Final Technical Report University of Massachusetts Lowell DE-FE0031895 Final Technical Report Cover Page Recipient Organization: University of Massachusetts Lowell Project Title: Wireless High-Temperature Sensor Network for smart boiler systems Date of Report: February 10, 2026 Award Number: DE-FE0031895 Total Project Costs: [Total $ 432,206] DOE Share: [DOE Share $ 432,206] Recipient Cost Share: [Cost Share $ 0] Project Manager: Andrew Downs Andrew.Downs@netl.doe.gov Project Partner(s): Xuejun Lu, Xuejun_Lu@uml.edu Xingwei Wang, Xingwei_Wang@uml.edu Dongsheng Li, dongsheng.li@amllcct.com Table of Contents 1.0 Executive Summary .......................................................................................................................................1 2.0 Background....................................................................................................................................................1 3.0 Project Objectives..........................................................................................................................................2 4.0 Description of Activities Performed ..............................................................................................................3 1. High-temperature sensor designs .................................................................................................................3 2. 300 MHz Lamb filter design...........................................................................................................................4 3 900 MHz Lamb filter design...........................................................................................................................4 4 2.4 GHz Lamb filter design.............................................................................................................................5 5. Developed the fabrication process for the wireless high-temperature sensor..................................................6 6. Investigate the DoodleLabsยฎ Smart Radio system to set up the wireless high-temperature sensor network testbed....................................................................................................................................................................9 7. Simulate and optimize the wireless high-temperature sensor ....................................................................... 10 8. Fabricate a 2.4 GHz test antenna pattern on a 1x 100mm SiC wafer ............................................................. 15 9. Perform four-point probe measurement of the DC resistance of the antenna pattern ................................. 16 10. Add a Cu tape ground plane and measure the S-Parameters improvement ................................................ 16 11. Measured the wireless coupling between the transmitting and the receiving antennas ............................ 17 12. Demonstrate the wireless temperature sensing working principle with wireless RF antenna mutual coupling ............................................................................................................................................................... 18 13. Sensing and readout patch antennas and the wireless temperature measurement set up......................... 19 14. Develop antenna mutual coupling model ..................................................................................................... 20 13. Set up the wireless sensor network and demonstrate data streaming ........................................................ 27 14. Fabricated new high-temperature sensors based on a 4H-SiC substrate ..................................................... 29 15. Simulate the new designed RF high temperature sensor ............................................................................. 29 16. Set up a high temperature measurement oven ............................................................................................ 30 17. Measure the high temperature sensor response at various temperatures .................................................. 31 18. Test the reliability of the wireless sensor ...................................................................................................... 32 19. Simulate a wireless sensor network .............................................................................................................. 33 20. Position-dependent eavesdropping attack ................................................................................................... 34 21. Simultaneous Authentication of Equals (SAE) authentication process ......................................................... 34 22. AES-128-GCM Encryption and Decryption process ....................................................................................... 35 23. Set up Python server running on a tiny Jetson nano platform with Wi-Fi access network .......................... 36 24. Develop dashboard web service to display temperature sensor data .......................................................... 39 25. Set up Doodle-lab smart radio networks for connecting sensor networks far apart.................................... 40 26. Conduct field test and display temperature sensing data using dashboard http service ............................. 41 27. Set up dual-link connections through WiFi via T-Mobile cellular 5G gateway and Doodle-lab mesh radio . 43 28. Monitoring the traffic interruption when one of the connections is dropped ............................................. 44 29. Calculated the error rate during the service transition and interruption ..................................................... 44 5.0 Conclusions and Recommendations .......................................................................................................... 45 6.0 Lessons Learned ......................................................................................................................................... 45 Final Technical Report University of Massachusetts Lowell DE-FE0031895 1.0 Executive Summary This project aims to develop a new wireless high-temperature sensor network for real-time continuous boiler condition monitoring in harsh environments. Such a wireless high-temperature sensor network enables network-based automatic temperature sensing and data collection, which combined with artificial intelligent (AI) algorithms allow the construction of smart boiler systems with boiling condition management and optimization for significant energy-saving and reliability improvement. 