Project Progress

Summary

A new data format convention was proposed by the software developer (Dr. Choon Park) to maximize the downloading speed and compatibility between the current hardware design base (that uses the LabVIEW TDMS format) and the software development base (that uses the ParkSEIS format). Considering the vast amount of data that, when completed, will be coming from the acquisition system for the pseudo-real-time onboard analysis, there should be a compact data format without unnecessary overhead information that will allow the software package focus directly on the analysis algorithm rather than spending extra time for data importing. Click for a brief demonstration of the new format on YouTube. A temporary data format, however, will be used as a communication tool to deliver data sets already collected by using a similar hardware system developed for other related projects at the Lund University until the hardware system for this project is fairly ready to be tested with the software package. The KGS format (a modified SEG-Y format) will be used 
until then. A field data set collected by using the 48-channel MEMS microphones at the Lund University has been delivered to Park Seismic for the evaluation, by using the ParkSEIS MASW software package, of the optimum number of channels per longitudinal array (NCh-opt) and the optimum channel (receiver) spacing (dx-opt).

Modification of the ParkSEIS software, the base software package to be  developed into the software packaged for this HMA project, has started. The first task is to upgrade its capacity to handle the highest frequency from 30 KHz to 100 KHz. This process is currently being applied to all related components in data import, display, FFT, and the 2D-wavefield transformation module to generate the dispersion image data sets.

Main hardware components have been specified and the first pricing quote has been received at the Norrfee Tech, AB. It seems the overall hardware budget (now re-estimated approximately as $35K) is higher than the one originally  proposed (~$30K). It has been discussed to submit an amended budget proposal by shortening the overall project period so that the part of the labor cost can be reallocated into the hardware budget. This topic will be  continuously discussed as the project proceeds.

Monthly Project Meeting (February 25, 2020) (Minutes)

Progress Table (Prime Contractor - Park Seismic LLC) - February, 2020

Progress Table (Sub Contractor - Norrfee Tech) - February, 2020

Summary

Hardware Parts Ordered

Parts to build multichannel acquisition system (MAS) have been ordered at  National Instruments (NI). They are the main components needed to build a 64-channel acquisition system (expandable up to 96 channels). Other accessory  parts (e.g., MEMS microphones) will be ordered soon.

Specifications of parts ordered (as listed on the quotation in English/Swedish)

Optimum Number of Channels Per Array (March, 2020)

The final system will consist of multiple linear (1D) MEMS microphone arrays arranged along the transverse direction side by side (see Figure 5 in the  proposal). It is critically important to figure out the optimum configuration of  the one (longitudinal) array in the channel spacing (dx-opt) and the number of  channels allocated per array (NCh-opt). By using a field data set acquired with a 48-channel MEMS microphones with a very small channel spacing (i.e.,  dx=0.752 cm) (i.e., spatially over-sampled) over a special type of asphalt pavement in Sweden, a series of test have been executed through two types of spatial re-sampling approaches; i.e., the simple re-sampling of decimation and the stack-re-sampling approaches. The former simply reselects existing traces  (channels) from the original full (48- channel) field record with a greater spacing (e.g., 2dx, 3dx, 4dx, etc.) by discarding in-between traces. On the other hand, the latter stacks (instead of discarding) them to make a new trace obtained with a greater spacing. This stack approach, called "receiver array", has been used for a long time in the seismic exploration for the purpose of attenuating the  strong, but propagating at very low velocity, ground roll surface waves. These two re-sampling approaches are compared in the performance of enhancing the signal (Lamb) waves while attenuating the noise (air) waves at the same time.

According to the test results, it seems a 16-channel array with a 5dx (=3.76 cm), or a 4dx (=3.008 cm), channel spacing that connects (in parallel) all in-between MEMS microphones arranged with the current spacing of 0.752 cm. This means the stack-re-sampling approach turned out more effective than the other simple re-sampling (decimation) approach. With this configuration, the system will be able to handle HMA layers in a thickness (H) range of 4 cm (Hmin) - 20 cm (Hmax).

see SUMMARY of the procedures and results.

see APPENDICES for procedural details.

