Two-parameter modulation and feature fusion method of electromagnetic excitation wireless QCM-D

Abstract

The invention discloses an electromagnetic excitation wireless QCM-D (Quartz Crystal Microbalance) two-parameter modulation and multi-modal feature fusion method for high-specificity recognition of volatile organic compounds, and designs a synchronous modulation method of frequency-voltage two parameters of driving voltage for a wireless electrodeless quartz crystal microbalance WE-QCM-D system. By extracting the dissipation variable quantity of sound wave energy generated by the change of driving voltage under different excitation frequencies on the surface of a quartz crystal, the steady-state response dimension of WE-QCM-D to VOCs is improved, and the fusion of steady-state and dynamic response characteristics is carried out by combining machine learning and a neural network algorithm. High-selectivity recognition of various similar VOC gases and mixed VOCs samples is achieved, and the recognition rate reaches 95% or above.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of gas phase chemical detection, in particular to a dual-parameter modulation method for the discrimination of volatile organic compounds (VOCs) on an electromagnetically excited wireless QCM-D, aiming to utilize advanced wireless quartz crystal microbalance QCM-D for the discrimination of volatile organic compounds VOCs. BACKGROUND

[0002] Crystal microbalance with dissipation monitoring is a highly sensitive sensing technology that utilizes the piezoelectric properties of quartz crystals to detect small changes in mass and viscoelastic properties at the nanoscale. Since its development in the 1960s, QCM technology has been widely applied in scientific and industrial fields due to its ability to measure nanogram-level mass changes, significantly surpassing the sensitivity of traditional electronic balances. By converting changes in surface mass to changes in resonant frequency, QCM systems can provide accurate, real-time, and label-free detection data. These characteristics make QCM-D particularly valuable in monitoring chemical and biological interactions and detecting volatile organic compounds in environmental, industrial, and medical scenarios.

[0003] Traditional QCM-D systems usually rely on mechanical or electrical excitation methods and operate at a fixed resonant frequency. These systems mainly interpret the interaction of volatile organic compounds through changes in resonant frequency (Δf), which limits their ability to distinguish VOCs with similar molecular weights or overlapping binding characteristics. To address this issue, dissipation monitoring (ΔD) was introduced to complement frequency-based sensing, enabling dual-parameter detection and increasing information about viscoelastic changes. However, traditional dual-parameter systems are usually bulky, complex, and rely on wired connections, limiting their portability and practical deployment in field environments.

[0004] The introduction of electromagnetic excitation provides a wireless alternative by utilizing the principle of Lorentz force to excite quartz crystals. This eliminates the need for direct electrical contact, supporting structural simplification, system miniaturization, and remote operation within the four harmonic frequency limits, thereby limiting the ability to distinguish volatile organic compounds, especially VOCs with similar molecular sizes.

[0005] The applicant has applied for a Chinese invention patent CN110441182A, which discloses an electromagnetic excited wireless QCM-D multi-frequency time division multiplexing gas detection virtual array system and method, which comprises a quartz crystal wafer placed in the middle of the bottom wall of a gas detection chamber, an excitation coil and a receiving coil arranged outside the bottom of the gas detection chamber, a signal generator for generating a driving signal, an impedance matching network for transmitting the driving signal to match the impedance of the excitation coil, a multi-band narrowband amplification filter circuit for filtering and amplifying the detection electrical signal in the receiving coil, and an experimental controller running on a computer for controlling the multi-frequency point switching work of the virtual array system; the receiving coil receives the detection electrical signal in the alternating electromagnetic field generated by the quartz crystal wafer; the gas detection chamber is connected with a gas flow controller for controlling the nitrogen flow rate into the gas detection chamber, so as to maintain the gas flow rate at a desired constant value. On this basis, the applicant further researches and develops a dual-parameter modulation method suitable for the electromagnetic excited wireless QCM-D system. This novel method realizes the synchronous modulation of resonance frequency and energy dissipation by accurately controlling the excitation waveform and magnetic field conditions. By expanding the sensing feature space, the present application can capture unique volatile organic compound (VOC) response characteristics in multiple fields without additional hardware components or sensor arrays. This method promotes more accurate identification and classification of volatile organic compounds by utilizing dynamic excitation parameters and programmable control. SUMMARY

[0006] The present application provides a dual-parameter modulation method for wireless quartz crystal microbalance, which creates a high-dimensional virtual sensor array by adopting a mixed multi-frequency and multi-amplitude modulation strategy to achieve efficient identification of volatile organic compounds (VOCs). The modulation method is a synchronous modulation of the working frequency and excitation voltage of WE-QCM-D. This dual modulation strategy significantly enriches the information content of sensor response, thereby enhancing the pattern recognition ability. The performance of the system is verified through experiments involving ten different volatile organic compounds and various natural fruit aromas. The results show that the method effectively captures the distinguishing sensor features from a single oscillator and uses advanced feature extraction and machine learning algorithms for analysis. This improves the VOC gas selectivity and concentration robustness in complex chemical environments compared to traditional single-parameter WE-QCM-D systems.

