental setup, using a pigtailed coherent 1.5 µm laser source and a moving mirror. (b) Detector output measured by an oscilloscope for a motion of 100 µm/s. (c) Detector output measured by an electronic spectrum analyser for a motion of 100 µm/s (black), 50 µm/s (blue), and 25 µm/s (red)’ [4] ... 17
Figure 11 ‘Comparison of typical resolutions and working distances of the optical distance measurement techniques discussed in this article. The entry for triangulation is divided into a filled region, corresponding to measurement from a single observation point, and a shaded region for multiple observation points’ [4] ... 17
Figure 12 Ultrasonic sensor working principle - the ultrasound hits an object and it is reflected back to the sensor ... 19
Figure 13 ConfocalDT IFS2402 - confocal miniature sensor from Micro-Epsilon [5] ... 20
Figure 14 Specifications of the ConfocalDT IFS2402 sensor[5] ... 20
Figure 15 zygo® MK-2 Interferometer [6] ... 21
Figure 16 optoNCDT 1320 // Compact Laser Triangulation Displacement Sensor from MICRO-EPSILON [5] ... 21
Figure 17 List of specifications of optoNCDT 1320 laser triangulation displacement sensor [5] ... 22
Figure 18 RPS-412A High Accuracy Sensor - Ultrasonic sensor from Migatron Corp.[7] ... 22
Figure 19 InduSENSOR LVDT - inductive displacement sensors from MICRO-EPSILON [5] ... 23
Figure 20 List of specifications for InduSENSOR LVDT inductive displacement sensors [5] ... 23
Figure 21 GX-F/H type inductive sensors [8] ... 24
Figure 22 Specifications of GX-6 inductive sensor[8] ... 24
Figure 23 DI-145 DAQ device from DATAQ INSTRUMENTS ... 25
Figure 24 capaNCDT 6200 - capacitive sensor from MICRO-EPSILON [5]... 25
Figure 25 Specification table of the capaNCDT 6200 capacitive sensor [5] ... 26
Figure 26 Congruency between the angles formed by the bridge and the x axis and, respectively, the angle formed by the body parts with the y axis ... 27
Figure 27 3D model of the plasma cutting machine and theoretical sensor implementations ... 28
Figure 28 3D model of a sensor setup designed to determine the angle to which the bridge side components drift ... 29
Figure 29 3D SketchUp model of the inductive sensing setup: two inductive sensors placed in extremity points on one face of the machine, pointing at a metal flat, parallel to the machine plate ... 30
Figure 30 3D SketchUp model of the inductive sensing setup: two inductive sensors placed in extremity points on one face of the machine, pointing at a metal flat, parallel to the machine plate (another perspective) ... 31
Figure 31 Geometrical explanation of the determination of the desired 'd' distance, where the H shaped drawing represents the two body parts united by the bridge, viewed from the top, and where the line uniting A' and B' is the metal plate, A is the point where one of the proximity sensors will be, and, respectively, B will be the location of the other sensor. ... 31
Figure 32 Provided micrometre ... 33
Figure 33 USB icon in Oracle VM VirtualBox ... 33
Figure 34 Calibration in software ... 34
Figure 35 30 distance values taken with the micrometre and with the sensor software, on the x-axis being the measurement number and on the y-axis being the value of the reading in mm (the table of values can be found in Appendix D) ... 35
Figure 36 Graph of the sensor output in Hardware Manager ... 36
Figure 37 Puncher temperature measurement system sketch made by the student ... 38
Figure 38 Oil contamination measurement system sketch made by the student ... 39
Figure 40 Sketch – front view of torch system with a radius laser scanner attached to the metal support of the torch ... 40
Figure 41 Sketch – front view of torch system with a radius laser scanner attached to the metal support of the torch, under an inclination to the right ... 40
Figure 42 Sketch – front view of torch system with a radius laser scanner attached to the metal support of the torch, under an inclination to the left ... 41
Figure 43 Sketch - side view of torch system with an angular sensor attached to the metal support of the torch when it is pointing straight to the plate and when it is at an angle ... 41
Figure 44 Steel plate on bed to which a rotating laser scanner is attached to measure the thickness of the plate ... 41
Figure 45 Example of serial output for the electrical cabinet which can display each temperature and humidity value for each sensor separately, the lowest/highest and average values every time a sample is taken, while data is stored at the end of each day. The daily values reset on a daily bases while long term arrays store the values for the whole running period. The values are also compared from a day to another and in case there are big differences, a warning message is displayed. Certain thresholds activate symbol display next to the alarming values. ... 42
Figure 46 Environmental temperature measurement system (can be applied to humidity as well) ... 43
Figure 47 3D model of the plasma cutting machine and proposed sensor implementations ... 43
Figure 48 3D model of a sensor setup designed to determine the angle to which the bridge side components drift ... 44
Figure 49 3D model of placement of the circuitry (microcontroller + gyro-accelerometer + radio module + battery) on the plasma torch ... 45
Figure 50 3D model of two proximity sensors capable of detecting the thickness of the plate and, respectively, the x-axis position of the torch. ... 46
Figure 51 3D model of the sound capturing system formed out of a pole, two microphones, a data acquisition device and a laptop ... 46
Figure 52 Micrometre - google picture ... 47
Figure 53 Zig zag plate ... 48
Figure 54 3D SketchUp model of the sensor testing prototype setup – perspective 1 ... 52
Figure 55 3D SketchUp model of the sensor testing prototype setup – perspective 2 ... 53
CHAPTER I RATIONALE
‘Industry 4.0’, also called the fourth industrial revolution is a term for the automated manufacturing technology and data exchange via Internet of Things (IoT), its name being given by the German government. A response to the German ‘Industry 4.0’ was given by the Dutch: ‘Smart industry’. According to the document ‘Smart industry – Dutch industry fit for the future’, the definition of Smart Industry comes as follows: ’Smart Industries are industries that have a high degree of flexibility in production, in terms of product needs (specifications, quality, design), volume (what is needed), timing (when it is needed), resource efficiency and cost (what is required), being able to (fine)tune to customer needs and make use of the entire supply chain for value creation. It is enabled by a network-centric approach, making use of the value of information, driven by ICT and the latest available proven manufacturing techniques.’[1]
Because currently Voortman Steel does not have predictive breakdown/maintenance functions on their numerous pieces of machinery, but they have communication and sensors, they would like to upgrade and get the necessary qualification for making a ‘Smart Industry’.