Monthly Project Meeting (March 31, 2020) (Minutes)

Progress Table (Prime Contractor - Park Seismic LLC) - March, 2020

Summary

Acquisition System Being Constructed (16-bit max. 96-channel AD converter)

Building the first (and the main) hardware component (AD converter) has  started. It is the NI PXI System that can acquire a maximum of 96 channels with a 16-bit dynamic range. Description of the component items and  corresponding specifications are listed here. It is "parallel prototyped" by referencing to the previous system developed for a similar project at Lund University (LTH) in 2019 ("SYS-RYD- 2019") (see photo below). The new system is more compact in size and lighter in weight. It consists of four (4) 24-channel ports allowing a maximum of 96-channel acquisition. However, only 64  channels will be used for this project. It will also include additional capacities of recording GPS coordinates and temperature of the HMA surface at the  measurement points. The GPS receiver and infra-red (IR) temperature sensor are shown in a photo below.

Software Approach to Attenuate Impact-Generated Sound Wave

When using microphones as receiver to record the leaky-mode surface waves that are "leaked" from the HMA pavement surface, all ambient sound waves are also recorded as noise. The most troublesome sound wave is the one generated at the time of impact. It is always generated when an impact is applied on the pavement surface to generate seismic (surface) waves and it is very strong in
comparison to the level of signal seismic waves as shown in Figure 1 below. If not properly handled, it can cause a low signal-to-noise (SN) ratio at the high-frequency part of seismic signal (e.g., > 20 kHz) that can ultimately lead to a reduced accuracy in the evaluated frequencies (e.g., 1-10 kHz) that is critically important to accurately evaluate the thickness (H) of the pavement layer. Therefore, it is rucialcomparison to the level of signal seismic waves as shown in Figure 1 below. If not properly handled, it can cause a low signal-to-noise to reduce the level of this type of sound wave not only at the time of data acquisition but also during the data analysis steps.c This is  especially important when dealing with an HMA pavement (rather than a concrete pavement) because of the more significant attenuation of seismic waves occurring with the HMA pavement as noticed in Figure 1b. A possible effort to reduce the level at the time of data acquisition has been outlined in one of the documents from previous month's progress.

Description of the test results obtained by using different software approaches is provided here.

Figure 1. (a) 48 channel MEMS microphone array and (b) acoustic data recorded while driving at 30 km/h. (From Ryden et al., 2019)

Summary

Acquisition System Being Constructed (16-bit max. 96-channel AD converter)

The system is under construction and scheduled for a field test in June-July by using the MEMS microphone receiver array used for SYS-RYD-2019.

Software Approach to Attenuate Air Waves (Continued Development)

Direct air (sound) waves generated at and coming from the impact point are the most troublesome noise for the microphone measurement of leaky-mode Lamb waves that have to be suppressed by means of both acquisition and post-acquisition efforts. In previous summary of the development, several  different multichannel data-processing approaches were considered for their 
effectiveness in attenuating the air waves (noise) while preserving the Lamb  waves (signal) as mush as possible. They included fk-filtering, muting, and fk-muting approaches. They all turned out having pros and cons for the final goal of velocity (Vs) and thickness (H) evaluation of the HMA layer. This indicates all of them can be useful in different combinations under different condition of the
acquired data; e.g., data with a low SN ratio vs. a high SN ratio, etc.

The methodological development further continued to come up with a more effective approach based on the linear-move-out (LMO) approach. This report summarizes a new successful approach that uses a moving-window LMO stack to attenuate the air wave event of a specific propagation velocity (e.g., 334 m/s). Test results with synthetic and real field data sets indicate the effectiveness is superior over the previous conventional approaches (e.g., fk, mute, etc.), especially for the Impact Echo (IE) method that can evaluate the
thickness (H) independently from the analysis of propagating waves. This part of the result (IE) will be reported in next summary report.