[0007] The specific technical solutions are as follows: The application discloses a method for high-specificity recognition of volatile organic compounds (VOCs) by electromagnetic excitation wireless QCM-D (WE-QCM-D) dual-parameter modulation and multi-modal feature fusion.

[0008] In some embodiments, the method for electromagnetic excitation wireless QCM-D dual-parameter modulation and multi-modal feature fusion, the driving amplitude modulation only starts after the inductive response of the wireless QCM-D sensor to the volatile organic compound analyte reaches equilibrium, thereby ensuring stable feature extraction during the amplitude cycle.

[0009] In some embodiments, the method for electromagnetic excitation wireless QCM-D dual-parameter modulation and multi-modal feature fusion, the time interval between every two adjacent amplitudes is 3-24 seconds.

[0010] In some embodiments, the method for electromagnetic excitation wireless QCM-D dual-parameter modulation and multi-modal feature fusion, the WE-QCM-D is driven to work alternately at four resonance frequency points from the fundamental frequency to the seventh harmonic frequency by sweep frequency, and the driving voltage is modulated during the steady-state response stage of the sensor for each selected harmonic frequency.

[0011] In some embodiments, the method for electromagnetic excitation wireless QCM-D dual-parameter modulation and multi-modal feature fusion, the amplitude of the excitation voltage is modulated in a step signal mode. Specifically, when the response of the sensor to the volatile organic compound analyte reaches stability, the driving voltage is gradually increased from 1V to 10V in a step signal mode (for example, with an increment of 1V), and then symmetrically reduced back to 1V, and the frequency shift and energy dissipation change corresponding to each pair of frequency-amplitude are extracted and used as multi-dimensional features for analyte recognition.

[0012] In some embodiments, the method for electromagnetic excitation wireless QCM-D dual-parameter modulation and multi-modal feature fusion, during the detection of volatile organic compounds, the real-time changes of the sensor response signals including the resonance frequency Δf and the energy dissipation ΔD are collected under each frequency-amplitude combination, and the signals are processed by advanced time-domain and frequency-domain feature extraction algorithms, including decay envelope tracking, harmonic distortion analysis and phase shift detection.

[0013] In some embodiments, the electromagnetic excitation wireless QCM-D dual parameter modulation and multi-modal feature fusion method, the frequency change Δf and the energy dissipation change ΔD are continuously monitored and calculated as a dynamic function of time in the transient state phase before reaching equilibrium.

[0014] In some embodiments, the electromagnetic excitation wireless QCM-D dual parameter modulation and multi-modal feature fusion method, the identification of volatile organic compounds is carried out using advanced classification algorithms, including K-nearest neighbor KNN, support vector machine SVM, random forest RF and long short-term memory neural network LSTM, wherein the LSTM model is specially configured to capture time-dependent relationships and extract binding dynamic information from time-resolved sensor responses at different excitation voltages, thereby revealing how the driving amplitude affects the adsorption-desorption behavior of VOC molecules.

[0015] The WE-QCM-D system of the present application can refer to the applicant's prior application of Chinese invention patent CN110441182A, which includes a quartz crystal oscillator enclosed in a gas phase detection chamber, an excitation coil and a receiving coil, which are externally arranged to realize wireless non-contact excitation and sensing of the oscillator, and further includes a signal receiving and conditioning circuit with frequency-specific filtering and amplification unit, a signal acquisition module, and a data processing unit capable of extracting and correlating the frequency change and energy dissipation response of each frequency-amplitude pair. The WE-QCM-D system further includes a non-contact excitation and sensing configuration through a planar spiral coil, which allows wireless electromagnetic excitation of the electrodeless AT-cut quartz crystal resonator placed in the gas phase detection cabin. A signal generation module and an impedance matching network are used to drive the excitation coil, while the receiving coil collects the damped oscillation signal, which is then processed through a frequency-specific narrowband filtering and amplification circuit. A computer-controlled experimental platform based on LabVIEW software is used to manage frequency switching, amplitude control and timing synchronization. During testing, the stimulation frequency of the pulsed sinusoidal signal is modulated from the fundamental resonance frequency to the seventh harmonic of the quartz crystal oscillator in sequence, and the modulation process is controlled by a LabVIEW-based software interface that can achieve automatic multi-frequency switching and time-division excitation.