2.0 Background Boilers and furnaces are extensively used virtually everywhere, from thermal power plants to home heating systems. These systems consume the most significant amount of energy. Due to the enormous energy consumption, optimization of the operations of the boilers and furnaces to improve fuel efficiency and reduce energy consumption can lead to tremendous energy-saving and thus bring significant benefits to the Earth and its environment. On can achieve such optimization through analyzing the real-time data collected during operation (e.g., temperature, fuel feed rate, etc.) and applying self-learning algorithms to allow boilers to self- adjust their operating parameters to achieve the highest efficient fuel economy in responding to the energy demands. To achieve smart boiler and furnace systems, real-time and continuous monitoring of the temperatures of boilers and furnaces are essential. Existing boiler temperature monitoring technologies include: (1) Thermocouples Thermocouples consist of an electrical junction formed at the interface of two metals with different chemical potentials. The temperature-dependent chemical potential difference provides a simple and efficient approach for temperature measurement. Thermocouples are widely used in the high-temperature measurement. However, the significant limitations are: (a) limited precision in temperature measurement, where the voltage generated by the thermocouples can change with the purity of the metals, the reference temperature as well as the uncertainties introduced by the alloy manufacturing; (b) material aging, which causes the chemical potential shift and change the temperature reading. (2) Resistance Temperature Detector (RTD) RTD is based on the resistance variation with temperatures of platinum, copper or nickel resistors, i.e. the ๐œ•๐‘… temperature coefficient of resistance ๐›ผ = ~0.0038/โ„ƒ. RTD thermometers are simple and give better ๐‘…๐œ•๐‘‡ stability and accuracy than thermocouples, but they have a limited temperature range of < 600 ๏‚ฐC due to the resistance changes induced by the surface impurity doping in high-temperature boilers. Besides, thermometers have the self-heating error issues and a slower response time of a few seconds. (3) Thermistors Thermistors are made from polycrystalline ceramics such as barium titanate (BaTiO3), which measure the temperatures based on the resistance change of the polycrystalline ceramics. Thermistors can provide a much higher sensitivity than RTDs, but they can only measure a limited temperature range up to 250 ๏‚ฐC, which is not useful in high-temperature measurement in the 1,800 ๏‚ฐC regime. (4) Microstrip patch antenna temperature sensors[1-3] Microstrip patch antennas are narrow-band antennas. The operating frequency depends on the dimensions of the patch antenna. The temperature-induced dimensional change can cause a shift of the operating frequency through both the substrateโ€™s coefficient of thermal expansion (CTE) and the dielectric constant change. Therefore, by measuring the resonant frequency shift, one can measure the temperature. This approach has the advantages of capable of performing simultaneous strain and temperature sensing with a relatively simple structrure[1-3]. However, the Q value of a microstrip patch antenna depends on the loss tangent ๐›ฟ and antenna radiation, which is not high, usually in the range of 100 -500[4]. Such relatively low Q values lead to Page 1 of 45 broad resonant spectral profiles, which limits the temperature measurement resolution. Besides, due to the temperature-dependent material loss tangent ๐›ฟ variation, the patch antenna temperature sensors have a limited temperature range of < 1,000 ๏‚ฐC. (5) Surface acoustic wave (SAW) temperature sensors SAW temperature sensors[5, 6] employ interdigital transducers (IDTs) on piezoelectric substrates such as lithium niobite (LiNbO3), lithium tantalate (LiTaO3), and lead zirconate titanate (PZT)[7]. The transducer can convert acoustic waves to electrical signals and vice versa through the piezoelectric effect of the substrate materials[6]. SAW temperature sensors can be operated wirelessly and passively by taking RF signals from an RF antenna, which enables wireless temperature sensing[8, 9]. The SAW temperature sensor can have a high Q of more than 8000[10], which gives a sharp spectrum for high-resolution measurement. However, since various acoustic modes can be involved in the sensing, the transmission spectra show different side modes, which also change with temperatures and thus complicate the temperature measurement. Also, to our best knowledge, there is no report for temperature sensors with an ultra-high sensing temperature of over 1,800 ยฐC. This is primarily due to the difficulties of getting suitable high- temperature piezoelectric materials[5, 6] and the high-temperature electrodes. 3.0 Project Objectives This DOE project will develop a new wireless high-temperature sensor network for real-time continuous boiler condition monitoring in harsh environments. We will optimize the designs of the wireless high- temperature sensor network, and fabricate the sensor network and characterize its performance and explore their application in smart boiler systems. The Milestones and their completion dates are listed in Table 1. Table 1, Milestone Table updated 03/28/2024. Milestone Task Milestone Title and Planned Verification Subtask Description Completion Date Method #1 1.1 Optimal design of the 12/31/2020 Simulation high- temperature sensor (finished) data #2 1.2 Design the broadband 01/31/2020 Simulation antenna (finished) data #3 1.3 Antenna coupling results 3/30/2021 Simulation through the Lamb filter (finished) data #4 1.4 Design of the sensor 4/30/2022 Simulation network and network (finished) data simulation #5 2.1 A high-temperature sensor 04/30/2023 Hardware (finished) device #6 2.2 A wireless high-temperature 04/30/2023 Sensor sensor network (finished) network #7 3 Test report of the sensor 07/31/2023 Test reports network #8 4 Test report of the reliability 09/01/2024 Test reports of the sensor network (finished) Page 2 of 45 #9 Added Incorporation of security 09/01/2024 Algorithm measures to ensure data (finished) and integrity and prevent simulation unauthorized access. #10 Added Field test of the network's 03/01/2025 Test reports performance. (Finished) #11 Added Validation of the network's 09/01/2025 Test reports fault tolerance. (Finished) 4.0 Description of Activities Performed 1. High-temperature sensor designs The electromechanical transduction factor ฮ“ can be written as: ๐น ฮ“= = ๐‘’31 โˆ™ ๐‘™, (1) V where F is the force, and V is the voltage. The equivalent motional inductance ๐ฟ๐‘š can be expressed as: ๐‘š 2 , (2) ๐œŒ๐‘ค๐‘™๐‘ก where ๐‘š๐‘’๐‘ž = 2 is the equivalent mass, and ๐œŒ is the density of the material. The equivalent motional capacitance ๐ถ๐‘š is: ฮ“2 ๐ถ๐‘š = , (3) ๐‘˜๐‘’๐‘ž ๐ธ๐‘™๐‘ก where ๐‘˜๐‘’๐‘ž = 2๐‘ค . The resonant frequency ๐œ”๐‘š can be written as: 1 1 ๐œ”๐‘š = โˆš๐ฟ ๐ถ = ฮ“2 โˆš๐‘˜๐‘’๐‘ž /๐‘š๐‘’๐‘ž , (4) ๐‘š ๐‘š The equivalent motional resistance is: โˆš๐ฟ๐‘š /๐ถ๐‘š ๐‘…๐‘š = ๐‘„ , (5) The admittance of the Lamb RF filter is: ๐ถ 1 1+ 0 โˆ’๐œ”2 ๐ถ๐‘š ๐ฟ๐‘š +๐‘—๐œ”๐ถ0 ๐‘…๐‘š ๐ถ๐‘š ๐‘Œ= 1 + ๐‘—๐œ”๐ถ0 = 1 , (6) ๐‘…๐‘š +๐‘—๐œ”๐ฟ๐‘š + ๐‘…๐‘š +๐‘—๐œ”๐ฟ๐‘š + ๐‘—๐œ”๐ถ๐‘š ๐‘—๐œ”๐ถ๐‘š The acoustic velocity ๐‘ฃ is related to Youngโ€™s modulus ๐ธ by: ๐ธ ๐‘ฃ = โˆš๐œŒ, (7) ๐‘‘๐ธ The temperature coefficient of Youngโ€™s modulus ๐œ… = โ‰ˆ โˆ’0.02 ๐บ๐‘ƒ๐‘Ž/โ„ƒ2and the coefficient of thermal ๐‘‘๐‘‡ expansion (CTE) is ๐›ผ โ‰ˆ 2.2๐‘๐‘๐‘š/โ„ƒ. In this task, we design the filter in three frequency ranges, i.e. 300 MHz, 900 MHz, and 2.4 GHz. Page 3 of 45 2. 300 MHz Lamb filter design The design parameters are shown in Table 2. Table 2, Design parameters for 300 MHz filter Length Thickness Width (mm) CTE (/K) Meq (kg) e31 (C/m^2) E (Pa) (mm) (mm) 6.0 0.2 0.080 -42e-6 1.25๏‚ด10-6 0.2 7.0๏‚ด1011 Figures 2(a), 2(b), and 2(c) show the simulated filter profiles for the quality factor (Q) of 300, 500, and 1,000 respectively. Figure 1, Filter transmission profilers designed at 300 MHz at various temperatures of 1600 ๏‚ฐC, 1700 ๏‚ฐC, and 1800 ๏‚ฐC (a) Q = 300; (b) Q = 500,; (c) Q = 1000. 3 900 MHz Lamb filter design The design parameters are shown in Table 3. Page 4 of 45 Table 3, Design parameters for 900 MHz filter Length Thickness Width (mm) CTE (/K) Meq (kg) e31 (C/m^2) E (Pa) (mm) (mm) 6.0 0.2 0.0285 -42e-6 1.25๏‚ด10-6 0.2 7.0๏‚ด1011 Figures 2(a), 2(b), and 2(c) show the simulated filter profiles for the quality factor (Q) of 300, 500, and 1,000 respectively. Figure 2, Filter transmission profilers designed at 900 MHz at various temperatures of 1600 ๏‚ฐC, 1700 ๏‚ฐC, and 1800 ๏‚ฐC (a) Q = 300; (b) Q = 500,; (c) Q = 1000. 