TDMS Data Format

Although it seems to contain unnecessary components that can slow down data transmission speed between the hardware and the onboard software modules, the TDMS format will be condered for possible adoption after evaluating its structural format.

Summary

Acquisition System Being Constructed (16-bit max. 96-channel AD converter)

The system is under construction and scheduled for a field test in June-July by using the MEMS microphone receiver array used for SYS-RYD-2019.

Thickness (H) Evaluation [Lamb Dispersion (A0) and Impact Echo (IE) Approaches]

Thickness (H) of HMA layer can be evaluated from the same Lamb-wave  dispersion curve (A0) that is used to evaluate the shear-wave velocity (Vs). On the other hand, it can also be evaluated from the spectral characteristics of  obtained seismic data through the Impact Echo (IE) method. Both approaches are examined during the month of June, 2020.

In theory, thickness of HMA layer (H) can be measured by the same Lamb  dispersion curve, the fundamental anti-symmetric (A0) mode, that is used to evaluate the shear-wave velocity (Vs) of the layer. An A0 curve consists of a curved part that has increasing phase velocity with frequency and occurs at the low-frequency side of the curve (e.g., 1-10 kHz) and a relatively flat part with a fairly constant phase velocity that occurs at the higher-frequency side of the curve (e.g., 10-50 kHz). The constant phase velocity of the curve corresponds to the Rayleigh-wave velocity (VR) of the HMA layer, which is about 92% of the shear-velocity (Vs) for the common range of Poisson’s ratio (e.g., 0.15-0.35). For a given HMA layer of VR, the onset frequency of the curved part from the flat part of the curve changes with the thickness (H) of the layer; i.e., the thicker  layer will result in the lower onset frequency. In consequence, the overall 
shape of the curved part will also change with H. All these properties of A0  curve are graphically illustrated in this report.

Conventional seismic measurements through contact approach (e.g., accelerometer) well depicted both curved and flat parts of A0 curve. However, the non-contact rolling measurements by using MEMS microphones did not show clear definition of the curved part although they delineated the flat part with superb quality. It is believed that this can be closely related to the spectral  characteristics of the light impact source that tends to generate more energy at higher frequencies (e.g., 10-50 kHz). This, however, will be verified soon in the near-future when more microphone data sets are collected by using multiple  impact sources of different impact energy. However, the phenomenon may also be related to the near-field effect of surface waves as well as attenuation properties of the pavement layer.

In this report, these two aspects (i.e., near-field effect and attenuation property) are examined by using field data sets acquired in the past by using both contact (i.e., accelerometer + hammer) and non-contact (i.e., microphones +  bouncing ball) approaches.

In parallel to the Lamb curve (A0) approach to evaluate the thickness (H),  another independent approach that works on the spectral characteristics of the measured seismic wavefields is now considered. It is the Impact Echo (IE) approach that measures the spectral peak that is associated with the seismic resonance within the solid material. The peak occurs at the resonance frequency (fr) that is determined by the P-wave velocity (Vp) and thickness (H or d) of the layer; i.e., = BVp / (2d) where B (=0.96) is called a correction factor that slightly changes with Poisson’s ratio, but is commonly set to a constant value of 0.96.

First, two different approaches are tested to construct the amplitude spectrum;  i.e., stacking individual amplitude spectrum from each channel’s data  [frequency-domain stacking (FDS)] and stacking all channels seismic data first  and then construct the amplitude spectrum [time-domain stacking (TDS)]  (Ryden, 2016). It turned out TDS is far more effective than the former. Second, the most recent development by Bjurstrom and Ryden (2016) that applies negative phase velocity to construct a frequency-phase velocity spectrum is tested. This approach (Bjurstrom and Ryden, 2016) turned out highly effective  and will be used as the main IE method in this project. 

The future plan is to apply both approaches (A0 curve and IE) to evaluate the  thickness (H). Then, both values will be used to come up with an average value (Have) by applying an appropriate weight to each value based on a few quality  factors.