[0016] A method for detecting volatile organic compounds using a dual parameter modulation wireless QCM-D virtual sensor, comprising the steps of: S1, adjusting the signal generator to set the trigger signal as a sinusoidal excitation pulse signal and configure the gate mode; S2, adjusting the impedance matching network at the highest resonance frequency point to ensure impedance matching between the signal generator and the excitation coil at each harmonic frequency point, so that the quartz crystal resonator can oscillate; S3, connect the gas phase detection chamber to a nitrogen gas source, open the gas flow controller, flush the detection chamber 5 to 6 minutes to expel residual gases and establish a stable baseline atmosphere, then close the gas flow controller; S4, inject the target volatile organic compound sample gas into the detection chamber, allowing the quartz crystal resonator surface to adsorb gas molecules under resonant excitation conditions until the frequency shift (Af) and energy dissipation change (AD) signals reach a steady-state equilibrium; S5, after completing a detection cycle, control the gas flow control valve to introduce nitrogen into the chamber to promote desorption of volatile organic compounds (VOCs) adsorbed on the quartz crystal resonator; allow 5 to 7 minutes of desorption time during which the system passively restores the resonator to its baseline frequency and dissipation level; S6, after desorption, maintain a 10-minute delay to fully purge the chamber; then, inject the next VOC sample and allow 4 minutes of adsorption time to ensure stable interaction between the resonator surface and the analyte; S7, according to the virtual sensor architecture, define virtual channel one by setting the signal generator output to a specific excitation frequency f1 and a corresponding amplitude A1, the amplitude is dynamically controlled to follow a predefined modulation sequence (e.g. a stepwise ramp between 1V and 10V) during the equilibrium phase of analyte binding; repeat steps S4 to S7 to perform consecutive detection cycles and ensure measurement repeatability; S8, the receiving coil detects the attenuated oscillation signal from the quartz crystal resonator, and the LabVIEW-based experimental controller configures a multi-band filter and an amplification circuit to isolate and process the signal at frequency f1; the filtered signal is transmitted to an oscilloscope, and the computer extracts dynamic response parameters, including frequency shift (Af), dissipation change (AD), and amplitude-dependent binding behavior, based on envelope analysis, phase shift deviation, and harmonic distortion; S9, every 3 seconds, the system switches to the next virtual channel by updating the excitation frequency and amplitude settings to f2, A2; f3, A3; and so on, until all defined frequency-amplitude combinations are scanned; this time-division multiplexing strategy constructs a high-dimensional sensor matrix from a single resonator, capturing multi-parameter features for robust volatile organic compound (VOC) identification.

[0017] The wireless QCM-D system of the present application uses a dual-parameter modulation method to identify volatile organic compounds, integrating electromagnetic excitation, time-division multiplexing, multiple harmonic frequencies, and synchronous amplitude modulation. The present application can construct a high-dimensional virtual sensor array from a single resonator, achieving sensitive, selective, and repeatable gas phase detection while minimizing hardware complexity.

[0018] The two-parameter modulation QCM-D system includes an electrodeless AT-cut quartz crystal resonator fixed at the center of the bottom of the gas-phase detection chamber. The resonator is wirelessly driven by a burst-mode sinusoidal excitation signal transmitted through an external excitation coil, while the response signal is captured by a spatially aligned receiving coil. Both coils are designed as planar spiral structures and mounted on the outer surface of the detection chamber to achieve non-contact electromagnetic coupling.

[0019] The programmable signal generator is configured to output a pulsed sinusoidal signal, the frequency of which is automatically swept from the fundamental resonance frequency of the quartz crystal to the seventh harmonic. Meanwhile, during each frequency dwell, the amplitude of the excitation signal is dynamically modulated between 1 V and 10 V, gradually increasing and then decreasing back to 1 V. The system also implements a novel amplitude modulation mechanism, in which the amplitude of the excitation signal can be controlled according to predefined modulation waveforms, such as sinusoidal, step, or ramp modulation, operating in the frequency range of 10-100 Hz. This dynamically amplitude-modulated excitation signal can more deeply probe the viscoelastic properties of adsorbed analytes, providing richer sensing information than traditional frequency-only excitation methods.

[0020] To ensure proper excitation and signal fidelity in this broad modulation spectrum, the excitation signal is transmitted through an impedance matching network. The receiving coil captures the decaying oscillation signals of the resonator, which are then transmitted to a multi-band filtering and amplification circuit. This circuit is managed through a LabVIEW-based software interface, enabling frequency-specific signal conditioning and channel switching. A time-division multiplexing control system ensures that each frequency-amplitude combination is processed sequentially with synchronized timing, creating a virtual sensing channel matrix.