4 2.4 GHz Lamb filter design The design parameters are shown in Table 4. Page 5 of 45 Table 4, Design parameters for 2.4 GHz filter Length Thickness Width (mm) CTE (/K) Meq (kg) e31 (C/m^2) E (Pa) (mm) (mm) 6.0 0.2 0.0104 -42e-6 1.25๏‚ด10-6 0.2 7.0๏‚ด1011 Figures 3(a), 3(b), and 3(c) show the simulated filter profiles for the quality factor (Q) of 300, 500, and 1,000 respectively. Figure 3, Filter transmission profilers designed at 2.4 GHz at various temperatures of 1600 ๏‚ฐC, 1700 ๏‚ฐC, and 1800 ๏‚ฐC (a) Q = 300; (b) Q = 500,; (c) Q = 1000. 5. Developed the fabrication process for the wireless high-temperature sensor Figure 4 shows the flow chart of the fabrication process for the wireless high-temperature sensor. Page 6 of 45 Figure 4, Flow-chart of the fabrication process. The fabrication process starts with a SiC substrate. The substrate is cleaned using the isopropanol alcohol (IPA) solution, then acetone, and then deionized water (DI) rinse. The wafter is then baked in an oven at 90๏‚ฐC for 30 minutes. After cleaning the SiCsubstrate wafer, the flowing steps will be performed: Step 1: Deposit a SiO2 layer using the plasma-enhanced chemical vapor deposition (PECVD) Table 1 shows the parameters for the SiO2 layer deposition. The total thickness of 400 nm SiO2 layer will be deposited by the PECVD process. Page 7 of 45 Table 5 Parameters for the SiO2 layer deposition SiH4 flow rate N2 flow rate N2O flow rate RF power (W) SiO2 depostion (sccm) (sccm) (sccm) rate (nm/min) 20 380 400 200 310 Step 2: Pattern the SiO2 layer using the photolithography and the wet-etching process (a) The SiO2 layer will be first spin-coated with an S1812 photoresist (PR) at a spin speed of 4000 rpm followed by soft-baking at 90๏‚ฐC for 5 mins. (b) Photolithography UV exposure using a photomask and develop the pattern in a developer. (c) Wet-etching using buffer-oxide etchant (BOE) to leave the central SiO2 rectangular pattern. Step 3: Deposite the metal film outside the central SiO2 rectangular pattern (a) Photolithography define the PR pattern on the centralO2 rectangular pattern (b) E-beam deposition of a thin tantalum (Ta) layer on the device and perform lift-off process to leave the metal surrounding the central SiO2 rectangular pattern. Due to the high melting point (3,017ยฐC) of the Ta metal, the E-beam process needs to be extremely careful and needs to be closely monitored. A special pre- machined slug Ta metal instead of pellets are required as the E-beam target for the metal deposition. After the E-beam deposition, the sample is put into the Acetone to finish the lift-off process. This finishes the sample A fabrication. (c) Repeat Step 1 to 2 to get another SiC wafer (sample B) and deposit a whole layer of Ta through the E-beam deposition process. Step 4: Flip-chip bonding Sample A and Sample B to form the device sample The flip-bond process will be performed at Harvard University CNS nanofabrication center using the Fintech flip- chip bonder (a) put Sample A on the stage (b) flip Sample B and align it with the Sample A (c) heat up the stage to 400 ยฐC and apply 10 Kg force and hold for 15 min and cooling down while holding the force. Step 5: Photolithography patterning the device, deep reactive ion etching (RIE) and removing SiO2 (a) Lapping and polishing the flipped surface of sample B (b) Repeat Step 1 to 2 to create a design device pattern on the polished surface of sample B (c) E-beam deposit nickel (Ni) and lift-off to use the No as the mask for the deep RIE etching. The Ni mask gives a very high deep RIE selectivity of more than 46. The deposited Ni film thickness is 1 ยตm. Multiple depositions is needed to achieve the desired mask for the deep RIE etching. Perform deep RIE etching using the SF6/ O2 plasma. Table 2 shows the parameters of the deep RIE etching. Page 8 of 45 Table 6, Parameters for the deep RIE etching SiF6 flow rate O2 flow rate RF frequency RF power (W) SiC etching rate (sccm) (sccm) (MHz) (ยตm/min) 25 6.7 13.56 300 0.4 (f) Remove the SiO2 by wet etching After the RIE process, the device will be put into a buffered oxide etchant (BOE) to remove the SiO2 under the structure. Step 7: Deposit the top RF antenna layer (a) Perform photolithography UV exposure using the antenna photomask and develop the pattern in a developer. (b) E-beam deposition of thin tantalum (Ta) layer on the device and perform lift-off process to leave the RF antenna on top of the device. The fabrication of the device is finished after the metal lift-off process. 6. Investigate the DoodleLabsยฎ Smart Radio system to set up the wireless high-temperature sensor network testbed We investigated the DoodleLabsยฎ Smart Radio system and plan to use it for the wireless sensor network. Figure 5 shows the mesh network configuration (figure from DoodleLabsยฎ smart radio configuration document). Figure 5, Mesh network configuration using DoodleLabsยฎ Smart Radio system (figure from DoodleLabsยฎ smart radio configuration document). DoodleLabsยฎ Smart Radio system has the following advantages: โ€ข Large coverage with a communication distance > 10 km โ€ข High throughput (~35 Mbps) โ€ข Low latency (< 100 ms) โ€ข MIMO technique high-performance โ€ข Integrated spectrum scanner to mitigate interference for high interference resistance Page 9 of 45 โ€ข Multiple protocols The Smart Radio system supports the following protocols: โ€ข Web GUI โ€ข SSH โ€ข JSON-RPC API โ€ข SNMP We will manage the mesh network using the Web GUI through a web browser. The communication band can be 2.4 GHz and 900 MHz. In this project, we will use the 2.4 GHz band with a 15 MHz bandwidth. The IP addresses of the sensor network will be set up by the DHCP server in the network. All the other sensor nodes will be set up as the DHCP clients. Each ultra-high temperature sensor is connected to one of the GPIO ports on the evaluation board, which is connected to one of the smart radio stations. Figure 6 shows the evaluation board and its ports that can be accessed by the ultra-high temperature sensors (figure from DoodleLabsยฎ evaluation board specification document). Figure 6, Evaluation board and its ports (figure from DoodleLabsยฎ evaluation board specification document). We have ordered the DoodleLabsยฎ Smart Radio system. After the parts arrive, we will set up the testbed of the wireless sensor system based on the DoodleLabsยฎ Smart Radio system. 7. Simulate and optimize the wireless high-temperature sensor Figure 7 shows the schematic structure of the mesh network simulation. The WSN network consists of the sensor nodes and a gateway node, which connects the sensor nodes with external networks. Page 10 of 45 Figure 7, Schematic diagram of the Mesh network configuration. The WSN network consists of the sensor nodes and the gateway node. The WSN communication protocols are based on IEEE 802.15.4 with a data transmission rate of 250 kilobits per second (kbps). The data are transmitted through packages. Figure 8 shows the components in a data package, which includes two leading guarding time slots of ๐‘‡๐‘” = 100 ๐œ‡๐‘ , an 8-byte header followed by a 20 to 120-byte data, a response time slot of ๐‘‡๐‘” = 100 ๐œ‡๐‘ , and a 12-byte acknowledgment. Figure 8, Components of a data package. As shown in Figure 2, the data size varies from 20 bytes to 120 bytes. The time for transmitting the data ๐‘‡๐ท๐‘Ž๐‘ก๐‘Ž is: ๐‘‡๐ท๐‘Ž๐‘ก๐‘Ž = 32๐‘€๐ท ๐œ‡๐‘  (8) where ๐‘€๐ท is the number of bytes in the data package. The total time for a package including the guarding time and the response time is: ๐‘‡๐‘†๐‘™๐‘œ๐‘ก = (684 + 32๐‘€๐ท )๐œ‡๐‘  (9) The data package size needs to be optimized based on the power consumption and SNR. The power consumption for each sensor node includes the power consumption of the PCB board ๐‘ƒ๐‘ƒ๐‘Ÿ๐‘œ๐‘. = 5.4 ๐‘š๐‘Š, and power consumption of the