In this report, the general scheme of the IE method is briefly explained. Then, a synthetic seismic record is generated to be used to test each type of IE  approach. Finally in this report, Bjurstrom and Ryden (2016) is used for real data  sets from both contact and non-contact approaches.

TDMS Data Format

A project to build a C++ module to read the TDMS file of seismic data saved by the acquisition system being developed at Norrfee Tech is launched. The project will complete a separate C++ program that reads TDMS files and  converts them into the ParkSEIS data format, which is the standard format for the software package being developed.

Progress Table (Prime Contractor - Park Seismic LLC) - June, 2020

Summary

Field Data Sets from Newly Constructed Asphalt Road near Lund, Sweden (July 22, 2020)

Field data sets acquired on July 22, 2020, on a newly constructed asphalt  pavement near Lund, Sweden, are processed and evaluated for various  purposes. The data sets were collected by using the old system (“SYS-RYD-2019”). The overall data quality is evaluated as “excellent” from raw field records and corresponding dispersion images that clearly show well-developed Lamb wave arrivals and the fundamental-mode anti-symmetric Lamb dispersion (A0), respectively. Below are one of the several hundred records obtained (> 500) in the field (left) and corresponding dispersion image and other spectral characteristics (right). An approximate inversion results show velocity (Vs =  400 m/), thickness (H = 0.1 m), and Poisson’s ratio (POS = 0.3) (bottom). These are the results Norrfee Tech provided.

Further Testing of Air-Wave Alleviating Techniques

Testing of various multichannel techniques previously tested/developed that  can alleviate adverse influence of air waves on the target analyses (i.e., Lamb  and Impact Echo analyses) are tested for ten (10) randomly selected records  from the data set on July 22, 2020. Three previously developed techniques are tested that can alleviate the air-wave energy as much as possible without compromising other useful attributes of obtained wavefields [e.g., wavefields for Lamb and Impact Echo (IE) analyses]. They are surgical mute (MUTE), fk-filter (FK), and air-wave subtraction by using moving-window LMO stack (AIR-SUBTRACTED). This report contains more details.

Quality Control (QC) Parameters

Testing of several quality control parameters for a particular type of wave (e.g., air or Lamb waves) such as overall amplitude (“Amp”), arrival time in ms [“msT0 (ms)”], velocity [e.g., “Vair (m/s)”], and peak frequency [“Fpeak (kHz)”] is performed by using the aforementioned ten (10) field record. This report contains more details.

TDMS Data Format

A project to build a C++ module to read the TDMS file of seismic data saved by the acquisition system being developed at Norrfee Tech was launched as of June, 2020. The project will complete a separate C++ program that reads TDMS files and converts them into the ParkSEIS data format, which is the standard format for the software package being developed. It seems necessary to use some library files provided by National Instrument (NI) for some functions to work properly within the C++ codes. This project is about to finish. 

(One, #9, out of several hundread records collected)

