NDC Infrared Engineering X-Ray Tranflectance (XRTF) Project EPA pilot demonstration on technology EP05WO May 2006 Bo Franklin/Terry Bates

Transcription

1 1) Description of award NDC Infrared Engineering X-Ray Tranflectance (XRTF) Project EPA pilot demonstration on technology EP05WO May 2006 /Terry Bates Pilot demonstration of technologies requiring further development to reduce the use of radioactive sealed sources in industrial gauging devices 2) Introduction Thickness and weight per unit area measurements using Beta and Gamma Isotope sources are common in the web process industries (e.g. paper, plastics, coating, rubber, paint etc). These sensors typically serve the range 0.1 mil 400 mil ( microns). Most of current applications and challenges for these sensors are in the thinner materials (0.1 50mil, microns) The challenge is that the existing technologies are in widespread use and are delivering a precision of measurement and accuracy that currently satisfies many of the user requirements (Typically better than 0.5% and in many cases better than %) at a currently accepted price point. NDC has already manufactured an alternative solution using X-Ray absorption in a transmission sensor. The price point of this sensor is comparable with the Isotope sensors. This X-Ray technology has a higher susceptibility to product composition (Z Atomic Number) and although high precision (0.05% or better) of measurement is possible, the technology is sufficiently composition sensitive to be only usable on materials where production composition variation is well controlled. Page 1

2 . 2) Technology description Proposal This proposal is to replace current Beta and Gamma Isotope thickness measurement systems for the web process industries with a novel non-contacting X-Ray sensor operating in both Transmission and Backscatter modes simultaneously. The aim is to compensate the known composition sensitivity of the X-Ray Transmission sensor with a backscatter detector operating in the composition in-sensitive Compton Backscatter mode. Some exploratory work on suitable small form factor CZT detectors has already taken place. NDC also have experience of the correlation of multiple disparate sensor responses in a matrix solution based on previous sensor combinations. This design will therefore employ some novel detector and sensor correlation technology. This proposal will address technologies that require research and enhancements for future implementation into existing manufacturing practices. The principle employed is to utilize the factors that make a Gamma gauge relatively composition insensitive (by operating the X-Ray in a Compton Backscatter mode). The X-Ray transmission mode allows very high flux with a highly collimation design that minimizes sensitivity to sheet position, maximizes streak resolution and allows a noncontact 20mm gap design. The proposed design would have better performance than a Beta gauge (0.1% % precision) and comparable composition sensitivity. The design aim would be to equal or exceed the performance of current Beta gauge designs. Most components of this gauge are all currently available in existing Beta/Gamma/X-Ray detectors or in available sensor developments. Stable X-Ray tubes and PSU's are available and are already applied in NDC s X-Ray Transmission sensor. The novel feature is the dual sensor composition compensation combined with the proven transmission sensor performance. Page 2

3 Methodology a) Experimental Design and Test Protocol The steps to this design and test were 1) Design and characterize the response of an X-Ray Backscatter head (XRB) 2) Integrate the detector from an X-Ray Transmission sensor (XRT) to produce a combined Transmission/Reflectance sensor (X-Ray Trans-Flectance - XRTF) 3) Characterize the response of the XRTF sensor 4) Develop processing algorithms to give a compensated combined XRTF response 5) Compare the combined response to equivalent (Beta gauge ) performance Introduction: The measurement concept uses a combination of XRT and XRB. XRT technology provides a gauge with performance characteristics superior to beta gauges in all respects save one: material composition sensitivity. This composition sensitivity can create measurement errors due to product formulation changes in the process, and is a primary reason why beta gauges are still preferred devices over X-Ray. XRB technology is stated to be far less composition sensitive, however due to the lower efficiency of the backscatter signals this sensor mode offers less precision and poorer collimation. The concept of the novel XRTF device is to provide a single X-Ray source and monitor both reflected (XRB) and transmitted (XRT) energy signals. The transmitted signals will be used to provide all other measurement data, and the backscattered data will be used to provide a compensation for material composition effects. NDC already manufacture and sell a device based on X-Ray transmission technology, so the first effort was to evaluate an XRB device to determine its performance and degree of composition sensitivity, in order to validate the design concept. Page 3