processors, ๐‘ƒ๐‘ƒ๐ถ๐ต = 6.0๐‘š๐‘Š, plus the antenna transmission (Tx) power ๐‘ƒ๐‘ก๐‘ฅ . The Tx power depends on the desired SNR of the network and the loss of network from the transmitter to the receiver ๐ฟ๐‘–๐‘— (๐‘‘๐ต), where i and j are the transmitter and the receiver, respectively. Page 11 of 45 ๐ฟ๐‘–๐‘— (๐‘‘๐ต) = โˆ’20๐‘™๐‘œ๐‘”10 (๐ท๐‘–๐‘— /๐ท0 ), (10) where ๐ท๐‘–๐‘— is the distance from the transmitter to the receiver, and ๐ท0 is the reference distance for the 0 dB loss. The SNR at the receiver node in dB scale is thus: ๐‘†๐‘๐‘…๐‘Ÿ (๐‘™) = ๐‘ƒ๐‘ก๐‘Ÿ (๐‘™) โˆ’ ๐ฟ๐‘–๐‘— โˆ’ ๐‘ƒ๐‘ , (11) where ๐‘ƒ๐‘ is the noise power of the sensor network. We take a typical value of ๐‘ƒ๐‘ (๐‘‘๐ต๐‘š) = โˆ’93. The probability of error is thus ๐‘๐‘’ (๐‘™) = ๐‘„(โˆš2๐‘†๐‘๐‘…๐‘Ÿ ), (12) The probability of successful data receiving is: ๐‘๐‘–๐‘— (๐‘˜) = 1 โˆ’ ๐‘„(โˆš2๐‘†๐‘๐‘…๐‘Ÿ ), (13) The successful data transmission probability for an n-byte packet is: 8๐‘› ๐‘๐‘–๐‘— (๐‘™, ๐‘›) = [1 โˆ’ ๐‘„(โˆš2๐‘†๐‘๐‘…๐‘Ÿ )] , (14) The successful transmission probability for 12-byte acknowledgment packet is: 96 ๐‘๐‘—,๐‘– (๐‘˜, 12) = [1 โˆ’ ๐‘„(โˆš2๐‘†๐‘๐‘…๐‘Ÿ )] , (15) To ensure the successful transition of a whole data package, the repeat times for the data package is: 1 ๐‘€ = ๐‘ (๐‘™,๐‘›)๐‘ (๐‘˜,12), (16) ๐‘–๐‘— ๐‘—,๐‘– The total energy consumption for the successful transmission of one n-byte packet data package is: = 43.8 ๐œ‡๐ฝ + ๐‘€[๐‘ƒ๐‘ก๐‘ฅ (๐‘™)๐‘‡๐‘‘๐‘Ž๐‘ก๐‘Ž + ๐‘ƒ๐‘Ÿ๐‘ฅ (๐‘™)(๐‘‡๐‘†๐‘™๐‘œ๐‘ก โˆ’ ๐‘‡๐‘‘๐‘Ž๐‘ก๐‘Ž )], (17) From Eq. (17), the energy consumption for a -byte packet data package depends on the distance between sensor nodes and the package size. Figure 9 shows the bit error rate ๐‘๐‘’ (๐‘™) v.s. the Tx antenna power for various distances between the sensor nodes of L = 120 m, 150 m, and 200 m. An ultra-low bit error rate ๐‘๐‘’ (๐‘™) < 10โˆ’20 can be achieved with a Tx power of ๐‘ƒ๐‘ก๐‘ฅ = โˆ’10 ๐‘‘๐ต๐‘š for a distance of 200 m. Page 12 of 45 Figure 9, bit error rate ๐’‘๐’† (๐’) v.s. the Tx antenna power for various distances between the sensor nodes of L = 120 m, 150 m, and 200 m. Figure 10 shows the energy consumption to transmit a package with different data package sizes for different Tx powers. The energy consumption is lower for the large data package size when the Tx power is higher. Figure 10, Energy consumption to transmit a package with different data package sizes for different Tx powers. Page 13 of 45 Figure 11 shows the data delays v.s. data package sizes for different Tx powers. The delay is low for a high Tx power. Figure 11, data delays v.s. data package sizes for different Tx powers. Figure 12 shows the data delays v.s. data package sizes for noise powers. The package size needs to be small for a high noise power environment to reduce the delay. Figure 12, Delays v.s. data package sizes for noise powers. Page 14 of 45 8. Fabricate a 2.4 GHz test antenna pattern on a 1x 100mm SiC wafer The fabrication process of is showing below: โ€ข HMDS adhesion layer. โ€ข nLOF 2070 baked at 112.5ยฐC and developed for 6 minutes. โ€ข RIE to remove adhesion layer and descum in open areas. โ€ข Cu/Ti sputter deposition. Cu sputtered at 100W RF for 21000s. โ€ข Lift-off with remover 1165 and solvent clean. โ€ข RIE to descum. โ€ข Dice into antenna dies. Table 7 lists the parameters of the fabrication process. Table 7, Antenna fabrication parameters Figure 13 shows a picture of the fabricated antenna connected to an PNA for testing. Figure 13, Picture of the fabricated antenna connected to an PNA for testing. Page 15 of 45 9. Perform four-point probe measurement of the DC resistance of the antenna pattern Four point measurements were performed on the SiC and on the antennas to ensure the DC characteristics of the fabricated antennas. Table 8 and Table 9 list the sheet resistance and the resistivity of the SiC and the antennas, respectively. Table 8, Antenna fabrication parameters SiC above top of antenna R (ฮฉ/โ–ก) Rs (ฮฉโˆ™cm) V/I Thickness (ยตm) 1.70E+08 5.96E+06 3.99E+08 350 8.16E+08 2.86E+07 1.87E+08 350 2.73E+10 9.57E+08 1.38E+10 350 8.47E+09 2.96E+08 1.94E+09 350 Table 9, Antenna fabrication parameters Cu/Ti at top of antenna R (ฮฉ/โ–ก) Rs (ฮฉโˆ™cm) V/I Thickness (ยตm) 1.76E-02 4.00E-06 4.47E-03 2.3 1.80E-02 4.00E-06 4.80E-03 2.3 1.74E-02 4.00E-06 4.38E-03 2.3 10. Add a Cu tape ground plane and measure the S-Parameters improvement The antenna performance can be improved by adding a Cu ground plane. Figure 14 shows the comparison of the measured S11 parameters of the fabricated antenna with and without the Cu-backplane. The Cu ground plane can improve the S11 parameters. Page 16 of 45 Figure 14, Comparison of the measured S11 parameters of the fabricated antenna with and without the Cu- backplane.. 11. Measured the wireless coupling between the transmitting and the receiving antennas We also measured the wireless coupling between the wirelss antennas. Figure 15 shows experimental set up for the wireless coupling measurements using a PNA network analyzer. Figure 15, Experiment setup for the antenna wirelss coupling. Figure 16 shows the measured S11 parameters of the transmitting antenna with and without the wireless coupling. The coupling can be clearly seen from the S11 papameters. Page 17 of 45 Figure 16, Measured S11 parameters of the transmitting antenna with and without the wireless coupling. 12. Demonstrate the wireless temperature sensing working principle with wireless RF antenna mutual coupling Figure 17 shows the wireless temperature sensing test set up. Two RF antennas were used in the test. One antenna was placed on a hot plate. This is the temperature sensing antenna (i.e. Antenna #1). It is a passive antenna without any RF cables or connectors. The other antenna was set a distance away from the sensing antenna. This is the temperature-reading antenna (i.e. Antenna #2). The temperature-reading antenna is connected to a network analyzer to read the S11 parameter shift due to the coupling from the temperature- sensing antenna. We put a ceramic plate between the two antennas to demonstrate the invisible wireless temperature reading without the capability to penetrate the dielectric layers. Figure 17, Experimental setup for wireless temperature sensing demonstration. Page 18 of 45 This setup is for the wireless temperature principle demonstration. We will put the high-temperature antenna (i.e. Antenna #1) in a high-temperature oven for the high temperature sensing test. Figure 18 shows the measured S22 parameters of the temperature-reading antenna (i.e. Antenna #2) at different temperature on the temperature-sensing antenna (i.e. Antenna #1). Figure 18, Measured S22 parameters of the temperature-reading antenna (i.e. Antenna #2) at different temperature on the temperature-sensing antenna (i.e. Antenna #1). Figure 19 shows the measured frequency shift of antenna #2 at different tempearures of antenna #1. Figure 19, Measured frequency shift of antenna #2 at different tempearures of antenna #1. 13. Sensing and readout patch antennas and the wireless temperature measurement set up The sensing and readout antennas are both patch antennas operating at a resonant frequency of 2.4 GHz. Figure 20(a) shows an image of the patch antennas fabricated on a RO4730G3 printed circuit board (PCB) [16]. Figure 20(b) provides the detailed dimensions of the patch antennas, with all measurements given in millimeters (mm). Note that for the experiments presented in this paper, the measured temperatures did not exceed 162 ยฐC. Page 19 of 45 Therefore, we continued using the RO4730G3 PCB for the sensing antenna. For high-temperature measurements, we have used high 4H-silicon carbide (SiC) temperature semiconductors. Figure 20, 2.4 GHz patch antenna: (a) picture of the fabricated patch antennas on a RO4730G3 PCB; (b) detailed dimensions. All measurements are in mm. The sensing antenna has the same structure as the readout antenna. The two antennas were aligned in their high-gain direction to maximize signal reception and mutual coupling. 