Progress Table (Prime Contractor - Park Seismic LLC) - July, 2020

Progress Table (Sub Contractor - Norrfee Tech) - July, 2020

Muting Air and Pre-Lamb Wavefields

The importance of properly alleviating air wave energy existing in an acquired  raw field record has been addressed multiple times in the previous work reports (e.g., April, May, and July progress reports). This report summarizes outcomes  from one of those approaches when applied to a field data set acquired on  August 12, 2020, on the same road previously surveyed (see here for previous  survey results). Among multiple approaches, It is the “mute” approach  summarized in this report that erases (i.e., zeros) all wavefields outside a designated mute window in a time-offset display of a raw field record. It turned out to be the most effective when constructing an accurate Lamb dispersion  trend. It is also important to apply a pre-Lamb mute to reduce harmful influence from ambient noise. Therefore, both air-wave mute and pre-Lamb  mute are simultaneously applied to the raw field record. Three (3) different types of mute have been tested. One is called “MANUAL Mute” that sets the  mute window by visually inspecting the record to identify air-wave and Lamb-wave arrival times (msT0-air and msT0-Lamb) and propagating velocities (V-air and V-Lamb). With this approach, each record is manually muted one by one. The other type is called “SURGICAL Mute” that sets the mute window semi-automatically based on the presumptive values for msT0-air, msT0-Lamb, V-air, and V-Lamb. The arrival times (msT0-air and msT0-Lamb) can be approximately
estimated based on the acquisition parameter of pre-trigger time (e.g., 1.0 ms), which is the length of previous time after air waves triggered the recording and therefore should be included in the acquired record. The velocity values (V-air and V-Lamb) can be set arbitrarily based on a common range of each type of waves (e.g., V-air = 340 m/s and V-Lamb = 1500 m/s). Once these values are set, then the “SURGICAL Mute” applies the same mute window to all records existing in the input file automatically. The last type is called “AUTO Mute” that tries to detect all four parameters (i.e., msT0-air, msT0-Lamb, V-air and V-Lamb) automatically through its own detecting algorithm. The algorithm is based on the Linear-Move-Out (LMO) stacking to detect different waves (e.g., air vs. Lamb) arriving with different velocities. The “AUTO Mute” includes additional  algorithms that can detect other wave attributes that may exist in the raw field record such as reverberating air wave arrivals, air wave arrivals from the opposite end of the receiver array, too-low signal-to-noise (SN) ratio, etc. They are automatically applied whenever detected being necessary to improve the  accuracy in the evaluation of the mute window and also to improve the overall SN ratio of the recorded Lamb waves.

All three (3) approaches are applied to a data set acquired on August 12, 2020,  on the same road previously surveyed on July 22, 2020. Results of time-domain records and corresponding dispersion images are displayed in this report for various comparison purposes in this report. Results from all three approaches are comparable to each other in seismic data and dispersion image qualities. However, the “AUTO Mute” provided slightly superior quality. The comparison will continue to be made on further future data sets to confirm the superiority. Advantage of pre-Lamb mute is illustrated in this report by using a field record that shows dispersion image trend at high frequencies (e.g., > 20 kHz) is  improved due to the attenuation of ambient noise in the corresponding high
frequency range.

Importance of using a “proper” mute window is also illustrated in this report by using a field record.

Data Acquired on August 12, 2020

All (20) field records acquired on August 12, 2020, are displayed in this  report in raw data format as well as processed dispersion image format for the purpose of evaluating Lamb wave quality. They were obtained by using both  forward (FWRD) and reverse (REVS) impact sources attached at forward (i.e., #1 channel side) and reverse (i.e., #48 channel side) side of the receiver array, respectively (watch the video below). All reverse-shot records, however, are purposely flipped to look like forward-shot records by the software at the time
of data import, which is necessary to apply all mute approaches properly. In consequence, those reverse-shot records are separately marked by (*) to indicate so.

Comparing dispersion images from forward- and reverse-shot records, those from the reverse shots did not have proper Lamb dispersion trend as shown in this report. This seems to be related to a short pre-trigger time (0.5 ms) that was not long enough to catch the Lamb waves that arrived earlier than air  waves by more than 0.5 ms (e.g., 1.0 ms) due to the extended distance between  the array center and the impact point of the reverse shot. This will be  verified from a next field survey with a longer pre-trigger time (e.g., 1.5 ms). This hypothesis was tested in two slides by using rough estimation of impact distances for both shots (i.e., FWARD and REVE) from the receiver array by using a configuration shown on a photo.

Amplitudes of forward and reverse shots are compared by displaying peak  amplitudes of all (48) traces in each record in this report. The display indicates that, in general, the reverse shots were more powerful than the forward shots.

Decimation of channels (i.e., from 48 to 16 channels) was tested for each record from surveys performed during July and August for the purpose of evaluating any possible degrade of data quality in Lamb dispersion image. The results shown in this report indicate there is little difference, indicating that a 16-channel array will be able to provide an equally good quality.