The project focuses on exploring and realizing ‘Smart industry’. By ‘Smart industry’ Voortman Steel, the collaborating company, means an ensemble of industrial steel processing machineries that are fully automated, in the automation being included the predictions of upcoming breakdowns of particular components and of machineries as a whole, real time diagnosis, real time measurements of different mechanical, electrical, physical aspects with influence on the proper functionality of the machineries, recording of data, data analyzation and interpretation. To achieve that level of automated constant monitoring, a sensor system has to be designed. Because the sources of breakdowns are various, many parameters have to be measured and the output has to be observed, analysed, and interpreted, the results being able to possibly influence the future design of t he machinery itself. For that, research, trial and error tests and data analysis have to be done.[2]
Due to the fact that switching a whole industry department to ‘Smart Industry’ is a process that would require more than one person’s work during twenty weeks, the focus was set eventually on one of the matters of interest from one of the machines found at Voortman Steel Machinery. The machine is the V304 plasma/oxy-fuel cutter which processes steel plates and the issue in cause is the displacement of the bridge which impacts negatively the modelling process. The study case was chosen by the engineering managers from Voortman’s Steel due to their current needs. Mr. Martin Alberink, the machine vision engineer from Voortman believes that the plasma cutter’s issues are tougher to cope with, than the other machines’ due to the fact that it is often the case that the machine misbehaves, the reasons remaining unknown. It was agreed that although solving one problem of one piece of machinery is not classified as upgrading to ‘Smart Industry’, but contributing to it, finding a relatively inexpensive way to cope with the proposed matter would satisfy Voortman project-wise.
Voortman’s only requirements for the project were to use low-cost sensors, because sensors that are too expensive increase the price of the machine as well, and making the machine inaccessible to the client is not something the company wishes, and obtaining a bridge movement detection accuracy of 50 microns, which according to Mr. Alberink, it is the threshold from which visible errors appear on the steel plate when symmetrical cuts are made.
The research question of this project together with its sub questions are as stated below:
How can we detect with an accuracy of 50µm the difference between the position of the left extremity of the V304 plasma cutting CNC machine’s bridge with respect to the axis on which it moves regularly (along the rails) and the position of the right extremity of the bridge with respect to the same axis by using sensor systems that fit into a budget of a few hundreds of Euros?
- Is there a straight forward method of measurement or complex calculations will have to be derived in order to obtain the desired quantity? If so, what would those be?
- What is the best sensor or what are the best sensors to be used for this project’s purpose? - How can the sensor system be installed accurately enough to obtain reliable information? - How can the reliability of the sensor system be validated?
Figure 1 Voortman V304 plasma cutting machine [3]
A relatively imprecise (non-proportional) tri-dimensional model of the Voortman V304 plasma cutting machine was created for simulations and ease of illustration. This replica can be observed in Figure 2.
Figure 2 Voortman V304 plasma cutting machine 3D SketchUp model
The problem in discussion falls on the component named ‘bridge’ which is the solid metal beam that unites the two opposite machine body parts and that enables the plasma torch to move longitudinally. An illustration of the bridge from the 3D model designed in Figure 2 can be observed in Figure 3.
Figure 3 Voortman V304 plasma cutting machine bridge 3D SketchUp model
With the aim of making a better problem statement, a sketch of the machine viewed from above was created as it can be observed in Figure 4.
Figure 4 Sketch of the view from above of the Voortman V304 plasma cutting machine
The problem can be summed up in one question: How can we detect with an accuracy of 50µm the difference between the position of the left extremity of the bridge (A) with respect to the y axis and the position of the right extremity of the bridge (B) with respect to the same axis by using sensor systems that fit into a budget of a few hundreds of Euros.
Judging by the fact that the bridge is perfectly perpendicular on both left and right body parts, they can be regarded as one symmetrical solid object that rotates around the centre (O) when the motors that power the two body parts are out of sync, or when other deviations occur. Due to that reason, the extremities of the bridge (A and B) could be regarded as points on the extremities of the body parts since the effect of a deviation of the bridge would be seen on the body parts too.
The only constraint of the project is the budget, which obliges the use of sensors with basic functionality, which means that smart measuring ways have to be developed in order to compensate for the absence of certain features such as long range, durability, software/hardware ease of use, etc.
A side requirement is to have a response rate fast enough to ensure the proper working of the sensor system. However, this feature will not eliminate from the start concepts that give reasonably good performance, but that cannot fulfil this
A sensor system capable of such monitoring would have a direct effect on both, the machine and the product. When the extremities of the bridge are not in the same position the modelling of the steel plate suffers errors and the product becomes faulty, in which case the product either has to be made again or the client will not be satisfied. By having irregularities in the component positioning, the wear increases and problems with bearings or even motors may appear. The current situation is not quantified in money loss or material wasted, so coming up with a number to indicate efficiency improvement is not possible. However, such a system could also be used to warn and prevent breakdown or contribute to a future more complex
CHAPTER II
SITUATIONAL AND THEORETICAL ANALYSIS
The scope of the project was changed multiple times and, in consequence, much research and many conceptual designs were realized. All of them can be found in Appendix A. In this chapter only the research relevant to the bridge alignment is being discussed.
I. Principles
1) Optical noncontact distance measurement
All of the sensors that use any of the optical noncontact distance measurement principles described below have to point perpendicularly towards the target object in order to find the distance between them. In this paper, the measurement methods are divided into two major parts: y-axis measurement and x-axis measurement (see Figure 4). Direct optical noncontact Y-axis measurement would require sensors to be placed at the ends of the rails, pointing towards the body parts. This configuration requires a sensor range of ~4m-6m and an accuracy of less than 50µm. The second major category of measurement methods is x-axis measurement, which requires knowing accurately certain machine dimensions (e.g. bridge length), the use of an auxiliary flat screen to which the sensor that is placed on the body part will point. This configuration is necessary due to two reasons: machine’s flatness is not reliable and the body part is moving on rail with the length about three times as big as its length, which guarantees that if only one sensor is placed in front of the body part, there will be times when the body part will not be in the sensor’s detection area. For x-axis measurement, the sensor’s range is flexible since the screen’s placement can be adjusted to fit the needs. The accuracy that is needed is 50µm from which the flatness error of the screen has to be subtracted. For instance, if there will be a plate with a flatness that varies within 20µm, the sensor will have to have an accuracy of ≤(50µm-20µm), so ≤30µm.