14. Develop antenna mutual coupling model Figure 21 illustrates the schematic of the wireless temperature sensing system using an equivalent circuit model [17] with added mutual coupling voltage sources, where both the sensing and the readout patch antennas are modeled as RLC resonators [17]. The readout (i.e., antenna 1) and the sensing (i.e., antenna 2) antennas are wirelessly coupled through the induced voltages ๐‘‰๐‘š21, and ๐‘‰๐‘š12 in their respective circuits. The sensing antenna detects temperature changes, while the readout antenna measures these changes by observing shifts in the reflection parameter ๐‘†11. The ๐‘†11 reflection parameters were measured on the readout antenna using Keysight PNA N5227B Network Analyzer. The sensing antenna was connected to a 50โ„ฆ termination. Figure 21, Schematic the wireless temperature sensing system using an equivalent circuit model [17] with added mutual coupling voltage sources of ๐‘‰๐‘š21, and ๐‘‰๐‘š12. The sensing and readout antennas are both Page 20 of 45 modeled as RLC resonators [17]. Due to the wireless mutual coupling, the temperature change in the sensing antenna can be wirelessly measured through the shift of the reflection parameter ๐‘†11 in the readout antenna. To obtain the antenna equivalent circuit parameters, the sensing and readout antennas were first individually measured without the presence of the other antenna, i.e. the standalone (SA) antenna. The impedance ๐‘๐ฟ of the equivalent circuits at different frequency ๐‘“ can be expressed [17]: 1 ๐‘…0 ๐‘๐ฟ = ๐‘—2๐œ‹๐‘“๐ฟโˆž + ๐‘—2๐œ‹๐ถ + , ๐‘“ ๐‘“0 (18) โˆž 1+๐‘—๐‘„0 ( โˆ’ ) ๐‘“0 ๐‘“ where ๐ฟโˆž and ๐ถโˆž determine the off-resonant impedances approaching the upper and lower frequency band limits. Because the temperature changes were measured based on the resonant properties of the antennas, the off-resonant parameters ๐ฟโˆž and ๐ถโˆž can be disregarded without affecting the simulation results in the resonant 1 frequency regions. ๐‘…0 is the radiative resistance and ๐‘“0 = is the resonant frequency with ๐ฟ๐ด and ๐ถ๐ด 2๐œ‹โˆš๐ฟ๐ด ๐ถ๐ด being the inductance and capacitance of the antenna, respectively. ๐‘„0 is the Q-factor of the resonator, defined as the ratio of the power stored in the resonator ๐‘ƒ0 to the antenna's radiation power loss ๐‘ƒ๐ด , after ignoring the Joule heating power loss. The reflection parameter ๐‘†11 of the SA antennas is determined by the impedance ๐‘๐ฟ of the equivalent circuits and the characteristic impedance ๐‘0 of the RF probes [18]: ๐‘ ๐ฟ โˆ’๐‘0 ๐‘†11 = , (19) ๐‘ ๐ฟ +๐‘0 Figures 22(a) and 22(b) show the simulated reflection parameter ๐‘†11 profiles of the readout and sensing SA antennas at different temperatures, respectively. The solid curves are the measurement data, and the dashed curves are the simulation results. The simulation data match the measurement, indicating the accuracy of the equivalent circuit model and effectiveness to using the model to get antenna parameters. We also simulated and measured the reflection parameter ๐‘†11 profiles (not shown in the figure) for a few other temperatures between 25 โ„ƒ and 162 โ„ƒ (i.e., 41 โ„ƒ, 85 โ„ƒ, and 120 โ„ƒ) and verified that the simulation results match the experiment data for both the sensing and the readout antennas. Page 21 of 45 Figure 22, Comparison of the simulated and measured ๐‘†11 profiles at the standalone (SA) antenna temperatures of 25 โ„ƒ and 162 โ„ƒ: (a) readout SA antenna; (b) sensing antenna. Solid curves: measurement data, dashed curves: simulation data. The simulation data match the experiment measurement results. The match also shown in other temperatures between 25 โ„ƒ and 162 โ„ƒ (not shown in the figure). Table 10 lists the RLC parameters obtained from the ๐‘†11 profile simulation. The RLC parameters were calculated ๐‘‘๐‘…0 based on the dimensions and the material of the antennas, whereas the temperature coefficients of ๐‘…0 (i.e., ๐‘… ๐‘‘๐‘‡ ) 0 ๐‘‘๐ฟ ๐‘‘๐‘Š and antenna dimensions such as length ๐ฟ and width ๐‘Š (i.e., ๐ฟ๐‘‘๐‘‡ and ๐‘Š๐‘‘๐‘‡) were obtained by doing the curve fitting using the equivalent circuit model. Table 10, RLC parameters obtained from the standalone antenna ๐‘†11 profile simulation Standalone (SA) Antenna parameters Readout antenna Sensing antenna ๐‘ช๐‘จ (pF) 58.1 58.0 ๐‘ณ๐‘จ (pH) 75.7 75.5 ๐‘น๐ŸŽ (ฮฉ) 48.0 44.3 ๐‘ธ๐ŸŽ 69.0 64.2 ๐’…๐‘น๐ŸŽ ๐‘น๐ŸŽ ๐’…๐‘ป (K-1) 0.26 ร— 10โˆ’3 0.26 ร— 10โˆ’3 ๐’…๐‘ณ ๐’…๐‘พ , ๐‘ณ๐’…๐‘ป ๐‘พ๐’…๐‘ป (K-1) 7 ร— 10โˆ’6 7 ร— 10โˆ’6 After obtaining the RLC parameters and the temperature coefficients for both antennas, we used the parameters to model the mutual coupling between the two antennas. The mutual couplings are due to the induced voltage terms ๐‘‰๐‘š12 and ๐‘‰๐‘š21 in the readout and sensing antenna respectively. The ๐‘‰๐‘š12 term indicates the voltage generated in the readout antenna (i.e., antenna 1) by the sensing antenna (i.e., antenna 2). 1/2 4๐œ‹๐ท ๐œ† ๐‘–( +โˆ†๐œ™) ๐‘…01 ๐‘…02 ๐‘‰๐‘š12 = ๐ด๐‘‰๐‘–๐‘› (๐บ1 ๐บ2 )1/2 (4๐œ‹๐ท) ๐‘’ ๐œ† [ ๐‘“ ๐‘“ ๐‘“ ๐‘“ ] , (20) 1+๐‘—๐‘„01 ( โˆ’ 01 ) 1+๐‘—๐‘„02 ( โˆ’ 02 ) ๐‘“01 ๐‘“ ๐‘“02 ๐‘“ Page 22 of 45 where ๐ด is the attenuation factor in signal transmission, ๐บ1 and ๐บ2 are antenna gains, ๐‘‰๐‘–๐‘› is the input voltage, ๐œ† is the wavelength of the RF waves, ๐ท is the distance between the readout and sensing antenna, and โˆ†๐œ™ is the additional phase shift in the transmission and probing measurement. ๐‘‰๐‘š21 has a similar relationship with the input voltage of antenna 2. Since we only measured the ๐‘†11 in the readout antenna, we focus on the ๐‘‰๐‘š12 term in the readout antenna. Assuming the antennas have the same gains, i.e., ๐บ1 = ๐บ2 = ๐บ, the ๐‘†11 of the readout antenna considering the mutual coupling is thus: 1/2 4๐œ‹๐ท ๐œ† ๐‘–( +โˆ†๐œ™) 1 ๐‘ ๐ฟ โˆ’๐‘0 ๐‘…01 ๐‘…02 ๐‘ โˆ’๐‘ ๐‘†11 = ๐ด๐บ ( ) ๐‘’ ๐œ† ( ) (1 + )[ ] + ๐ฟ 0, (21) 4๐œ‹๐ท ๐‘ ๐ฟ +๐‘0 ๐‘ ๐ฟ +๐‘0 1+๐‘—๐‘„01 ( ๐‘“ โˆ’๐‘“01 ) 1+๐‘—๐‘„02 ( ๐‘“ โˆ’๐‘“02 ) ๐‘ ๐ฟ +๐‘0 ๐‘“01 ๐‘“ ๐‘“02 ๐‘“ 4๐œ‹๐ท 4๐œ‹๐ท ๐‘–( +โˆ†๐œ™) The term ๐œ† in ๐‘’ ๐œ† of Eq. (4) represents the phase changes accumulation in the round trip from the readout antenna to the sensing antenna and back to the readout antenna. To verify the effectiveness of the simulation and obtain the parameters of ๐ด and โˆ†๐œ™, we measured the ๐‘†11 of the readout antenna at various sensing antenna temperature of 25 โ„ƒ, 41 โ„ƒ, 85 โ„ƒ, 120 โ„ƒ, and 162 โ„ƒ, at different readout and sensing antenna distances ๐ท of 35 ๐‘š๐‘š, 65 ๐‘š๐‘š, and 165 ๐‘š๐‘š. Figures 23(a), 23(b), and 23(c) show the measured ๐‘†11 profiles of the readout antenna compared with the simulated ๐‘†11 profiles using Eq. (4) at different readout and sensing antenna distances ๐ท of 35 ๐‘š๐‘š, 65 ๐‘š๐‘š, and 165 ๐‘š๐‘š, respectively. The dashed curves are the simulation data, and the solid curves are the measurement data. For each readout and sensing antenna distance ๐ท, ๐‘†11 profiles for temperatures of ๐‘‡ = 25โ„ƒ and ๐‘‡ = 162 โ„ƒ are presented with the blue curves corresponding to ๐‘‡ = 25 โ„ƒ and the red curves to ๐‘‡ = 162 โ„ƒ. Table 11 lists the Eq. (4) model parameters of ๐ด, ๐บ, and โˆ†๐œ™ for different distances ๐ท of 35 ๐‘š๐‘š, 65 ๐‘š๐‘š, and 165 ๐‘š๐‘š. The distances in terms of the RF wavelength are also presented in the table. Table 11, ๐ด, ๐บ, and โˆ†๐œ™ parameters for different distances Model parameters ๐ท = 35 ๐‘š๐‘š (0.28๐œ†) ๐ท = 65 ๐‘š๐‘š (0.52๐œ†) ๐ท = 165 ๐‘š๐‘š (1.32๐œ†) ๐ด 0.024 0.024 