TDMS Data Format

A separate project to make a C++ module to import a TDMS data file and convert into a ParkSEIS (PS) data file has been launched in mid July. The project was executed by a freelancer who used to work on building PXI systems (a similar hardware system for this project) that saved output data in TDMS format. The final code set has been delivered and is currently under modification to be merged into the ParkSEIS-HMA software package. It is not an entirely independent C++ code set, but has a dependency on the NI's built-in library that needs to be downloaded online and installed on the computer that runs the ParkSEIS-HMA package.

A 48-channel MEMS microphone array rolling with both forward (right) and reverse (left) bouncing impact sources.

Progress Table (Prime Contractor - Park Seismic LLC) - August, 2020

Progress Table (Sub Contractor - Norrfee Tech) - August, 2020

Summary

Project Updates at 2020 NRRA Pavement Workshop

Project updates have been presented as a 2020 NRRA workshop series on Thursday, September 3, 2020. The initial development, progress made so far,  and the future direction of the project development have been presented during a 1-hour workshop. The overall summary of the project and the software development were presented by the principal investigator, Dr. Choon Park, while the hardware development was presented by co-investigators, Dr. Nils Ryden and Dr. Josefin Starkhammar. A total of 45-min presentations were followed by a 15-min Q&A session. NRRA posted the entire presentation at YouTube.

Automatic Muting (AUTO Mute) - Refinement Testing with Data Acquired on September 1, 2020

New field operations were conducted on September 1, 2020, at the same place used previously for data sets collected during July and August. Among many 48-channel records collected (about 500 of them), two TDMS files were obtained from the Team Sweden; i.e., “2020-09-01_19_12.TDMS” and “2020-09-01_19_24.TDMS” files that contained 100 and 50 field records, respectively. The same pre-trigger time of 0.5-ms was used. This 0.5-ms time, firstly used in August, turned out later (late August) too short to capture the earlier-arrival Lamb waves for the records obtained from the reverse shots (i.e., from the impact by the source attached on the back-side of the receiver array off the 48th channel) because of the longer distance between the impact point and the array. There was a total of 32 reverse-shot records for the “2020-09-y01_19_12.TDMS” data set that contained a total 100 records, while all 50 records were from the forward shots for the “2020-09-01_19_24.TDMS” data set.

First, the AUTO Mute algorithm previously developed and tested on a limited amount of field records (e.g., data sets from July and August) has been tested on these two data sets to evaluate the accuracy in the muting windows for air-wave and pre-Lamb waves. Results for all 150 records showed the AUTO Mute was highly effective when the raw records and the muted records were visually
compared side-by-side. These comparisons of seismic records are presented in this report. Second, data qualities are examined in dispersion images to valuate the effective recording of Lamb waves. While all reverse-shot records did not  have any Lamb waves in the target frequency bands that are needed for the velocity (Vs) and thickness (H) evaluations (e.g., 10 kHz – 30 kHz), 54 forward-shot records (out of 68 such records) were in good quality that can allow a reliable Vs and H evaluations for the “2020-09-01_19_12.TDMS” data set. On the other hand, 39 records (out of 50) were in good quality for the “2020-09-01_19_24.TDMS” data set. These results put the “accept” rate of 80% and 78% for the two respective data sets. 

Study of the optimum configuration for the 1D array has been one of the main goals of using the old (48-channel) system to collect field data under various different conditions (e.g., different section of the road, different temperatures, different source characteristics, etc.). Based on the test results from a limited amount of field data sets (e.g., the ones from July and August), it has been tentatively assumed that the 16-channel array with 3 times receiver spacing (i.e., 3dx) would be the most optimum configuration. With relatively vast amount of field records now available (i.e., 150), the test has been repeated to reinsure this 3-dx-16-channel assumption. This time, the AUTO Mute was applied to both full 48-channel records and the resampled 3dx-16-channel records to examine if there is any reduction in the effectiveness due to the reduced number of channels. The AUTO Mute results of 48-channel and 3dx-16-channel records  are presented side-by-side in this report for visual comparison.