Regarding optical noncontact distance measurement methods, two things have to be taken into consideration: the range and the accuracy with which the desired quantities can be measured by using the respective method.
a) Intensity-based sensors
Figure 5 ‘Illustration of a fibre optic intensity sensor for distance measurements. (a) Two-fibre sensor, showing the object and the operational principle (b) Various geometries for fibre bundle sensor heads’ [4]
b) Triangulation sensors
Figure 6 ‘Principle of optical triangulation sensor. The unknown distance, D, is determined from the known distances 𝐸,𝐹 and the measured value of G—the distance to the pixel in the position sensitive detector (PSD) recording the image of the laser spot on the measured object’ [4]
c) Time-of-flight sensors
Figure 7 ‘(a) Pulsed time-of-flight measurement showing three regimes where the time of flight is (i) much longer, (ii) similar to, and (iii) shorter than the pulse width. For nanosecond pulses these cases correspond to distances (in air) of >50 m, a few meters, and <1 m, respectively. (b) Intensity modulated time-of-flight is an alternative method suitable for distances in range (ii). The phase shift is indicated by the double-headed arrow.’ [4]
Page 14
d) Confocal sensors
Figure 8 ‘Principle of fibre optic monochromatic and polychromatic confocal sensing. (i) Monochromatic confocal sensor with an object at the image plane. (ii) Monochromatic confocal sensor with an object displaced from the image plane. (iii) Polychromatic sensor where the image plane position varies with wavelength and a different wavelength will satisfy the confocal geometry for each object position’ [4]
e) Interferometric sensors
Figure 9 ‘Basic setup for WLI and typical interference pattern as a function of the reference arm mirror movement. A low-coherence source is used, in this case a fibre-coupled 1.5 µm super luminescent diode with 60 nm spectral width. The interferogram shows an oscillation period of half the wavelength, and width corresponding to the coherence length of the source’ [4]
f) Doppler sensing
Figure 10 ‘Demonstration of fibre-optic-assisted Doppler velocimetry. (a) The experimental setup, using a pigtailed coherent 1.5 µm laser source and a moving mirror. (b) Detector output measured by an oscilloscope for a motion of 100 µm/s. (c) Detector output measured by an electronic spectrum analyser for a motion of 100 µm/s (black), 50 µm/s (blue), and 25 µm/s (red)’ [4]
Figure 11 sums up the optical noncontact available options in one graph. Since the optical noncontact sensors would have to be used for measuring distances up to ~6m, which is the length of the rails, along which the two body parts are able to move, this project’s point of interest would be at the intersection of the µm line from the resolution axis with the m line from the absolute working distance axis. In that area, only the laser interferometry is found, since it reaches ranges of a meter with micrometric resolution, but the real range of such devices does not cover a 6m range. Moreover, this is a method that implies the use of very expensive sensors (>10000 EUR), which is out of scope.
Figure 11 ‘Comparison of typical resolutions and working distances of the optical distance measurement techniques discussed in this article. The entry for triangulation is divided into a filled region, corresponding to measurement from a single observation point, and a shaded region for multiple observation points’ [4]
It can be concluded that regular optical noncontact measurement methods are not good enough to measure distances with an accuracy ≤50µm. One thing that I do not agree with from the graph in Figure 11, however, is the resolution of the triangulation sensors which can actually reach the µm line. An example of triangulation sensor that contradicts the graph can be found later in this report, its specification table being shown in Figure 17. The linearity is not limited by the ‘mm’ threshold and goes to the ‘µm’, since linearity values such as a few tens of microns are presented. On the plot, that would appear as an ellipse that is longer in the bottom left side. The plot could have been made when the respective performant sensors were not on the market, and therefore, the error that was spotted can be regarded as due to modernization in time.
However, these sensors are also out of scope due to their regular high price (exceeding 500 EUR), which is also the case for confocal sensors.
The optical sensors that can measure with a resolution good enough for the project are, therefore, mostly unusable due to the constraints of the project.
2) Inductive distance measurement
Inductive proximity sensors are noncontact pieces of electronics that are able to detect the position of metal objects. The nature of the metal influences the range of the sensor, iron and steel being better for longer ranges. A magnetic field is created by a coil and an oscillator. The oscillations are damped when a metal is in the proximity. A subdivision of the phenomenon is known under the name of Eddy-Current and there are sensors named ‘Eddy-Current sensor’ or ‘Eddy-Current Inductive Sensor’ on the market.
The aforementioned sensors are considerably cheaper (even less than 40EUR) than the optic ones that reach the desired accuracy. The inductive sensors have low range but good accuracy and repeatability as it will be seen later in this paper. These sensors seem to be able to perform high accuracy measurements, so the range problem will have to be solved in order to use them.
The advantages of the inductive and Eddy-Current sensors are that dirt or other materials in between the sensor and the metal that has to be detected will not interfere in the measurement, the low price and the robustness.
It is noticeable that inductive measurement would not fit into the Y-axis category due to the 4-6m range requirement. However, because X-axis measurement is not range dependant and the accuracy terms are met by the inductive sensors, the configuration described in the optical measurement introduction for x-axis measurement methods occurs here as well.
Similar to the inductive sensors, there is the capacitive measurement which, unfortunately, has to be excluded from a first sight due to its ineffectiveness in dirty conditions.
3) Mechanical distance measurement a) Draw-wire sensors
A good description of these sensors is given on the µƐ (micro-epsilon) website, from which such sensors can be bought as well: ‘Draw-wire sensors measure linear movements using a highly flexible steel wire. The wire is wound on a drum, whose axis is coupled with a potentiometer or an encoder. The end of the wire is attached to the measurement object. A rotating movement of the drum is produced when a change in the distance of the measurement object from the sensor occurs. This rotating movement is converted to an electrical signal and output using an encoder or potentiometer.’[16]
The accuracy and range of these sensors are precarious, since neither is fitting the requirements.
However, assuming that the sensors would be performant enough for a Y-axis measurement, the configuration would be as follows: two such sensors would be attached to two poles placed at the two ends of one rail (each sensor on one pole), the wires would be connected to the body part that moves on the respective rail, the connection points being at the same level from the ground and the sensors too. Knowing the standard sum of the values of the two sensors would mean that any rotation of the bridge will be noticed by a change in that sum. Further, the desired values could be calculated. A concept design using these sensors will be later in this report in Figure 28.
b) Dynamometer
As it is commonly known, the elastic force can be used to determine the change in length of a spring if the elastic constant of the spring is known. Using spring push buttons that get activated by bridge displacements and measure the elastic force accurately could in theory measure separately the positions of the body parts and enable us to obtain the difference. However, an accuracy of 50µm is next to impossible to achieve with such a mechanical device, since there will be too much deviation from the spring.