0.024 ๐บ 1.1 0.75 3.2 โˆ†๐œ™ (ยฐ) 26 76 353 At the distance ๐ท = 35 ๐‘š๐‘š, the ๐‘†11 minimum shifts from the standalone position (Fig. 3(a)) of 2.401 ๐บ๐ป๐‘ง to 2.411 ๐บ๐ป๐‘ง, whereas for the distance ๐ท = 65 ๐‘š๐‘š and ๐ท = 165 ๐‘š๐‘š, the ๐‘†11 minimums shift to 2.405 ๐บ๐ป๐‘ง and 2.408 ๐บ๐ป๐‘ง, respectively. The larger shift indicates a stronger coupling for shorter distance. At the distance ๐ท = 35 ๐‘š๐‘š (Fig. 4(a)), the simulated and the measured ๐‘†11 profiles are not well matched. However, for distances 65 ๐‘š๐‘š and 165 ๐‘š๐‘š, the simulated ๐‘†11 profiles closely match the measured ones, suggesting that the equivalent circuit model presented in Eq. (4) provides a good approximation for cases of weak mutual coupling but is less accurate for cases of strong mutual coupling. At other temperatures such as 41 โ„ƒ, 85 โ„ƒ, 120 โ„ƒ, the model also shows good agreement with the measurement data for the for distances 65 ๐‘š๐‘š and 165 ๐‘š๐‘š. Page 23 of 45 Page 24 of 45 Figure 23, Comparison of the simulated and measured ๐‘†11 profiles at the antenna temperatures of 25 โ„ƒ and 162 โ„ƒ for various readout and sensing antenna distances ๐ท: (a) 35 ๐‘š๐‘š; (b) 65 ๐‘š๐‘š; (c) 165 ๐‘š๐‘š. The distance of 35 ๐‘š๐‘š shows the largest ๐‘†11 minimum shift, indicating strongest mutual coupling. The model shows accurate agreement with the measurement data at the distances of 65 ๐‘š๐‘š and 165 ๐‘š๐‘š, while the agreement is less precise at the distance of 35 ๐‘š๐‘š. As listed in Table 11, the gain has a standard patch antenna value of ๐บ = 3.2 at the distance of ๐ท = 165 ๐‘š๐‘š, whereas it shows lower values of ๐บ = 1.1 and ๐บ = 0.75 at the distances of ๐ท = 35 ๐‘š๐‘š (0.28๐œ†), and ๐ท = 65 ๐‘š๐‘š (0.52๐œ†), respectively. Please note that the lower ๐บ = 0.75 corresponds to 0.52๐œ†, likely due to destructive interference between the antennas. This occurs when the sensing and readout antennas are positioned at the amplitude minimum of the standing wave patterns generated by multiple reflections. Figure 24 shows a plot of phase delays โˆ†๐œ™ versus the distance ๐ท. A linear relationship is obtained with a correlation of determination ๐‘… 2 = 0.99. The slope ๐‘† = 2.59 ยฐ/๐‘š๐‘š can be written in terms the propagation 2๐œ‹ constant ๐‘˜ = ๐œ† , i.e. ๐‘† = 0.90๐‘˜. Therefore, Eq. (4) can be rewritten as: 1/2 ๐œ† ๐‘ โˆ’๐‘ ๐‘…01 ๐‘…02 ๐‘ โˆ’๐‘ ๐‘†11 = ๐ด๐บ ( ) ๐‘’ ๐‘–(0.45๐‘˜2๐ท+๐œ™0 ) (1 + ๐ฟ 0 ) [ ] + ๐ฟ 0, (22) 4๐œ‹๐ท ๐‘ ๐ฟ +๐‘0 1+๐‘—๐‘„01 ( ๐‘“ โˆ’๐‘“01 ) 1+๐‘—๐‘„02 ( ๐‘“ โˆ’๐‘“02 ) ๐‘ ๐ฟ +๐‘0 ๐‘“01 ๐‘“ ๐‘“02 ๐‘“ where ๐œ™0 = โˆ’77.03ยฐ is the distance-independent phase delay obtained from the y-interception from Fig. 24. The 0.45 factor in the propagation term ๐‘’ ๐‘–(0.45๐‘˜(2๐ท)+๐œ™0 ) might be because of the multiple paths involved in the mutual coupling. Page 25 of 45 Figure 24, Phase delays โˆ†๐œ™ versus the distance ๐ท plot. A linear relationship is obtained with a correlation of determination ๐‘… 2 = 0.99. As shown in Fig. 23, the ๐‘†11 profile is much sharper for the distance ๐ท = 65 ๐‘š๐‘š (0.52๐œ†) with a low ๐‘†11 magnitude minimum of lower than โˆ’ 45๐‘‘๐ต, whereas it is wider for other distances. Since temperature sensing relies on measuring the shift in the ๐‘†11 magnitude minimum, a sharper profile can provide enhanced sensitivity. Even sharper ๐‘†11 profiles with a lower magnitude minimum can be achieved by adjusting the phase delay โˆ†๐œ™ while keeping other parameters constant. Figure 25 shows the simulated ๐‘†11 profiles for various phase delay โˆ†๐œ™. The phase delay can significantly change the ๐‘†11 profiles, making it shaper or wider. Since the wireless temperature sensing relies on detecting the shift in the ๐‘†11 magnitude minimum, a sharper ๐‘†11 profile can enhance the measurement resolution. Figure 25, ๐‘†11 profiles for various phase delays โˆ†๐œ™. โˆ†๐œ™ values can significantly affect the sharpness of the ๐‘†11. A sharper ๐‘†11 profile can help to enhance the resolution of the temperature measurements. Figure 26 shows the measured ๐‘†11 magnitude minimum shifts at different temperatures compared with the simulation results for different distances. The solid lines are the simulation and the points (i.e., dots, triangles and diamonds) are the measurement data points. The sensitivities, measured through the slope of the curves in terms of frequency shift (๐‘˜๐ป๐‘ง) per โ„ƒ temperature change, are indicated next to the respective curves. At the distance of ๐ท = 35 ๐‘š๐‘š (0.28๐œ†), the largest frequency shifts are observed, due to the strong coupling. In Page 26 of 45 contrast, the ๐ท = 165 ๐‘š๐‘š (1.32๐œ†) gives lower frequency shifts, while the lowest frequency shift occurs at the distance of ๐ท = 65 ๐‘š๐‘š (0.52๐œ†), potentially due to the ๐œ†/2 destructive interference coupling. However, the sensitivities (๐‘˜๐ป๐‘ง/โ„ƒ) are similar. Additionally, the distance of ๐ท = 65 ๐‘š๐‘š (0.52๐œ†) provides more consistent readings, benefiting from the higher resolution provided by the sharp ๐‘†11 profiles. Figure 26, ๐‘†11 magnitude minimum frequency shifts at different temperatures for various distances. The solid lines are simulation, whereas the points are the measurement data. The slope of the curves are the temperature measurement sensitivities in terms of frequency shift (๐‘˜๐ป๐‘ง) per โ„ƒ temperature change. 13. Set up the wireless sensor network and demonstrate data streaming We have finished setting up the wireless sensor network. Figure 27 shows a picture of a three-node sensor network. It consists of three sensor nodes each with a Jetson nano edge computer for sensor data processing. The sensors also communicate with the servers through DoodleLabโ€™s smart radio MIMO antennas. Figure 27, Picture of the three-node wireless sensor network with edge-computing for sensor data processing. The sensors communicate with the servers through DoodleLabโ€™s smart radio MIMO antennas. Page 27 of 45 The communication scripts of the wireless sensor network between the clients and the server are shown in figure 28. The initial tests shows high data transmission rate and low latency of < 10 milisoconds (ms). The sensor network will be tested for long distances communication in the next research period. Figure 28, Python script for communication between the senor clients and the servers. Page 28 of 45 14. Fabricated new high-temperature sensors based on a 4H-SiC substrate Figure 29 shows a flow chart of the fabrication process. It is based on the standard silicon CMOS fabrication process. Figure 29, Flow chart of the 4H-SiC high-temperature sesnsor fabrication process. Figure 30 shows the fabricated 4H-Sic high-temperature sensors on a 4H-SiC wafer (not diced). The sensor simulation scheme is also shown in Figure 2 to indicate the application of the sensor. Figure 30, Measured S22 parameters of the temperature-reading antenna (i.e. Antenna #2) at different temperature on the temperature-sensing antenna (i.e. Antenna #1). 15. Simulate the new designed RF high temperature sensor Figure 31 shows the simulated S parameters of the new RF antenna high temperature sensor. The S11 resonant frequency of the new design is located at the resonant frequency of 2.4 GHz. Page 29 of 45 S-Parameters for the SiC Antenna Temperature Sensor 0 -10 Magnitude (dB) -20 -30 S11 S21 S12 S22 -40 2.30 2.32 2.34 2.36 2.38 2.40 2.42 2.44 2.46 2.48 2.50 Frequency (GHz) Figure 31, Simulated S parameters of the newly designed RF high-temperature sensors. 