[0021] The modulated oscillation response is captured and processed in real-time by a computer-based controller. The system employs advanced signal processing algorithms for feature extraction, including time-domain decay envelope analysis and frequency-domain harmonic distortion analysis. These techniques can capture phase changes, viscoelastic damping behavior, and nonlinear response characteristics induced by modulation, which are crucial for accurate identification of volatile organic compounds. The extracted features for each frequency-amplitude pair are used to construct a high-dimensional response matrix, which is input into machine learning models such as K-Nearest Neighbors, Support Vector Machines, Random Forests, and Long Short-Term Memory (LSTM) networks to improve selectivity and sensitivity to VOC species.

[0022] To support repeatable measurements, the gas-phase detection chamber is equipped with a gas flow control system, including a mass flow controller and bidirectional check valves. Nitrogen gas is used to purge the chamber before and after sample injection to ensure consistent baseline and prevent contamination between cycles. The unidirectional gas path stabilizes flow conditions and reduces disturbances during sensing.

[0023] The quartz crystal resonator used in the present invention is a bare AT-cut quartz wafer, coated on one side with a sensitive film designed for specific adsorption of volatile organic compounds molecules. After film processing, the resonator has a fundamental frequency of 6 MHz, supporting harmonics including 18 MHz, 30 MHz, 42 MHz and higher frequencies, facilitating multi-frequency electromagnetic excitation in a wireless and contactless manner.

[0024] As a preferred solution, both the excitation and receiving coils are constructed as planar spiral structures and are symmetrically fixed on the external base of the gas-phase detection chamber, enabling stable electromagnetic excitation and signal reception through non-contact coupling with the quartz resonator.

[0025] For sensor positioning and system stability, the gas-phase detection chamber is preferably designed as a sealed flask with a circular recess at the bottom for accommodating the electrodeless AT-cut quartz crystal oscillator. This configuration allows for precise and repeatable placement of the resonator and reduces disturbances from internal gas flow during measurement.

[0026] As a preferred option, the bottom of the gas-phase detection flask is connected to a gas inlet duct connected to an external nitrogen supply. The gas inlet duct is equipped with a first one-way gas flow valve. Upstream of the valve, a gas flow controller is installed to regulate the nitrogen flow rate during chamber cleaning and baseline recovery operations.

[0027] As a preferred solution, the upper neck of the gas-phase detection flask is connected to a gas outlet duct, which in turn is connected in series with a second one-way gas flow valve. The outlet of the gas outlet duct is connected to a volatile organic compound exhaust treatment device to ensure safe and controlled discharge of the analyte after the measurement period.

[0028] As a preferred solution, the mouth of the flask is sealed using a plug or cap that is integrated with the gas injection interface, connected to the gas sampling system to introduce volatile organic compound analytes or fruit aroma samples into the chamber in a controlled manner.

[0029] The use of bidirectional one-way gas flow valves ensures directional control of gas flow within the gas-phase reaction system, minimizes the effects of reverse pressure waves, and improves cleaning efficiency. Gas inlet flow is stabilized by a precision flow controller to maintain consistent conditions during volatile organic compound exposure and nitrogen flushing. The recess machined into the bottom of the flask ensures correct positioning of the quartz wafer, preventing misplacement caused by mechanical impact or turbulent flow, thereby improving test repeatability and signal fidelity.

[0030] The method of the present invention for a dual-parameter modulation wireless QCM-D system based on electromagnetic excitation can specifically include the following steps: 1) Adjust the signal generator to output a gated sinusoidal excitation signal and configure through the LabVIEW-based interface to enable synchronous triggering and modulation control. The signal generator is set to generate burst-mode waveforms that can be sequentially applied across multiple harmonic frequencies.

[0031] 2) Adjust the impedance matching network at the highest operating harmonic frequency to achieve impedance matching between the signal generator and the excitation coil at each selected frequency point (from the fundamental to the seventh harmonic) to enable stable oscillation of the piezoelectric quartz resonator.

[0032] 3) Connect the gas-phase detection chamber to a nitrogen gas source. Open the flow control valve to flush the detection chamber with nitrogen, expel residual gases, and establish a clean baseline environment. Maintain nitrogen flow for 10 to 30 minutes before closing the valve to begin the measurement cycle.

[0033] 4) Inject the sample volatile organic gas into the chamber and allow the analyte to interact with the resonator. Maintain this exposure until the system reaches equilibrium, typically requiring about 4 minutes of adsorption, at which point the frequency change (Δf) and dissipation (ΔD) responses stabilize.