4 14. Detector Selection: Backscatter detector selection is key to project success. Performance, form factor, and cost are all key to determining product performance and market viability. The detector technolgies evaluated were: Ionization chamber NaI + photomultiplier CZT( solid state) The Ionization Chamber option was selected for the following reasons: Flat energy response Temperature stability Compactness of design Durability Cost The CZT detector was rejected primarily because of cost and concern regarding temperature stability. The NaI + photomultiplier was rejected because of cost, physical size and the fact that energy differentiation is not required for this measurement. This detector saturates at relatively low X-Ray tube output levels which would preclude obtaining the full benefit of an X-Ray tube compared to a discrete Isotopic source, and improved measurement precision 15. Tube/PSU selection: X-Ray Tube: The tube requirements were based upon the experience obtained with our X-Ray Transmission sensor (already in production), and are, as follows: Page 4 HV range Beam Current Durability Stability Longevity

5 An additional requirement for this unit was : Compactness of design Two tube options were selected that satisfy the above requirements and afford the added security of alternative suppliers. PSU (Power Supply) Selection: Again, requirements were based upon past experience, as stated above, and are as follows: Reliability HV and Current Stability Compactness of design A new PSU from our current supplier was selected because it provided all of the above and it is physically smaller than the model used in our current XRT sensor. This reduced size is critical because, with a back scatter unit, the detector and X-Ray source must be located in one head. 16. Interface Electronics(X-Ray Source/PSU): The XRB is a new sensor specifically required for this project but it is interfaced to the rest of the measurement system in a similar way to existing XRT sensors. However, many functions of the earlier interface board have been relocated, internal to the head, which effectively reduces the complexity of the new interface board and at the same time enables the XRB to be more self contained when compared to the existing XRT unit. 27. Product Specification: A mechanical and electrical specification has been completed and is integral to the XRB Product Specification 31. Shutter Design: The shutter design was completed. This is integral to an X-Ray tube mounting block assembly. Also included in this assembly is a pneumatic piston actuator, a shutter status PCB, and a shutter status LED linked via fiber optics to the head outer surface. An X-Ray tube mounting block assembly was built and subjected to an accelerated life test in which in excess of 1.4 million operations were achieved and the unit remained operational. This is equivalent to a working life of at least 80 years. Page 5

6 Page 6

7 XRB and Integrated device assembly and testing: The final XRB Head was tested as an individual sensor and then in combination with an XRT Detector in the final XRTF Combination Prototype XRTF Sensor XRTF Combination Sensor XRB XRT Detector Page 7

8 XRB Component (Top Head) PSU Cooling XRay Tube Pre-Amp Detector Polarizing Shutter Collimation Page 8

9 Data Collection Method and Tests The XRB Prototype Tests were as follows: 1) Measurement Precision 2) Optimum Vertical Plateau 3) Calibration Range 4) Measurement Footprint 1) Measurement Precision This is the fluctuation of the calibrated output in measurement units equivalent gsm. The term gsm is grams per square meter and is gauging industry standard measurement units for mass per unit area. 1sigma 2sigma 2sigma Gsm 1second 1second 1sec % Page 9

10 output in gsm output in gsm 2) Optimum Vertical Plateau This is the optimum vertical gap spacing and is depicted by a change in the calibrated output with distance. This test was performed using three samples ranging from 400 to 1700 gsm. Results shown graphically below: 423 Z Plateau Data Plotted Nov3rd gsm sample sample/head spacing in mm Z Plateau Data Plotted Nov3rd gsm sample sample/head spacing in mm Page 10