16. Set up a high temperature measurement oven Figure 32(a) shows the schematic structure of the high temperature oven and the RF measurement setup. The front metal door of the oven is replaced with a ceramic door to allow RF signals to be transmitted through the door without a high attenuation. The passive RF sensor inside the oven is a patch antenna with a resonant frequency of 2.4 GHz at room temperature (RT). The passive sensing antenna has no RF connectors. Figure 32(b) shows the pictures of the oven and the probing antenna outside of the oven. The probing antenna is also a patch antenna with the same resonant frequency of 2.4 GHz at RT. It is connected to a vector network analyzer (VNA). Figure 32, Schematic of the high temperature oven and the RF measurement setup. Figure 33 Left: a picture of the high temperature oven. The metal door of the oven is removed and replaced with a ceramic door to allow RF signals to be transmitted through. The sensing patch antenna is placed inside the oven. Right: probing antenna connected to a VNA. Page 30 of 45 Figure 34(a) shows a picture of the sensing patch antenna, which has no RF connector. (a) (b) (a) Picture of the sensing patch antenna. Itโ€™s a passive RF antenna without any RF connectors. (b) Picture of the sensing patch antenna inside the high temperature oven. Figure 34, Sensing patch antenna and the placement inside the high temperature oven. 17. Measure the high temperature sensor response at various temperatures Figure 35 shows the frequency minimum shifts of the probing antenna when the oven is heated up to various temperatures. From the figure, the frequency minimums shift to low frequency as the temperature of the oven increases. The frequency shift is sharp in the temperature range from 100 หšC to 500 หšC. It saturates as the temperature is above 600 หšC. Figure 35, Frequency minimum shifts of the probing antenna at various oven temperatures. Figure 36 shows the minimum magnitudes of the S22 of the probing antenna at different distances between the oven and the probing antenna. It shows a periodic variation with the distance, indicating that the interference depends on the distance. Page 31 of 45 Figure 36, Minimum magnitudes of the S22 of the probing antenna at different distances between the oven and the probing antenna. 18. Test the reliability of the wireless sensor The wireless high temperature sensor had been tested under a period of two weeks. The temperatures of the were cycled from room temperature to above 1000 หšC. The temperature cycling was repeated three times during the two weeks testing period. Figure 37 shows the measured resonant frequencies of the S11 profile of the wireless sensor for three repeated measurements. The wireless senor shows consistent performance after the temperature cycling, which verifies its high reliability. Figure 37, Measured resonant frequencies of the S11 profile of the wireless at different temperatures sensor for three repeated measurements. Page 32 of 45 19. Simulate a wireless sensor network We simulated a network with 100 nodes. The nodes are randomly positioned within the covered area of 100 m x 100 m. Figure 388 shows the plot of the network topology. Each node can communicate with any other nodes within the effective communication distance of 20 m. Figure 38. WSN topology with 100 modes randomly positioned within a 100 m ร— 100 m area. Figure 39 shows the simulated energy consumption of each node, assuming each node transmits 4000 bits, and the transmitting and the receiving powers are both 50 nJ/bit. Figure 39, Energy consumption for each node in the WSN. Page 33 of 45 20. Position-dependent eavesdropping attack Since the nodes of the WSN are randomly positioned, eavesdropping attacks depend on their locations to the nodes. We assume that attacker can capture communication by eavesdropping on a packet when it is within the effective communication distance. Figure 40 shows the simulated eavesdropped packets at different eavesdropping locations. The highest eavesdropped packets happen at the closest locations to the WSN nodes. Figure 40, Simulated eavesdropped packets at different eavesdropping locations. 21. Simultaneous Authentication of Equals (SAE) authentication process To reduce the eavesdropping attack, authentication processes must be implemented. The WSN employes the SAE key exchange processes followed by the AES-128-GCM encryption and decryption process. The SAE key exchange mechanism is also used in WPA3 (Wi-Fi Protected Access 3) to provide secure authentication. In SAE, the two parties in the communication verify the keys used in the encryption and decryption process for the communication. The keys include a private key and public key. The private key of a party (noted as Private KeyA) is generated from the partyโ€™s password using a Harsh 256 function. The public key (noted as Public KeyA) of the party is derived from the private key (Private KeyA). The Public KeyA is shared with the other party (B). Party B will do the same private key and the public key generation process and share Public KeyB with the other party A. The two parties combine their own private keys with the otherโ€™s public keys and check whether they match. If they match, the SAE authentication process is finished. The key exchange for the encryption and decryption processes is then successful. Python has libraries for the SAE authentication process. The hash 256 function can be iterated multiple times to increase the difficulty level to break the keys. Figure 41 shows the time it takes to generate the session key for various iterations. Page 34 of 45 Figure 41, Simulated SAE performance at various iterations. 22. AES-128-GCM Encryption and Decryption process After the key exchange, the information is then encrypted using the keys. The AES (Advanced Encryption Standard) is a symmetric key encryption algorithm. It is standardized by NIST using a 128-bit key. Other length keys such as 192, and 256 can also be used. Such encryption provides high security. The estimated time to brute- force attack WPA3 is calculated to be 10,790,283,070,806,013,952.00 years. In the AES-128-GCM Encryption process, the plaintext is first encrypted in blocks using AES in counter mode (CTR), where each block is combined with a unique counter value. The combined data is then encrypted with the AES key to produce the ciphertext. A unique value (nonce) is used during each encryption operation to ensure that the encryption process will produce different encrypted texts for the same plain text. Using Galois mode, an authentication tag is also generated at the same time. This ensures the integrity and authenticity of the encrypted texts. Using Matlab and importing Java cryptography classes below, we simulated the encryption and decryption times for a plaintext = 'This is a secret message.' โ€ข import javax.crypto.Cipher; โ€ข import javax.crypto.spec.SecretKeySpec; โ€ข import javax.crypto.spec.GCMParameterSpec; Figure 42 shows the simulated encryption and encryption and decryption times. Figure 42, Simulated encryption and decryption times for a plaintext = 'This is a secret message.' Page 35 of 45 23. Set up Python server running on a tiny Jetson nano platform with Wi-Fi access network We use a small computer Jetson nano running Ubuntu 20.04 LTS (Linux). It offers a robust embedded sensor network platform. The Jetson nano runs as the server, whereas the sensors are clients. The sensors communicate to the server through WebSocket, which can update the sensor data in real-time. The Jetson Nano WebSocket server runs server.py and hosting dashborad.html. Figure 43 shows the schematic connection of the sensor network. Figure 43. Block diagram of the portable wireless sensor using wireless client-server connection. The codes of server.py are below: Page 36 of 45 The codes of sensor clients are below: Page 37 of 45 Page 38 of 45 24. Develop dashboard web service to display temperature sensor data The sensor data will