Then, dispersion images for the 3dx-16-channel records were obtained and  compared side-by-side with those from the full (48) channel records by displaying them for the “2020-09-01_19_12.TDMS” data set in this report. No noticeable visual differences were observed from both the raw-seismic-data and the dispersion-image comparisons for the target frequency band (e.g., 10 kHz – 30 kHz). However, more systematic comparisons will be made in the near future by comparing key analysis results (e.g., Vs and H) from both data sets.

TDMS Data Format

The C++ codes written by a freelancer have been refined so that they can be incorporated into the ParkSEIS-HMA software module without any clash. It seems the codes were written in C codes (instead of C++) because the NI  library components support only C codes. Some parts of the codes were restructured so that they contain only C components without foreign C++ components (e.g., classes). It seems that they are properly working, at least, with all TDMS files collected so far. The new round of testing will be executed
when the TDMS file of finalized structure is collected by using the new system for 1D array (sometime in November). 

Progress Table (Prime Contractor - Park Seismic LLC) - September, 2020

Progress Table (Sub Contractor - Norrfee Tech) - September, 2020

Summary

Quarterly (Q3) Report

The 3rd quarterly report has been prepared and submitted. It is posted here.

TDMS Data Conversion Module Completed

The module in ParkSEIS software that converts a TDMS file into the ParkSEIS (PS) format has been finalized into three (3) types. No further development to accommodate other variation of the TDMS format will be attempted. 

Modules for Velocity (Vs) and Thickness (H) Evaluation - Algorithm Completed

The modules in ParkSEIS that can evaluate the shear-wave velocity (Vs) and thickness (H) of an HMA layer automatically have been completed in  algorithmic development. They have been tested through both numerical  modeling and actual field data sets in accuracy and reliability. The modules will be further tested and improved once the 1D acquisition system being developed is available for more extensive field measurements within the next few months (e.g., November - December). The algorithmic development is further explained in this report. Two methods (Methods I & II) are developed as temporary options for the calibration and evaluation purposes through further field tests. Results obtained from a set of field data (September 1, 2020) are displayed below. The data set includes seismic measurements at 100 consecutive points along the road (approximately 100-m long distance) by  using a field approach presented here. 

The values of velocities (Vs’s) from the two methods (I & II) fall within a reasonable range of HMA layer (e.g., 1300 m/s ≤ Vs ≤ 1500 m/s). The overall  variation trend is quite smooth, indicating measured values from both methods are realistic. Velocities (Vs’s) from both methods are similar approximately within 1.0% difference. The Vs’s from Method II are slightly higher than those  from Method I (approximately by 1%). The S/N values from Method II, however, are significantly higher approximately by 5%. In consequence, Vs’s from Method II are believed to be more accurate. On the other hand, thicknesses  (H’s) evaluated from the two methods are different approximately by as much as 30% overall. In addition, the changing trends are fairly abrupt and irregular,  indicating less realistic trends than those of the velocity (Vs). In consequence, thickness values are much less reliable than the velocity values. The S/N values for H evaluation from the two methods are almost (99%) identical. The H trend from method II, however, seems to be more consistent than that from method I. In this sense, H results from method II are believed to be more reliable, which is consistent with the result from the modeling data. For more details and further information, please see this report. 

Progress Table (Prime Contractor - Park Seismic LLC) - October, 2020

Progress Table (Sub Contractor - Norrfee Tech) - October, 2020

Summary

Field Laptop Computer

A rugged laptop computer has been procured and customized by installing the ParkSEIS-HMA (PS-HMA) software package to be used for field data acquisition and pseudo-real-time on-borad data processing purposes. It has an enhanced computing power, fast solid-state hard drive, embedded touch-screen capability, and sufficient RAM memory (16 GB) to meet the challenging demands. It has been used at Park Seismic to test the PS-HMA software developed to handle the 1D-array data and then shipped to Norrfee Tech for the 1st Joint Field Test (JFT) scheduled on December, 2020. 