The downside of the mechanical methods is the wear that occurs unavoidably and the poor accuracy that can be achieved. It is a rather limited method.
Configurations for this type of sensor could be developed in Y-axis as the system described for the draw-wire sensors which would measure the strain of the cable connected to the body part, and for X-axis measurement, one long pushbutton or more smaller buttons could be placed parallel to the rail and very close to the body part, so as to be pushed by the machine when in deviates from its track. But, again, this method is highly unreliable and the wear is inevitable.
4) Ultrasonic
Figure 12 Ultrasonic sensor working principle - the ultrasound hits an object and it is reflected back to the sensor
The ultrasonic sensors are composed out of an emitter and a receiver. The emitter sends ultrasounds that bounce back to the sensor when they hit an object, and that are further taken by the emitter, which enables the sensor to calculate the time between sending and receiving. Since we know that distance is time multiplied with the speed, and since the sensor uses sound, which has known speed, the distance can be calculated.
With the Y-axis measurement configuration discussed in the previous chapters, the goals of the project would not be achievable with such a sensor due to the range requirement, which cannot coexist with the accuracy of 50µm. For the X-axis measurement configuration, such a sensor could be implemented. The only drawback is that ultrasonic sensors, despite being mostly inexpensive, when it comes to accuracy as such, the prices rise more than the budget of the project can support.
5) Radio
RADAR systems have been used for years to detect distances to moving or stationary objects. However, the radio detection is mostly meant for very long ranges, returning errors in lengths of the order of tens of metres. The multitude of options for this type of sensors is vast, but an accuracy such as the required 50 microns is out of discussion, so radio detection is not suitable for this project’s goal.
6) Visual inspection
The Kinect webcams have been used for years to detect movement smartly and sometimes to detect distances. However, the camera presents bad linearity when experimenting measurements along a measuring tape and its accuracy is nowhere near the micrometric one.
If it were not for the high accuracy that is desired and for the bad linearity of the device, the Y axis measurement setup could be made with the camera.
With normal cameras, image processing software and code bands placed on the body parts, the exact position of the body part could be detected, but not with an accuracy of 50µm.[15] Such high accuracy in regular webcam measurement is not possible. When looking for sensors suitable for the project, the main focus has to be the accuracy. The range can be disregarded for the moment since there are geometrical or mechanical ways (extra hardware) of measuring longer distances.
II. Sensors
From the discussed principles it can be concluded that the sensors that can reach an accuracy that fits the requirements should follow one of the following principles: confocal sensing, interferometry, triangulation, ultrasonic, inductive or capacitive sensing.
1) Confocal sensing
Figure 13 ConfocalDT IFS2402 - confocal miniature sensor from Micro-Epsilon [5]
Figure 14 Specifications of the ConfocalDT IFS2402 sensor[5]
As it can be observed from the specifications of the ConfocalDT IFS2402 sensor from Figure 14, and more specifically, from the values of the linearity, resolution and range, this is an extremely accurate sensor with a very short range, meaning that it could be used in the X-axis configuration.
Its price is not specified on the website, but is expected to be too high for the budget due to the fact that the company has to be reached to find out the price.
2) Interferometry
Figure 15 zygo® MK-2 Interferometer [6]
An example of interferometer can be seen in Figure 15. This device costs 9200 EUR and it is second hand, needless to say that it is completely out of scope. Moreover, the interferometer is not entirely independent, while it is rather a tool device and not a sensor.
3) Triangulation
Figure 16 optoNCDT 1320 // Compact Laser Triangulation Displacement Sensor from MICRO-EPSILON [5]
Figure 16 depicts a short range laser triangulation displacement sensor that measures with extreme accuracy as it can be seen in Figure 17 (10-50mm range, 12-60µm linearity, 1-5µm repeatability, 14bit signal processing).
Figure 17 List of specifications of optoNCDT 1320 laser triangulation displacement sensor [5]
An inquiry was made to find out the prices of the products from MICRON-EPSILON that have an accuracy of ≤50µm. The answer was as expected disappointing due to the high pricing which exceeded 500 EUR.
4) Ultrasonic sensing
Figure 18 RPS-412A High Accuracy Sensor - Ultrasonic sensor from Migatron Corp.[7]
RPS-412A is an industrial ultrasonic sensor with a stainless steel casing and that can measure between 3 and 16 inches (76.2mm-406.4mm) with an accuracy of the highest value between 50µm and 0.05% of the range, at 25oC.
5) Inductive sensing
Figure 19 InduSENSOR LVDT - inductive displacement sensors from MICRO-EPSILON [5]
Figure 20 List of specifications for InduSENSOR LVDT inductive displacement sensors [5]
As it can be observed from the specifications of the sensors from Figure 19 in Figure 20, there are models that can achieve the desired accuracy, while the range is very short (1-25 mm range, 3-300 µm linearity, 1-5 kHz response frequency). The price for such a sensor starts from over 500$. However, in this area there are also regular sensors that are less durable and with worse casing, but that are more financially accessible.
Figure 21 GX-F/H type inductive sensors [8]
Figure 22 Specifications of GX-6 inductive sensor[8]
As it can be seen from Figure 22 the desired accuracy can be reached (<20% Hysteresis, 1.3mm reliable range, 40µm
repeatability), while the range is really small. This sensor’s price is below 50$, hence very accessible. However, there is a huge problem with this sensor: its output is digital so the best that could be done with it would be to warn when the 50µm distance is
It is less often met, but there are also analogue inductive sensors on the market which are more expensive than the switch digital ones, but which would be suitable for the project since they can be as cheap as 120EUR. To quote from the
correspondence with a sensor industry representative, there are sensors that start from 116 EUR with a range of 0.5-2mm, with an accuracy of +/-8µm, that can work with 0-10V and that has a 3m long cable. One of these sensors is BAW001L or BAW M12MG2-UAC20B-BP03 whose datasheet can be found in Appendix B.
Additionally, to obtain the output of such a sensor, a data acquisition device should be used.
The least expensive DAQ device I could find on internet can be seen in Figure 23. Its price is 29$, and there could be up to four analogue sensors connected to it simultaneously. If Voortman Steel has a PLC analogue processing card, it could be used instead of a DAQ device.
Figure 23 DI-145 DAQ device from DATAQ INSTRUMENTS 6) Capacitive sensing
An example of capacitive sensor is the capaNCDT 6200 from Figure 24. As it can be observed from its specifications from Figure 25, this is a sensor that can give more accurate measurement than all the previously presented ones, this one not being the best there is on the market.