be displayed for any device connected to the server via http port 8080 service. The dashboard provides data for any device visit the site through http. The sensor data is updated in real-time. The dashboard html codes are listed below. The dashboard.html file is hosted on the Jetson nano in the same directory as server.py. Page 39 of 45 25. Set up Doodle-lab smart radio networks for connecting sensor networks far apart We connect two sensor networks using Doodle-lab smart radio to form a big network covering a larger area. Figure 44 shows the schematic of the sensor network. The Jetson nano functions as both the WebSocket server and the data traffic gateway to route the data from one local sensor network to another sensor network connected through the Doodle-lab smart radio. Figure 44, Block diagram of the wireless sensor networks connected using Doodle-labs smart radio. Page 40 of 45 26. Conduct field test and display temperature sensing data using dashboard http service We conducted field tests of the wireless sensor networks connected using Doodle-lab smart radio networks. Figure 45(a) and Figure 45(b) show the wireless temperature sensors #1, and #2 in a bake oven and in a high temperature oven a basement with ESP32 wireless clients. Figure 45, Pictures of wireless temperature wireless sensor clients: (a) Sensor 1; (b) Sensor 2. Figure 46 shows the field wireless connection through Doodle-lab smart radio. Figure 46, Field wireless sensor networks connected through Doodle-labs smart radio. Page 41 of 45 Figure 47 shows the real-time temperature measurement messages captured by the WebSocket server running on Jestson Nano. Figure 47, Screen shot of the real-time temperature measurement messages captured by the WebSocket server. Figure 48 shows the Dashboard temperature reading chart that is accessible through the http port 8080 service. Page 42 of 45 Figure 48, Dashboard temperature reading chart that is accessible through the http port 8080 service. 27. Set up dual-link connections through WiFi via T-Mobile cellular 5G gateway and Doodle-lab mesh radio To ensure fault-tolerant network connection, we set up dual-links for the sensor networks. One link is Wi-Fi link through T-Mobile cellular 5G gateway, and the other link is through Doodle-lab mesh radio. Figure 49 shows the schematic connections of the sensor networks. Figure 49. Dual-link connections through T-Mobile 5G gateway and Doodle-lab mesh radio. Page 43 of 45 28. Monitoring the traffic interruption when one of the connections is dropped The Wi-Fi link through T-Mobile cellular 5G gateway is a redundant data link to test the sensor fault- tolerance capability of the sensor network when the Doodle lab smart radio. We conducted the data transmission tests when one of the connections was dropped. We monitored the missing (undelivered packages) through the IO port of one node using Wireshark. Figure 50 shows the number of data packages transmitted (dark grey trace) and the number of undelivered packages (red trace) through the IO ports. From the figure, the network can still successfully transmit data over the link when one of the links was dropped. The data traffic was automatically switched to the other link. This demonstrated the fault-tolerant sensor network connections with the dual-links. However, the number of undelivered packages (red trace) increased at the switching time when one of the links was dropped. The number of undelivered packages was higher when the Wi-Fi T-Mobile cellular Gateway was switched off. Figure 50, Number of data packages transmitted (dark grey trace) and the number of undelivered packages (red trace) through the IO ports. The network can still successfully transmit data over the link when one of the links was dropped. This demonstrated the fault-tolerant sensor network connections with the dual-links. 29. Calculated the error rate during the service transition and interruption The error rate is defined as: ๐‘ ๐‘…๐‘’๐‘Ÿ๐‘Ÿ = ๐‘๐‘™, (23) ๐‘ก where ๐‘๐‘™ and ๐‘๐‘ก are the number of the lost packages and the total number of transmitted packages, respectively. Table 12 lists the error rates ๐‘…๐‘’๐‘Ÿ๐‘Ÿ of the sensor network during the service transition and interruption. Table 12, Error rates ๐‘…๐‘’๐‘Ÿ๐‘Ÿ of the sensor network during the service transition and interruption Doodle-lab smart Transition/Stable stage Error rate Cellular Wi-Fi Status radio status ON ON Stable < 10โˆ’5 OFF ON Transition ~ 10โˆ’1 OFF ON Stable < 10โˆ’5 ON OFF Transition ~ 5 ร— 10โˆ’2 ON OFF Stable < 10โˆ’5 Page 44 of 45 5.0 Conclusions and Recommendations In conclusion, we have successfully completed the tasks for this project and developed a wireless high temperature sensor capable of measuring more than 1000 หšC without needing any cable or a transparent window. We also built a wireless sensor network with real-time data sharing and multi-platform data access, i.e., on-demand data request, http dashboard, smart phone APP. We also demonstrated low package loss and low network latency and Implemented SAE data authentication, encryption and decryption processes, calculated the time needed for the processes (a few ms). Additionally, we have successfully demonstrated data routing and fault-tolerance data connection. This research has set up a solid technical foundation for high-temperature sensor network. It also enables automatic high accuracy boiler data collection for subsequently data analysis using artificial intelligence (AI) algorithms to achieve smart boiler systems with extremely high efficiency to achieve tremendous energy saving. In addition to the boiler systems in power plants and home heating systems, the wireless high-temperature sensor network can also find important applications in wireless non-destructive monitoring engines for hypersonic vehicles and space aircraft. The project advances technology and student training in the following areas: โ€ข High-temperature materials โ€ข Advanced manufacturing of integrated RF sensors โ€ข Wireless sensor network communication and cloud-computing with high security โ€ข AI-enabled smart systems for high efficiency The research also trained graduate students in science and technology, especially in the strategic areas in improving the efficiency of fossil energy and related applications such as power plants, home heaters, and AI- enabled smart boiler systems with significantly enhanced efficiency and reliability. 6.0 Lessons Learned During the project, we encountered several technical and programmatic challenges, including student qualification requirements, fabrication constraints, and COVID-induced schedule pressures that required careful task plan and prioritization. One key obstacle was aligning fabrication processes with test measurement inputs to provide feedback to the device fabrication while maintaining technical rigor and meeting milestone deadlines. We addressed these challenges by conducting careful project task planning, periodic technical reviews, improving documentation clarity, and establishing more frequent communication checkpoints to proactively resolve issues. When technical uncertainties arose, we performed focused studies and validation tests to reduce the uncertainties prior to final confirmation. We also develop theoretical models to model the system measurement results. We are confident about the results when the theoretical model matches our measurement data. Through this process, we reinforced the importance of early risk identification, prioritized task management, and open communication across teams. These lessons have strengthened our approach to planning, task alignment, and proactive risk mitigation in subsequent phases of the project. Page 45 of 45