ParkSEIS-HMA (PS-HMA) Software Package for 1D-Array Data

The ParkSEIS software package for this project (ParkSEIS-HMA) has been developed to perform the in-field operation of raw seismic data collected by using a 16-channel MEMS-microphone-array acquisition system that has been developed at Norrfee Tech. The algorithmic development that can properly handle different types of wavefields had been finished by the end of last month (October 2020) as summarize in this previous report. Therefore, the Graphical User Interface (GUI) part of the software have been developed and tested  during this month as summarized in this report. 

Most components in the "IN-FIELD" module can perform the following tasks:

1. Open a "New Project" for a new survey.
2. Specify the folder where the raw TDMS data files are saved by the  acquisition hardware system. 
3. Specify the folder where the TDMS data files (*.tdms) are converted and saved into ParkSEIS (PS) format (*.dat) for subsequent analysis by PS-HMA. All intermediate and final analysis results are also saved in this  folder.
4. Some recording-related parameters (e.g., sampling rate, number of  samples, pre-trigger time, etc.) are set here and then passed into the acquisition-system control software installed at the PXI hardware system.
5. Some parameters that can control the automatic detection and analysis of signal wavefields (e.g., surgical mute parameters, wavefield transformation parameters, etc.) can also be specified.
6. Some parameters that can control visualization of the final results of the  shear-wave velocity (Vs) and thickness (H) of the HMA layer can also be specified.
7. Once "armed", it notifies the acquisition control software at PXI system to be ready for data collection.
8. It detects new raw TDMS files being saved by the acquisition system, converts them into PS-format files, and then executes all planned data-analysis steps to finally display the results in a fully automated manner.

All above steps are illustrated by using the Graphical User Interfaces (GUI's) in this report.

Hardware Development (16-Channel 1D-Array System)

The development of a 16-channel data-acquisition system by using MEMS  microphones is under progress. Some circuit boards had to be re-designed to use new MEMS microphones as the old components have been discontinued. The new circuit boards are currently being constructed and the beta version of the 1D acquisition system will be ready for the Joint Field Test (JFT) soon. 

Progress Table (Prime Contractor - Park Seismic LLC) - November, 2020

Progress Table (Sub Contractor - Norrfee Tech) - November, 2020

Summary

ParkSEIS-HMA (PS-HMA) Software Package for 1D-Array Data (Update and Improvement)

The ParkSEIS software package for this project (ParkSEIS-HMA) has been developed to perform the in-field operation of raw seismic data collected by using a 16-channel MEMS-microphone-array acquisition system that has also  been under development at Norrfee Tech. The Graphical User Interface (GUI) part of the software has been summarized in a previous report. The files describing communication between these GUI components and the hardware (PXI) system were previously sent to Norrfee Tech to optimally pass acquisition-related parameters to the PXI system as summarized in this file as well as in this appendix file. This communication part has been continuously modified and refined to meet various changes and updates made in the hardware (PXI) control software. The focus has been on the ability by the ParkSEIS software to notify the PXI system to "ARM" the system so that the acquisition-hardware system can collect data whenever the seismic source makes impact on the HMA surface. This effort has been summarized in this file. The first successful communication between the ParkSEIS and the PXI system had been made via a lab test on December 4, 2020.

Hardware Development (16-Channel 1D-Array System)

The development of a 16-channel data-acquisition system by using ("new") MEMS microphones has been under progress at Norrfee Tech. The beta version of the 1D acquisition system (see photo below) is ready for the Joint Field Test (JFT) soon. As soon as other peripheral components, such as cables, array-holding apparatus, impact-source device, etc., are ready, the JFT will be executed sometime in January, 2021.

Progress Table (Prime Contractor - Park Seismic LLC) - December, 2020

Progress Table (Sub Contractor - Norrfee Tech) - December, 2020