The huge disadvantage is that it is simply impossible to use this sensor in exposed industrial environments due to the fact that dirt destroys the measurements. It could, however be a good sensor to use internally, where dust cannot reach.
Figure 25 Specification table of the capaNCDT 6200 capacitive sensor [5]
After studying the sensors and principles presented in this chapter, it can be concluded that the only sensors worth investigating are the inductive sensors. The draw-wire sensors, despite not being suitable, will appear in an initial concept description due to a premature concept designing, before discovering its limitations.
CHAPTER III CONCEPTUAL MODEL
When there is a difference between the y axis positions of the two extremities of the bridge, due to the fact that the bridge is solid enough to not be able to bend, the edges of the body parts from will form an angle with the y axis. That angle will be congruent with the angle formed by the bridge with the x axis as it can be seen in Figure 26. Measuring one of these two angles could not only tell us if there is a difference between the two extremities’ position with respect to the y axis, but also the value of this difference through trigonometrical calculus.
Figure 26 Congruency between the angles formed by the bridge and the x axis and, respectively, the angle formed by the body parts with the y axis
Another possibility to explore is measuring separately the lengths from a fixed point on the y axis to the points of interest A and B (from Figure 4) and subtracting one from another to see what its absolute value is. On the same principle measuring the distances on the y axis of one body part from two opposite fixed sides could show both the position of the body part on the y axis and the angle at which it is inclined (if it is) by getting a bigger difference between the two measured values when there is an angle (≠0o) present.
Theoretically, a way of obtaining the value of the y-axis difference between the two extremities of the bridge is by direct measurement of the two separate parts so as to subtract one value from another and obtain the desired value. Another way is to derive that value by using known lengths of the machinery, which will have to be manually accurately measured unless the values are specified in the machine’s manual, and by measuring x-axis displacement with low range but very accurate sensors. Assumptions such as ‘The bridge and the two body parts form one solid component that does not bend, so when there is desynchronization between the two parts’ motors, the component makes a rotation movement and, therefore, the distance that needs to be measured will be the difference between the two extremity points because this will return the biggest value possible.’ and ‘The rotation of the component will be around the bridge’s centre.’ are made.
1) Y-axis measurement
As previously discussed, managing to measure the distances from both ends of one rail to the body part corresponding to that rail would allow us to obtain the displacement in angle and distance. Even simpler, measuring from the same side of the rails the distances from the rail edges to the faces of the body parts that correspond to the rails would be the most straight forward way of obtaining the difference between the two y-axis alignments. Aligned poles would have to be put in front of the machine
as shown in Figure 27. To the machinery, a flat piece of material will have to be attached, at which, sensors that are attached to the pole will point to find the distances.
Figure 27 3D model of the plasma cutting machine and theoretical sensor implementations
Realistically, this is possible only if we exclude the 50µm accuracy requirement (which is half of the thickness of a hair). Such exclusion is not possible, this being the most important factor of the project.
Because of the fact that the bridge, together with the two body parts move along a distance of a few metres, while the sensors measuring the distances will be fixed, the range of the sensors will have to cover all the possible positions of the component, which is impossible with a requirement of 50µm accuracy, not to mention the ‘relatively cheap sensors’ requirement. Normally such measurement accuracy comes along with very little range. As explained in the principle chapter, there is no way of achieving the straight forward way of measurement and respect the requirements of the projects in the same time. In other words, the accurate Y axis measurement at all times is not possible, so the X axis measurement has to be studied.
2) X-axis measurement
In Figure 48 a system capable of measuring the angle at which the bridge of the plasma cutting machine shifts is illustrated. An elastic cable or spring is connected to the machine and to a fixed pole in such way that the cable is parallel with the track or the ground. At the end that is closer to the pole, a light non-transparent flat object that cannot move along the cable but which will be loose around the cable in order to keep it perpendicular to the ground plane. A second pole would be placed in front of the flat object with a short range but high accuracy proximity sensor pointing at the object perpendicularly. By knowing the exact distances and angles when the bridge is in the right position and that are also drawn in Figure 48, the new distances and, implicitly, the new angle can be calculated when the bridge shifts.
Let there be a triangle ABC where A is the point of connection between the cable and the pole, B the point where the proximity sensor’s receptor is and C the point where the sensor’s range intersects with the flat object when there is no drift. All the lengths of the ABC edges and the angles can be measured and set as standards. Deviations from those values will imply a drift of the bridge. Let replace C when there is a drift. The interest will be the difference between the angles BAD and BAC. By using simple geometry we obtain the general formula AD=sqrt((BD-BC)^2+AC^2), hence BAD=arccos((AB^2 + AD^2 - BE^2)/2*AD*AB) from which we can subtract the known angle, BAC.
Figure 28 3D model of a sensor setup designed to determine the angle to which the bridge side components drift
This idea is most likely viable, but the durability is questionable due to its mechanical nature. The elastic item will eventually present signs of wear and it is not desirable.
Another way of measuring x-axis displacement mechanically, is by attaching to one body part two pressure buttons or
dynamometers that are able to accurately measure the contraction of the spring inside the sensor (Δl=Fe/k), at extremity points on that body part but at the same level (same distance to the ground), when getting in contact with a panel placed in front of the respective body part in a parallel position with the body part. During research, however, it was next to impossible to find such a sensor to be suited for the project.
The downside of using such a system is again the wear that will eventually occur due to contact.
Moving on from contact mechanical systems that wear down after a period of time, while taking into account the findings from the ‘Measuring principles’ and ‘Sensors’ sections, the only low cost, tolerant to industrial environment and accurate enough option seems to be inductive sensing. Placing on one body part just as described in the case of the pressure buttons, inductive sensors, a sensor system such as the one illustrated in Figure 29 and Figure 30 can result.
Page 29
Figure 29 3D SketchUp model of the inductive sensing setup: two inductive sensors placed in extremity points on one face of the machine, pointing at a metal flat, parallel to the machine plate
Inductive sensors
Figure 30 3D SketchUp model of the inductive sensing setup: two inductive sensors placed in extremity points on one face of the machine, pointing at a metal flat, parallel to the machine plate (another perspective)
This setup would allow measuring the displacement of the two points where the sensors are mounted. By knowing fixed distances such as from a sensor to the other and lengths of the components, the difference between the Y-axis positions of the two side points will be calculated geometrically. For that to happen accurately, the metal plate should be made of steel or iron (due to sensors’ affinity towards ferrous materials) and it should be as flat as possible, for a better accuracy, while being completely parallel to the rail of the corresponding body part.