[0034] 5) After the signal stabilizes, initiate the amplitude modulation cycle. Modulate the excitation amplitude from 1 V to 10 V and back down to 1 V in steps, while maintaining each excitation amplitude for 3 to 24 seconds. This probes the viscoelastic and mechanical properties of the adsorbed layer.

[0035] 6) After completing the detection cycle, reopen the nitrogen valve to allow a 5 to 7 minute rinse period to clear residual analyte from the chamber. Continue desorption for a total of 10 minutes until the resonator returns to its original baseline frequency. This prepares for the next sample injection.

[0036] 7) Repeat steps 4) to 6) for subsequent volatile organic compound sample detections to ensure signal repeatability and stable baseline recovery.

[0037] 8) For each measurement, designate a virtual sensing channel for a specific frequency-amplitude pair. Under LabVIEW control, the signal generator outputs the excitation signal for each virtual channel, and the receiving coil wirelessly collects the attenuated oscillatory response.

[0038] 9) The received signal is filtered and amplified through a specific frequency circuit and then transmitted to an oscilloscope and computer interface. Apply feature extraction algorithms to analyze the signal characteristics induced by modulation, including time-domain decay envelope tracking, phase offset deviation, and harmonic distortion patterns.

[0039] 10) Switch to the next virtual channel every 3 seconds, using time-division multiplexing logic. Loop through all defined frequency-amplitude combinations to build a multi-dimensional sensor dataset, capturing both steady-state and transient response characteristics.

[0040] 11) Aggregate all extracted parameter frequency shifts (Df) and dissipations (AD) across the virtual array and input them into machine learning models such as KNN, SVM, Random Forest or LSTM for classification, pattern recognition and quantitative VOC identification.

[0041] 12) The combination of amplitude modulation and advanced feature extraction significantly enhances the discrimination power for volatile organic compounds and complex odor signatures. Through experimental trials involving ten volatile organic compounds and multiple fruit aroma samples, the system achieves precise, repeatable, and high-resolution sensing suitable for real-time environmental and food quality monitoring.

[0042] Compared with the prior art, the present application has the beneficial effects of: The proposed dual-parameter modulation wireless QCM-D multi-frequency time-division multiplexing gas phase detection virtual array system and method realizes non-contact wireless excitation and dissipation monitoring of a quartz crystal sensor by using a single resonator. Through synchronous multi-frequency excitation and amplitude modulation, the system simulates a high-dimensional sensor array without the need for multiple resonators. This greatly simplifies the hardware requirements, reduces the system cost, and enhances the identification ability of the analyte. The modulation-based signal processing method makes it possible to extract more rich information, including phase dynamics and mechanical damping characteristics, thereby expanding the application potential of the QCM-D platform in portable, intelligent, and high-throughput VOC detection. BRIEF DESCRIPTION OF DRAWINGS

[0043] Figure 1 (a) is a time-based modulation pattern of the driving voltage, showing a cyclic step change between 1 V and 10 V, which is applied at the equilibrium phase of the sensor response to simultaneously probe the viscoelastic and mass adsorption properties of the quartz crystal; Figure 1 (b) is the energy dissipation response (AD) extracted by the QCM-D system under amplitude modulation of 1 V to 10 V, the data reveals the nonlinear relationship between the driving voltage and the energy dissipation of the adsorbed molecular layer, providing high-dimensional features for volatile organic compounds and odor classification.

[0044] Figure 2 is the classification result of different fruit odor samples using the dual-parameter modulation QCM-D features derived from Figure 1 (a) and 1(b) are the classification results of different fruit odor samples using the dual-parameter modulation QCM-D features derived from DETAILED DESCRIPTION

[0045] The application will be further described below with reference to the drawings and specific examples. It should be understood that these examples are only used to illustrate the application and not to limit the scope of the application. The operation method in the following examples without specific conditions is generally according to the conventional conditions or according to the conditions suggested by the manufacturer.

[0046] The dual-parameter modulation wireless QCM-D virtual sensor array system based on electromagnetic excitation includes the following components: a computer terminal, a programmable signal generator, an impedance matching circuit, an excitation coil, a receiving coil, a quartz crystal resonator, a multi-band filter and amplification circuit, a digital oscilloscope, a gas flow controller, a volatile organic compound gas phase detection cavity, a gas sample injector, a nitrogen supply tank and a waste gas treatment unit.