11 Voltage output output in gsm 1715 Z Plateau Data Plotted Nov3rd gsm sample sample/head spacing in mm The initial geometry yielded an optimum working gap of approximately 27 mm and this will need to be reduced by fine tuning the geometry for the final product. However, this is acceptable for a fixed bench test of the combined sensor. 3) Calibration Range This test demonstrates the maximum working range of the XRB component of the sensor and this exceeds the range of the existing XRT sensor: Calibration Curve Plotted Nov1st2005 Natural PVB sample set sample gsm Page 11

12 % max output 4) Measurement Footprint This is the size of the beam patch on the sample material and is an important aspect of the geometry of the package because it determines how close a measurement can be made to the edge of a production web. As this is application-dependent it is best to have at least two versions. Beam Patch (or Measurement Footprint) Plotted Plotted Nov8th fabricated 5 mm strip collimator production 5mm strip collimator no collimator dimension in mm 24 Initial Sensor and Geometry Assessment Some of this category is covered above but I have interpreted this to refer mainly to the proof of concept test which was performed as follows: 1) Geometry: Both the XRB Prototype and the XRT detector unit were secured in a test rig which was adjusted to give a gap between the two heads of 20 mm. This geometry was not yet optimized but this will be the approximate optimum gap for the final product. 2) Detector Adjustment Initial output level tests and adjustments were performed so that the detector outputs would a) not saturate b) yield a workable output level for the range of samples to be presented. Page 12

13 3) Sample Selection Initially a set of PVB (Poly Vinyl Butaryl) samples were selected because a good range were available including a batch with a high filler content 4) Method 4.1) The natural PVB samples were presented, in turn, at approximately mid-gap using a sample presenter. 4.2) The analog output from each detector was recorded, along with the sample gsm value, in an Excel file. 4.3) The procedures above were repeated for the PVB samples containing white filler. 4.4) The data was plotted in Excel. 5) Results: Conditions: X-Ray Tube: Working Gap:20mm(not optimized) HV=14.5 kev I= 0.5 ma Transmission mode Sample Output PVB gsm Volts Vnorm (White loading) vmin Sample Output PVB gsm Volts Vnorm (natural) Page 13

14 Backscatter Mode Sample Output PVB gsm Volts (White loading) vmin Sample Output PVB gsm Volts (natural) Page 14

15 Normalised analog output PVB sample set Transmission Characteristics Plotted Nov22nd PVB(natural) Transmission PVB(w hite loading)transmission sample gsm value Normalized analog output (Vnorm) = Vs- Vmin / Vmax - Vmin Where: Vs= Output volts for a sample Vmax= Output volts in air (no sample) Vmin= DC offset of the detector head amplifier Page 15

16 analog output 1.9 PVB sample set Backscatter Characteristics PlottedNov22nd PVB(w hite loading)backscatter PVB(natural)Backscatter sample gsm The outputs for the backscatter mode were not normalized because the maximum signal increases with increasing sample gsm unlike in the transmission mode where the maximum signal is determined by the air gap. However, as with Transmission, the air reading is the same for both Backscatter measurements, intercept on the vertical axis. Page 16

17 output volts ( normalised air output offset) 6) Additional Testing Following the above tests similar testing was performed on a wide variety of samples of different material types and compositions. Data was collected from both the backscatter and transmission devices. The plotted results are included below: Backscatter Mode Plotted Dec7th PET (clear) BOPET+SiO PS(white ) Rubber(black) Ballistic nonw oven(rubber+pe)dsm PVB(natural) PVB(w hite) PA (clear) PE+EVA (clear) sample gsm Page 17

18 Vnorm Transmiission Mode Plotted Dec2nd PVB white PVB natural Rubber(black) PET (Clear) BOPET+SiO2 dsm ballistic(non woven rubber+pe) PS PA PE+EVA(clear) sample gsm Page 18