Figure 31 Geometrical explanation of the determination of the desired 'd' distance, where the H shaped drawing represents the two body parts united by the bridge, viewed from the top, and where the line uniting A' and B' is the metal plate, A is the point where one of the proximity sensors will be, and, respectively, B will be the location of the other sensor.
Inductive sensors
niet gevonden.Figure 15), measuring tape, plasticine (for making pattern shapes to use for accurate placement of the objects and
Page 31
for measurement), the flat iron/steel plate and it is likely that other items such as vice, accurate protractors and inaccurate, inexpensive laser rangefinders will be used for fixations and measurement verifications. The need for other tools might occur. Other sensors such as accelerometers or temperature sensors and, eventually, a cooling system might be included too, in order to ensure a good measurement environment.
To conclude, the best way of measuring the difference in placement on y-axis of the lateral parts of the bridge is by deriving from the x-axis displacements that will be measured with the use of relatively inexpensive inductive sensors. Depending on how good the accuracy of the outcome will be, a predictive maintenance system will be developed. The system will be able to further contribute to predictive maintenance in combination with other future monitoring systems.
CHAPTER IV RESEARCH DESIGN
One of the research sub-questions focused specifically on the validation method. In order to validate that the sensor system is functioning properly, tests could be made either by placing a second identical setup on the other body part of the machine and comparing results which might raise budget issues, or bridge deviations may be induced by desyncing the two motors that power the bridge if it is allowed to do so. An already desynced bridge plasma cutter might be available and the setup could be tried on that machine. Manual measurements might also be viable for validation by measuring with a micrometre measured distances and comparing them. A combination of two or more methods from the previously mentioned ones could be used for validation.
The budget is supposed to be a few hundreds of euros, so finding suitable testing tools might be a problem, but renting/borrowing tools would be a good thing to do.
The durability of the system is equal to the durability of the sensors since they are wear-free, being non-contact, and the system being as simple as two sensors and one metal plate. On the Balluff website, where the BAW001L inductive proximity sensor was ordered from, the ‘mission time’ is specified to be 20 years, with a MTTF (40°C) (mean time to failure) of 640 years, but which is mentioned to not be a life expectancy guarantee. [17] The numbers, however, are overwhelmingly high, and would suit the durability of a plasma cutting machine. In case that after twenty years of functionality the sensors breakdown and the machine still works, the low price of the sensors would make their replacement less of a challenge.
The validation that has been done was the sensor accuracy validation which was tested with the use of a micrometre, the BAW001L sensor and a data acquisition device. The initial setup plan was as the ones described in Appendix C – Experimental Procedure and it is still the recommended setup. The setup on which the real life experiment was carried out has been adapted to the materials that were provided, which were as follows: the requested data acquisition device, one of the requested BAW001L sensors and one different micrometre as depicted in Figure 32.
Figure 32 Provided micrometre
In order to adapt to the circumstances, a piece of metal was attached to the tip of micrometre’s pin to have a minimum dimension detectable surface (see datasheet in Appendix B). The sensor was connected to the DATAQ device as such: the black wire of the sensor (BK) to the CH1+ slot of the DATAQ, the blue wire of the sensor (BU) to the CH1- slot of the DATAQ and to the GND from a power supply able to provide 24V, the brown wire of the sensor (BN) to the positive terminal of the 24V power supply. The software for the DATAQ requires 32 bits Windows. Due to having Windows 7 on 64 bits, a virtual machine was installed, namely Oracle VM VirtualBox, on which Windows 7 32 bits has been installed, along with the free software applications from DATAQ that can be found on the DATAQ website under the ‘Software’ tab from the product, which is DI-145 in this case, a trial version of Microsoft Office to sync the data with Microsoft Excel and Microsoft Visual Studio to be able to program the sensor. The DATAQ device was connected via USB to the laptop and from the VirtualBox it had to be recognized by right clicking on the USB icon (Figure 33) and selecting the ‘Unknown device’.
Figure 33 USB icon in Oracle VM VirtualBox
The metal attached to the micrometre was placed in front of the sensor in such way that it touched the piece of material that is placed on the tip of the sensor by default to get rid of the blind spot, and in such way that in the application ‘Hardware Manager’, after calibrating the display from voltage to distance in millimetres with the upper limit 10V= 1.49925mm (because in the appendix B it can be noticed that for 1mm the sensor outputs 6.67V) and the lower limit 0V=0mm as seen in Figure 34, the displayed value was 0.0000mm.
Figure 34 Calibration in software
Then, gradually, the metal was pushed towards the display of the micrometre and the values of the software display and the micrometre were read and recorded simultaneously.
CHAPTER V RESEARCH RESULTS
To make a short revision, the research showed that measuring unwanted movements of the bridge with inductive sensors is the best way of achieving the goal of this project. From the research, it was concluded that the sensor BAW001L, whose datasheet can be found in Appendix B is the most suitable for the requirements of the project: accuracy of ≤50µm and price as low as possible. The sensor could be more performant, but that would mean that the price would increase which would not be desirable because the price is elevated enough already. The sensor data has to be processed with a data acquisition device as the one that can be observed in Figure 23. Analysing its specifications, it can be observed that it can sense voltages of 10V, which fits the sensor analogue output, and that it has a resolution of 10 bits, which means that the smallest value that will be read by the DAQ will be the range divided by 210. Therefore, a range less than ~50mm would enable the DAQ to process values of
50µm. In our case, the sensor will have a range of 2mm which will means that the resolution of the DAQ will be able to handle the micrometric values.
CHAPTER VI EXPERIMENTAL RESULTS
The experiment that tested the sensor’s accuracy returned the following results:
Figure 35 30 distance values taken with the micrometre and with the sensor software, on the x-axis being the measurement number and on the y-axis being the value of the reading in mm (the table of values can be found in Appendix D)
As it can be seen, the two lines overlap perfectly until the distance reaches 1.5mm, where the sensor reaches its range and the micrometre is still able to take measurements reliably.
The values of read with the micrometre may present reading errors due to rough approximations. The micrometre has divisions of 10µm, but when the needle was in the middle of the division, the value was read as µm, so the error is ±5µm from the reading.
The highest difference between the measured and sensed distance was 21µm. CHAPTER VII
CONCLUSIONS
The sensor BAW001L behaved exceptionally well, returning accurate values, which, although it could not be tested due to lack of materials, it promises that with an identical sensor the concept is completely viable. The 50µm accuracy and the low-cost requirements were met and the system promises to be a solution to Voortman Steel Machinery’s bridge displacement issue.