[0047] The programmable signal generator is connected to the excitation coil through an impedance matching network, ensuring optimal power transmission and frequency matching at the fundamental frequency of 6 MHz to the seventh harmonic point. The excitation coil is externally located at the bottom of the detection chamber, generating a time-varying magnetic field that wirelessly excites the quartz crystal resonator through electromagnetic coupling. At the same time, the receiving coil is also externally mounted to capture the decay oscillation signal generated by the mechanical response of the resonator.

[0048] It is worth noting that both the excitation coil and the receiving coil are configured as planar spiral structures, and neither of them is in direct contact with the quartz crystal resonator. The resonator itself is an electrodeless AT-cut quartz chip placed in a circular groove at the bottom of the cavity, with a sensitive thin film coated on one side to enhance the adsorption and viscoelastic interaction of gas molecules.

[0049] During operation, the pulsed mode sinusoidal signal is triggered in gate mode and automatically sweeps across multiple harmonic frequencies through LabVIEW-based control. Once the frequency and dissipation response caused by the analyte reaches equilibrium, the system will start the amplitude modulation protocol shown in Figure 1 (a): a step mode from 1 V to 10 V, and then gradually returns to 1 V, maintaining each voltage level for 3 to 24 seconds. This control creates unique mechanical stress characteristics, probing the molecular rearrangement dynamics of the sensor interface, allowing detailed probing of viscoelastic and damping behavior beyond what can be achieved with fixed amplitude excitation.

[0050] The oscilloscope used in this example is a Tektronix model TDS5054B with a high-speed sampling rate of up to 5 GS / s. This performance is sufficient to meet the data acquisition requirements related to the highest operating frequency of the quartz crystal oscillator, 42 MHz, as seen in the seventh harmonic under full spectrum excitation. During each amplitude modulation cycle, the oscilloscope captures the dynamic response of the resonator to the voltage steps shown in Figure 1 (a), especially for Figure 1(b) The energy dissipation features presented in (b). These dissipation responses (ΔD) show voltage-dependent hysteresis phenomena as molecular fingerprints, revealing nonlinear viscoelastic transitions during adsorption and desorption. The oscilloscope not only enables real-time visualization of the electrical signals generated by the resonator oscillations, but also supports waveform storage and exportation functionalities. During each experimental cycle, the oscilloscope is configured to selectively capture the effective segments of the signal decay curves, focusing on the most relevant parts for the dissipation and phase analysis. Then, the collected data are transmitted to the computer, where the experimental controller performs real-time waveform decoding, extraction, and classification.

[0051] The quartz crystal resonator in this embodiment employs an electrodeless AT-cut bare quartz chip. A sensitive recognition layer is applied to its upper surface through a precise dot-coating technique. The resonator has a fundamental frequency of 6.0 MHz, a diameter ranging from 12 mm to 13 mm, and a substrate thickness of approximately 0.3 mm. When subjected to the amplitude modulation mechanism of (a), the resonator's dissipation response (in ΔD) shows analyte-specific hysteresis loops. Figure 1 Figure 1 (b) The energy dissipation features presented in (b). These dissipation responses (ΔD) show voltage-dependent hysteresis phenomena as molecular fingerprints, revealing nonlinear viscoelastic transitions during adsorption and desorption. The oscilloscope not only enables real-time visualization of the electrical signals generated by the resonator oscillations, but also supports waveform storage and exportation functionalities. During each experimental cycle, the oscilloscope is configured to selectively capture the effective segments of the signal decay curves, focusing on the most relevant parts for the dissipation and phase analysis. Then, the collected data are transmitted to the computer, where the experimental controller performs real-time waveform decoding, extraction, and classification.

[0052] The system architecture employs a single physical resonator sequentially operated at multiple harmonic frequencies, implementing a time-division multiplexing strategy to emulate multiple sensing channels - each corresponding to a unique frequency-amplitude pair. The impedance matching network is designed to be harmonic-compatible, such that when tuned to match the highest overtone frequency, it naturally provides a matching impedance at lower harmonics. This feature supports fast frequency switching and consistent energy transmission across the entire operating frequency spectrum.

[0053] The multi-band filtering and amplification circuit is implemented as a software-controlled frequency-selective amplifier bank. The circuit is managed by the LabVIEW-based experimental controller, dynamically reconfiguring to target the specific frequency band corresponding to the current excitation harmonic. During the amplitude modulation phase, the circuit isolates the harmonic distortion components related to the dissipation patterns in (b), enhancing the detection of voltage-dependent viscoelastic transitions. By filtering out out-of-band noise and enhancing the signal-to-noise ratio, it ensures high-gain, low-noise recovery of the oscillation data transmitted from the receiving coil. Figure 1 (b) The energy dissipation features presented in (b). These dissipation responses (ΔD) show voltage-dependent hysteresis phenomena as molecular fingerprints, revealing nonlinear viscoelastic transitions during adsorption and desorption. The oscilloscope not only enables real-time visualization of the electrical signals generated by the resonator oscillations, but also supports waveform storage and exportation functionalities. During each experimental cycle, the oscilloscope is configured to selectively capture the effective segments of the signal decay curves, focusing on the most relevant parts for the dissipation and phase analysis. Then, the collected data are transmitted to the computer, where the experimental controller performs real-time waveform decoding, extraction, and classification.