19 7) Summary or XRB and XRT tests : 1) The results above indicate that the response (slope) for both the Transmission and Backscatter modes are different for a given product and different again if the composition is altered, i.e., filler added. Transmission: Backscatter: The output reduces as the sample gsm increases, until there is not enough signal left with which to make a measurement. The output increases as the gsm value increases, until eventually a saturation level is reached such that an increase in sample gsm does not yield an increase in output. 2) The gsm value of the product can be written simply as follows : Sample gsm= K1*T + K2*BS Where, as a dual measurement, both outputs can be combined to provide the answer, but ideally K1 and K2 must be constant for a given mix. 3) The challenges are, in order: 3.1) To find a robust relationship between the two measurement modes for a given product. 3.2) Create the software to support this model. 3.3) Repeat this test on several different product types so that the amount of work required to characterize this type of measurement can be ascertained. 4) The geometry must be fine tuned to provide an optimum working gap of typically 20 mm. Page 19

20 Results analysis: Analysis of the detector response in the two modes (Backscatter and Transmission) shows a roughly symmetrically opposite response to the different material compositions. Utilizing this effect the intention is to use the Backscatter response to compensate the Transmission Sensor to give an output desensitized from composition. Phase 2 of the project took a wide range of samples of differing materials and compositions and presented these to the Experimental XRTF Sensor to obtain measurement data. This Phase concluded that the Transmission and Backscatter modes did show significant differences in output, however one disappointing fact that was learned from these experiments was that the X-Ray energy required for adequate Transmission sensitivity severely reduced the composition insensitivity in backscatter mode. This renders a simple compensation, based on the improved Backscatter insensitivity, unfeasible; however a linear solution using the two responses was proposed. Compensation Method In Backscatter mode the signal typically increases with increasing sample gsm whereas in the transmission mode the maximum signal occurs when there is no material in measurement path and decreases with increasing thickness. Test data for a wide variety of samples of different material types and compositions was collected from both the backscatter and transmission devices. The plotted results are included below Note the data plotted on these graphs from the Phase II report for the backscatter mode was not normalized. Page 20

21 This Graph shows a wide spread of data in Transmission mode, however this is not too easy to interpret in this Raw Normalized Voltage representation as the data is strongly nonlinear Similarly the Backscatter Raw data shows wide data spreads in this raw format Page 21

22 Final Step: Data Processing treatment: Background The measurement principle of a radiation absorption gauge follows the well known Beer-Lambert law : I(abs)=I(ref)e -μwt or I(abs)/I(ref) = e -μwt Where I(abs) is absorbed radiation, I(ref) is radiation with a reference absorber in the measuring gap, Wt is the weight per unit area of the measured material and μ is the absorption coefficient. From the above equations it can be seen that taking a natural logarithm will linearize the data Ln(I(abs)/I(ref)) = Wt * absorption coefficient (μ) In Beta, Gamma and X-Ray sensors μ gives the relationship between the scattering mechanism and the interaction of the incident radiation on the atoms and molecules of the material to be measured and has been empirically shown to depend both on the energy characteristic of the radiation and the atomic weight of the substance being measured. Backscatter radiation is defined as that radiation which leaves the surface of the material on the same side as the incident radiation beam. In the case of this combined sensor, one incident beam of radiation is passed though the material to the transmission detector and the same beam is backscattered to the backscatter detector The Backscatter response is also non-linear but the backscatter response typically increases with increasing thickness for most polymer materials. Again there is a limiting thickness the saturation thickness - above which no further increase in signal is possible. X-Rays interact with matter via three mechanisms, Photoelectric effect, Compton scatter and Raleigh Scatter At the lower energies required to obtain sufficient interaction with the absorber, giving adequate measurement sensitivity, the photoelectric effect predominates and this has been shown to be highly dependant on atomic weight of the absorber and for X-Ray transmission gages gives rise to significant composition sensitivity. Page 22