CHAPTER VIII RECOMMENDATIONS
A second sensor would have to be connected to the DATAQ in the same way the first one was, but to the CH2- and CH2+ slots instead of CH1- and, respectively, CH1+. The proposed testing procedures should be carried out instead of the one that was actually done due to better stability and reliability. When carrying out experiment no.2 use Figure 31 to obtain the angle and check with a protractor too and compare. The Hardware Manager application is a short term measurement graphic app that can return output as in Figure 36.
The placement of the sensors on the machine is out of scope, but, as a recommendation, it can be done accurately either mechanically with different hinges, nuts and bolts, or it could be incorporated in the body part of the machine in future machines, directly from the factory.
The metal plate should be stainless steel/steel/iron and it should be placed at a distance from the sensors equal to half the range limit of the sensor in order to make sure that the sensor will be in range no matter what the rotation of the bridge will be. The exact dimensions of the bridge would have to be measured or found out.
0,000 0,500 1,000 1,500 2,000 2,500 3,000 3,500 0 10 20 30 40 Software data Micrometre data
Figure 36 Graph of the sensor output in Hardware Manager
REFERENCES CITED
[1] FME-CWM, Chamber of Commerce, Ministry of Economic Affairs, TNO, VNO-NCW, 2014, [online article], ‘Smart industry – Dutch industry fit for the future’, available from: http://www.smartindustry.nl/wp-content/uploads/2014/07/Opmaak-Smart-Industry.pdf
[2] Mr. Martin Alberink, Voortman Steel Group, [description paper], ‘Assignments for internships and final projects – Designing a monitoring system for steel processing machines’
[3] Voortman Steel Machinery website, [online], available from: https://www.voortman.net/en/products/machinery/plateprocessing/v304
[4] Garry Berkovic and Ehud Shafir, 2012, [scientific article], Optical methods for distance and displacement measurements [5] Micro-Epsilon website, [online], available from: http://www.micro-epsilon.com/index.html
[6] Dipl. Ing. Michael Koch, [online], Astro Electronic, available from: http://www.astro-electronic.com/interf.htm [7] Migatron Corp., [online], RPS-412A High Accuracy Sensor, available from: http://www.migatron.com/high-accuracy-sensor/
[8] Digikey website, [online], available from: http://www.digikey.com/
[9] Pentronic, [online], ‘Temperature sensors – overall explanations’, available from: http://www.pentronic.se/home/temperature-sensors.aspx
[10] SHARP, [datasheet], ‘GP2Y0A21YK proximity sensor datasheet’, available from: http://www.sharpsma.com/webfm_send/1208
[11] Rosa Ciprian, Brad Lehman, 2009, [scientific article], ‘Modeling Effects of Relative Humidity, Moisture, and Extreme Environmental Conditions on Power Electronic Performance’, available from:
http://www.ece.neu.edu/groups/power/lehman/Publications/Pub2009/2009_9_Ciprian.pdf
[12] Branco, Carlos Moura, Ritchie, Robert O., Sklenička, Václav, 1996, [book], ‘Mechanical behaviour of materials at high temperature’, Springer, ISBN 978-0-7923-4113-0.
[13] ESAB knowledge center, 2013, [online article], ‘How loud is a plasma cutting torch’, available from: http://www.esab.ca/ca/en/education/blog/how-loud-is-a-plasma-cutting-torch.cfm
[14] Stone Nick, 2009, [scientific article], “Direct to Plate Photogravure: Catching Up With the Past.”
[15] Roger Y. Tsai, 1987, [scientific article], ‘A Versatile Camera Calibration Technique for High-Accuracy 3D Machine Vision Metrology Using Off-the-Shelf TV Cameras and Lenses
[16] Micro-Epsilon website, [online], available from: http://www.micro-epsilon.com/displacement-position-sensors/draw-wire-sensor/index.html
APPENDIX A
DISCARDED RESEARCH AND CONCEPTS
The focus of the project fell on the plasma cutting machine, although in the beginning a broader approach had to be done due to not knowing exactly what will be the machine of focus. For this reason, the student’s research phase was done in stages, each one ending after a progress meeting where updates were brought by Mr. Alberink, who is the contact between the author of this paper and Voortman Steel. Due to lack of complete technical knowledge regarding the machinery, information came to the student in waves and many concepts have been made and later discarded (at least project-wise). Initially, the research focused on general measuring methods for determining temperatures that are too high for usual tiny sensors. The conclusion was that contact sensors such as thermocouples could measure up to 2300oC, which was not enough, but it was the highest value a
contact sensor could measure, which meant that a sensor capable of remote measuring was needed.[9] Industrial IR sensors were found and recommended. The options for monitoring temperature in all the areas of the electric cabinets were discussed. Two concepts were developed: one for measuring temperature of the punchers in a perforation machine and another one for oil contamination detection.
Figure 37 Puncher temperature measurement system sketch made by the student
The design in Figure 37 was made considering that the highest temperature is reached by a puncher at the end of the punch (position and time-wise). The proposed solution was to use one industrial IR temperature sensor or pyrometer to measure the temperature of the puncher right after it perforates the steel plate, whose temperature could also be measured in the meantime by the same sensor. Above the sensor, there is a metal Petri dish with a hole in bottom big enough to not block the sensor measurements, but smaller than the bits of steel that fall from the plate perforation and another hole in the side wall, through which the bits of steel fall due to inertia when the servo motor spins the dish.
A better design than the one from Figure 37 would have been if instead of using moving parts such as servo motors, the Petri dishes could be inclined planes, so the gravity would act on its own.
Figure 38 Oil contamination measurement system sketch made by the student
Figure 38 depicts a concept for keeping track of the oil contamination, since research showed that oil contamination is one of the most common factors that lead to breakdown, due to the fact that impurities are turned into fine particles when they get in contact with the drills, so the air becomes stuffed and clogs the air conditioning system that cools down the machinery. Imagining the cutting oil being kept in a cylindrical tank of known mass, height, section area and, implicitly, volume, placing on top of the oil a light, opaque, floating lid that is immiscible with the oil, that has a reflective ratio between 18% (gray paper)-90% (white paper) and that has the sizes of the interior of the tank, and an accurate proximity sensor on the inside of the lid that is covering the tank would make it possible to measure the volume of the oil. [10]
Knowing the volume of the oil is helpful in determination of its density if we have the value of the mass of the oil. For that to happen, a scale with electronic output can be used, but the system without oil has to be weighted beforehand so as to subtract that value from the value that will be displayed with the oil inside.