[0054] ​The experimental controller uses LABVIEW (Laboratory Virtual Instrumentation Engineering Workbench) to write the software for device control and data processing, controlling the coordinated operation of the signal generator, impedance matching network, multi-band filter amplification circuit, oscilloscope, and gas flow controller throughout the experiment. It also performs data analysis on the effective data transmitted by the oscilloscope. The software performs interpolation fitting on the attenuated signal of the oscillator to obtain parameters such as resonance frequency f, phase φ, amplitude A, and attenuation coefficient τ.

[0055] The equation of the attenuation curve is: At the same time, the dissipation factor is calculated according to the following formula: Change in resonance frequency: ΔD = Δ5D - Δ1D.

[0056] Throughout the experiment, digital signal filtering and parameter calculation are performed online by the embedded experimental controller on the computer. Real-time plotting of the sensor response curve facilitates dynamic observation and tracking of the detection process. Nitrogen gas is first adjusted by a pressure regulator, then passes through a first one-way gas valve and a gas flow meter, and then is introduced into the gas phase detection chamber through the gas inlet pipe. At the bottom of the chamber, a precisely machined circular groove is embedded with a quartz crystal microbalance (QCM) sensor.

[0057] Below the QCM sensor, a planar spiral electromagnetic excitation coil and a receiving coil are placed. The excitation coil is driven by a radio frequency (RF) excitation signal from the signal generator, modulated by a two-parameter modulation circuit containing frequency and amplitude modulation, and transmitted through an impedance matching network. This modulation generates a high-dimensional virtual sensor array, improving the discrimination ability of VOCs and enriching the dimensionality of QCM-D signals.

[0058] Above the gas phase detection chamber, a gas sampling module is installed, which can be configured according to the characteristics of the experimental VOC sample. The chamber is further connected to an exhaust pipe equipped with a second one-way gas valve for discharging residual gas. Before discharge, the exhaust gas is treated by a tail gas treatment unit to ensure safe and environmentally friendly treatment of volatile emissions.

[0059] The above-mentioned wireless QCM-D system based on two-parameter modulation for gas detection method includes the combination of multi-frequency and amplitude modulation in time division multiplexing virtual sensor arrays, as follows: 1. Set the signal generator to output a sinusoidal excitation pulse signal in gated mode and enable signal output. The generator operates under two-parameter modulation, supporting real-time frequency and amplitude control.

[0060] 2. Adjust the impedance matching network to achieve impedance matching between the signal generator and the planar spiral inductor coil at the system's resonant frequency. This allows for stable excitation of the quartz crystal microbalance (QCM) over a wide range of frequencies.

[0061] 3. Connect a source of nitrogen gas to the gas inlet tube. Open the first and second one-way gas pass valves to allow nitrogen gas to flow into the system.

[0062] 4. Using the system controller, activate the gas flow controller to fill the gas phase detection chamber with high purity nitrogen gas, purge the tubing of residual gases, and establish a clean baseline condition.

[0063] 5. Maintain nitrogen gas flow for 10 to 20 minutes. After this stabilization period, the controller closes the gas flow valve and seals the chamber to prepare for sampling under inert conditions.

[0064] 6. Inject the target VOC sample gas through the gas sampler. Subsequently, the controller closes the second one-way valve, ensuring that the quartz crystal, now in resonance, interacts directly and exclusively with the gas. Changes in frequency and energy dissipation due to adsorption are tracked in real time.

[0065] 7. The experimental controller configures a virtual multi-channel system, setting the output signal amplitude (A) and frequency (f) on virtual channel 1. This output is generated by a LABVIEW-controlled signal generator that initiates amplitude modulation during the equilibrium phase of analyte binding, using a stepwise pattern from 1 V to 10 V and then a decreasing cycle back to 1 V. The time interval between each two adjacent amplitudes is 3 to 24 seconds, with a 1 V amplitude change to enrich the detection space.

[0066] 8. After completing the detection cycle, reopen the nitrogen gas valve to allow a 5 to 7 minute rinse phase to clear the cavity of residual analytes. Desorption continues for a total of 10 minutes until the resonator returns to its original baseline frequency. This prepares for the next sample injection.