23 Linearization The data was processed so that linear responses could be compared directly. The normal procedure for linearizing this type of data is to take the raw voltage, normalize into the range 0 1 (traditionally done for Beta gauges to also eliminate effects of source decay). Then Linearize as explained above by taking a Natural Logarithm (Ln) following this stage the data is predominantly linear however any residual non-linearity can be further corrected by various curve fitting techniques (eg Polynomial Interpolation) The Ln of the Transmission measurement cannot be compared to the Ln of the Backscatter readings unless the latter is processed so that the two measurements are in comparable scales. To do this the Tramission (Tx) data was first normalized where the data was given the traditional normalization:- (V- Vmin)/(Vmax-Vmin) In this case Vmax is Shutter open and Vmin is shutter closed reading In the case of the Backscatter (Bs) the normalization must be somewhat modified as the Maximum reading generally comes at infinite thickness (the saturation effect) In this case infinite thickness was assigned by an arbitrary estimation of a value beyond the measurement range such that this Vmax scaled the whole dataset into an appropriate range. Following this the Bs Vnor was inverted by subtracting it from 1 to allow similar slopes and logarithms to be generated Page 23

24 Following this processing we can visualize the data graphically: Note that the Backscatter (BS) data and the Transmission (TX) data are symmetrically opposed This Symmetry can be used to compensate the Transmission data by applying the difference as an offset Page 24

25 The Graph below shows how this is applied in practice: Note the PVB Backscatter and PVB Transmission responses and the PVB Adjusted response shown by the red line The same processing was applied to the complete data-set Normalization, Logarithm then a Polynomial interpolation to more accurately linearize the data The result of this processing is shown in the next two graphs Standard processing as done on a normal X-Ray XRT Adjusted processing using the Backscatter compensation to generate the XRTF response Page 25

26 Page 26

27 Discussion: The Graphs above show that a relatively simple correction from the Backscatter gauge significantly improves the Transmission data spread. It is likely that this technique can be further improved by linear equation methods. Whilst this looks promising as a first step to finding a general robust solution there are still some other areas of concern. The Backscatter response to the Aluminium, PVC and Barium loaded materials. In all these cases a reduction in signal was seen with increasing sample weight up to a saturation point when the measurement leveled off. These materials could possibly be measured at a higher energy and the XRB response will be different and less dramatic. This has still to be tested. It is possible that self absorption in the sample material is leading to early saturation at the low energy setting It would seem that up to the saturation point something similar could be done to manipulate the response to allow a correction, although the response would need different treatment and would require recipe stored calibrations and compensation switches. However the nonlinearity at saturation makes the remaining data lacking in resolution and discontinuous with the Transmission data The reasons behind these inverse responses need to be examined further. Page 27

28 Measures of Success Operating Performance Data A good basis for comparison would be obtained if we present the full sample set to the XRTF and to the Beta gauges (Kr85 an Sr90 are required as the data set exceeds the Kr range), this will allow a qualitive judgment on the value of this adjusted sensor Comparison to Beta Measurement Composition Sensitivity Examples of Beta measurements compared to the XRT and XRTF sensors on samples of a number of disparate materials ranging from Polystyrene (PS), Polyethelene (PE), PolyAmide (PA) through to PolyvinylButylate (PVB), Rubber and Papers The test was done by calibrating the sensors on PET (Mylar) and measuring the other materials with the same calibration to give Error from Mylar data. As most gauges are initially calibrated on Mylar this is a practical comparison method Value of the Compensation A typical gauge will be set up and calibrated on a production process that manufactures a much more limited range of materials than have been presented to these sensors in the course of these tests In practice the calibrations can be adjusted and optimized for the known materials. XRT is already used on many of the simpler materials where composition variations are not an issue. The main advantage of the technique presented is to limit the variation in response to the composition changes so that the gauge becomes simpler to calibrate and less affected by these composition changes. As the XRTF can have a much higher precision of measurement Page 28