A change in density will always mean a change in purity. A constant calculated density, however, will not mean in 100% of the times impurity absence. As an example, if there are two strange objects (impurities) of equal volumes, if their real densities summed are equal two double the density of the oil, the overall calculated density will still be equal to the real density of the oil, so the impurities will not be detected. Basically, in a scenario where the equation below is true, the oil will be falsely considered pure.
m(V+V1+V2+…+Vn)=V(m+m1+m2+…+mn) , where
m = mass of the oil V = volume of the oil
m1,2,..,n = mass of object 1, 2, 3, …, n
V1,2,3,…,n = volume of object 1, 2, 3, …, n
To avoid false alarms, another parameter has to be measure, and that is viscosity. There are many types of sensors capable of measuring viscosity, each using a different method. For the errors that a viscosity sensors cannot account, the density calculation will and vice-versa. The scenario when density will not trigger any alarm in case of impurity presence is rare. The case in which a system that measures both density and viscosity would not trigger any alarm in the presence of impurity would be even more rare and hence, negligible.
In the eventuality of an ‘alarm’, the oil has to be evacuated, preferably through a filtration device, and the tank has to be decontaminated. In order to not stop the machinery from working during this process, a second such system would have to be used and switched to during tank maintenance.
In Figure 37 and Figure 38, two concepts were drawn to potentially start working on one of the two. However, the focus was changed to the plasma cutting machine and it came along with a list of matters to look into: measurement of environmental temperature, of temperature in the control cabinet and the temperature of certain machine parts, measurement of environmental humidity, of gas pressure, of the steel plate thickness, of the position of torch, of voltages, currents and of vibrations.
According to the information given by Mr. Alberink, the plasma cutting machine is sometimes behaving badly due to various reasons among which misplacements of the moving parts, insufficient power, bearing breakdown and motor malfunction can be enumerated.
The research showed for every specified matter at least one suitable sensor, along with the measurement method.
Figure 39 Thickness measurement sketch
Figure 39 depicts a sketch made to show a concept about measuring the plate thickness with proximity sensors.
The difficulty here consist of finding a reliable high resolution sensor and determining what is the minimum amount of sensors that have to be deployed so as to always have at least one that can see the steel plate. The answer depends on the minimum size of a steel plate and it is as follows:
Necessary&sufficient_sensors = [table length/minimum plate length]*[table width/minimum plate width]
The following facts are taken into account: the plate edges are always parallel to the table’s edges, so there is no angle to worry about; the square brackets in the formula stand for whole part; all the table is used for cutting, so there are no redundant parts of the table.
At this point the question of the reliability of low cost sensors was raised; so further, the research was focused on finding good solutions with the use of industrial sensors. Also, an interest in the distance from the torch end to the steel plate was shown. For the height measurement multiple laser scanners have been taken into account.
Figure 40 Sketch – front view of torch system with a radius laser scanner attached to the metal support of the torch
Figure 41 Sketch – front view of torch system with a radius laser scanner attached to the metal support of the torch, under an inclination to the right
Figure 42 Sketch – front view of torch system with a radius laser scanner attached to the metal support of the torch, under an inclination to the left
The difficulty Voortman Steel has with the height measurement is shown when the torch is inclined. Therefore, the cases presented in Figure 40, Figure 41 and Figure 42. Applying the formula BC=AD-AB*cos(α) would solve the problem in those cases. Further, the torch can rotate in other directions too. Figure 43 illustrates that scenario with the addition of a tilt sensor to measure the angle the torch inclines at.
Figure 43 Sketch - side view of torch system with an angular sensor attached to the metal support of the torch when it is pointing straight to the plate and when it is at an angle
The second order equation B’C’’^2 + 2*AB*B’C’’-[2*AB*BC+BC^2+sin^2(a)*AB^2]/cos(a)=0 can give us the length of the segment determined by the end of the torch and the cutting point. The normal distance from the end of the torch to the steel plate can also be determined by using the following formula: B’C’=cos(a)*B’C’’. This led to the realization that there is no scanner needed on the torch support because the z axis is already measured and another angular sensor would be good enough. However, if the laser scanner is not placed in that spot, the thickness of the plate would have to be known. This is why the laser scanner is proposed to be placed instead at the extremity of the bed so it could measure distances to two sides and the angle between them. This way, because we are situated in the Side-Angle-Side triangle case, with cosine’s theorem we can find the thickness.
However, these will improve the quality of the product (cut steel plate) and it has no impact on predictive maintenance. For this reason, the dimension measurements are considered out of scope. Instead, the vibration of the components and the
temperature/humidity in the electric cabinet seem parameters worth monitoring because the wear of the material can be predicted from the numbers.
In Figure 45 the output of a sensor system implemented in an electrical cabinet is depicted. This is just a simulation, the sensor values being randomly generated. Alarms are triggered if thresholds are exceeded. Data is remembered and collected daily, and also compared from a day to another. This is not a final version of the code; it is just the current final version.
Figure 45 Example of serial output for the electrical cabinet which can display each temperature and humidity value for each sensor separately, the lowest/highest and average values every time a sample is taken, while data is stored at the end of each day. The daily values reset on a daily bases while long term arrays store the values for the whole running period. The values are also compared from a day to another and in case there are big differences, a warning message is displayed. Certain thresholds activate symbol display next to the alarming values.
Overall heat constant increment can announce a future breakdown. The same case applies to vibrations, but for those, data has to be collected and analysed to find patterns and behaviours.[11][12]
Figure 46 Environmental temperature measurement system (can be applied to humidity as well)
Figure 46 depicts a simple method to measure the environmental temperature and, eventually, humidity. Three or more sensors should be placed in extreme places of the machinery with the use of extensions, making this way sure that the heat of the machinery does not affect the sensors. If one sensor is, however, affected, by comparing all the sensors, the odd value can be excluded. In the case where the sensors do not have to be attached to the machinery, one, two or three sensors, depending on the reliability that is required, should be deployed in the room.
The focus of the project was narrowed down to measuring sound and distances/velocities/accelerations of all the moving parts of the machinery. The sound measurement was brought to the table due to a graph that I have stumbled upon while researching, where the supplied current had an influence on the amplitude of the sound made by the machinery. Mr. Alberink confirmed that inspectors are able to tell if the machine is misbehaving by the sound of it.[13] One other reason for changing the scope was the high price of the industrial sensors which exceeded by a lot the planned budget which is predicted to be a few hundreds of euros.
Figure 47, Figure 48, Figure 49,Figure 50, Figure 51 depict a tri-dimensional model designed to present a few concepts.