[0067] 9. The receiving coil detects the attenuated oscillation signals emitted by the QCM under resonant excitation. These signals are processed through a multi-band filter and amplifier circuit, isolating the signals within the selected frequency band (f), then transmitting them to an oscilloscope for visualization. The PC-based controller acquires the waveform and calculates key resonance parameters in real time, including: fundamental frequency, harmonic distortion, decay factor.

[0068] 10. Every 3 seconds, the system cycles to the next virtual frequency-amplitude channel until the complete set of designated channels is covered. Although the process is sequential, the virtual channel system simulates parallel perception on a multi-dimensional sensor array, thereby increasing throughput and discrimination ability.

[0069] Figure 2The enhanced classification ability of the system was verified by showing the classification results of different fruit odor samples using the bi-parameter modulated QCM-D signatures derived from Figure 1 (a) and Figure 1 (b). The enhanced classification ability of the system was verified by showing the classification results of different fruit odor samples using the bi-parameter modulated QCM-D signatures derived from

[0070] Furthermore, it is to be understood that various alterations and modifications can be made to the application herein disclosed in the above description without departing from the scope of the application as defined by the appended claims.

Claims

1. A method for electromagnetically excited wireless QCM-D dual-parameter modulation and multimodal feature fusion for highly specific recognition of volatile organic compounds, characterized in that, For the WE-QCM-D wireless electrodeless quartz crystal microbalance system, a synchronous modulation method with frequency and voltage dual parameters of the driving voltage is designed. By extracting the change in acoustic energy dissipation on the quartz crystal surface caused by the change of driving voltage under different excitation frequencies, the steady-state response dimension of WE-QCM-D to VOCs is improved. Furthermore, machine learning and neural network algorithms are combined to fuse steady-state and dynamic response characteristics, thereby achieving highly selective identification of various similar VOC gases and mixed VOCs samples.

2. The method for electromagnetically excited wireless QCM-D dual-parameter modulation and multi-modal feature fusion according to claim 1, characterized in that, Amplitude modulation is initiated only after the wireless QCM-D sensor has reached equilibrium in its sensing response to the volatile organic compound analyte, thus ensuring stable feature extraction during amplitude cycling.

3. The method for electromagnetically excited wireless QCM-D dual-parameter modulation and multi-modal feature fusion according to claim 1 or 2, characterized in that, The time interval between two adjacent amplitude values ​​is 3 to 24 seconds.

4. The method for electromagnetically excited wireless QCM-D dual-parameter modulation and multi-modal feature fusion according to claim 1, characterized in that, The WE-QCM-D is driven to operate alternately at four resonant frequencies from the fundamental frequency to its seventh harmonic by a frequency sweep method. For each selected harmonic frequency, the driving voltage is modulated during the steady-state response phase of the sensor.

5. The method for electromagnetically excited wireless QCM-D dual-parameter modulation and multi-modal feature fusion according to claim 1 or 4, characterized in that, The amplitude of the excitation voltage is modulated in a stepped signal pattern; when the sensor's response to the volatile organic compound analyte reaches a stable state, the driving voltage is gradually increased from 1V to 10V in a stepped signal pattern, and then symmetrically reduced back to 1V. The frequency shift and energy dissipation changes corresponding to each frequency-amplitude pair are extracted and used as multidimensional features for analyte identification.

6. The method for electromagnetically excited wireless QCM-D dual-parameter modulation and multi-modal feature fusion according to claim 1, characterized in that, In the detection of volatile organic compounds, the sensor response signals, including the real-time changes in resonant frequency Δf and energy dissipation ΔD, are collected at each frequency-amplitude combination. These signals are processed by advanced time-domain and frequency-domain feature extraction algorithms, including attenuation envelope tracking, harmonic distortion analysis, and phase shift detection.

7. The method for electromagnetically excited wireless QCM-D dual-parameter modulation and multi-mode feature fusion according to claim 1, characterized in that, In the transient phase before equilibrium is reached, the frequency change Δf and the energy dissipation change ΔD are continuously monitored and calculated as dynamic functions of time.

8. The method for electromagnetically excited wireless QCM-D dual-parameter modulation and multi-modal feature fusion according to claim 1, characterized in that, The identification of volatile organic compounds was performed using advanced classification algorithms, including K-nearest neighbor (KNN), support vector machine (SVM), random forest (RF), and long short-term memory neural network (LSTM). The LSTM model was specifically configured to capture time dependencies and extract binding dynamics information from time-resolved sensor responses under different excitation voltages, thereby revealing how the driving amplitude affects the adsorption-desorption behavior of VOC molecules.

Images