29 % Error PS Pa PE+EVA Rubber +Pe PBV Nat PVB Wt Rubber BOPET+Si Papers Percent Error from standard Mylar curve PS Pa PE+EVA Rubber +Pe PVB Nat PVB Wt Rubber BOPET+Si Papers Avg Err Statistics Beta XRT XRTF The sample materials are arranged in increasing order of difficulty as far as composition loading is concerned Note that the Beta gauge is typically within a few percent of the Mylar curve, normally well within 5 10 % The XRT is very variable on the same materials with 10 20% errors not unusual and much higher errors (>100%) possible The XRTF is close to the beta on the simpler compositions and generally within about 10% until the more complex products are seen here it is still about 3 times better than the XRT and has much improved statistics over the Beta gauge. Percent Error from standard Mylar curve Beta XRT XRTF Material Graphical plot of the above data Page 29

30 Comparison of Beta vs XRTF 4500 XRTF Adjusted Data PET PA PE+EVA 1500 PS PVB White DSM(Ballistic non-woven,rubber+pe) PVB Natural 1000 Papers BetaTiBa BetaTi Papers Wt 500 Papers Ye Linear PVBNatBeta PVBWtBeta Beta Standard Data PET PA PE+EVA PS PVB White DSM(Ballistic non-woven,rubber+pe) PVB Natural Papers Papers Wi Papers Ye Beta Ti Linear PVBNatBeta PVBWtBeta This shows the Beta gauge variation compared to the XRTF over the full range of samples tested. Page 30

31 Reliability and Ruggedness X-Ray based gauges operate in similar environments to Beta gauges, they require slightly more cooling as the thermal load of a sensor mounted HV PSU and X-Ray tube are higher than typical Beta gauge Electronics Tube life has been tested to be in excess of 5 years based on choice of operating current and HV and tube design, and the use of soft start strategies in the PSU design. The Prototype design will have a cost basis around 20% higher than a similar Beta gauge as normally sold, however where lifetime costs include source disposal the Beta gauge may well be more expensive. 3) Commercial Feasibility a) Market(s) for the alternative gauge All existing Beta gauge markets and some Gamma gauge markets b) Competitive advantages and disadvantages with existing nuclear gauges Efficacy of the technology (i.e., its ability to perform successfully and reliably), with emphasis on innovation and extent of use o Offer excellent pass line tolerance (geometry), high precision of measurement (over a Beta sensor) and good composition performance (comparable to a Beta Sensor); Capacity of the technology to replace existing radioactive devices and significantly impact the industry sector that uses the device o Replaces Gamma and Beta gauges in the gsm range o Cost structure 10-25% more than Beta technology Relative benefits and risks of the proposed technology compared with industrial gauges that utilize sealed radioactive sources; o Equivalent or better performance, no hazard when off, lower radiation scatter footprint when on easier to shield Page 31 Potential broader applicability of the proposed technology to in a variety of applications and industries

32 o Can cover much higher range of thickness than normally available Isotopes c) Opportunities, barriers Opportunities o There is a significant opportunity to replace beta gauges with this XRTF technology o There are very few applications where this technology could not replace beta gauges o Through further enhancement of the established linearization algorithms, perhaps further performance improvements can be made and additional features developed Barriers o Market acceptance of new technology o Licensing barriers by regulatory authorities o Finalization of design work (optimization of sensor optics, packaging, safety testing, etc.). o Additional funding could hasten the removal of these barriers Page 32

33 4) Conclusion summary, feedback on the EPA program In summation, the new technology holds significant promise. The mock-up device has proven the validity of the original design concept. A final product based on the design is, in our opinion, extremely viable. The prototype does have to be productized; and commercialization will include having to overcome licensing and market acceptance obstacles. NDC is prepared to move forward with the final productizing of the sensor, and would appreciate any inputs or assistance from the EPA or PSI to help facilitate / fast-track such a development. Page 33