Real-time detection method for crystallization of liquid sample based on intelligent sensor

Abstract

The application discloses a liquid sample crystallization real-time detection method based on an intelligent sensor and relates to the technical field of intelligent sensors.The method comprises the following steps: step 1, a liquid sample is combined with a toluene-acetone adjustable ratio ternary dewaxing medium composed of toluene, acetone and solvent oil according to a predetermined volume ratio, and a uniform mixed system is formed in a constant-temperature liquid preparation environment; step 2, the mixed system is placed into a sedimentation device; step 3, after an external physical field is terminated, the mixed system is allowed to stand and settle, wavelet analysis is performed to extract a wax precipitation signal, and a wax precipitation end point is determined according to a preset threshold; step 4, a wax-containing layer is separated at the wax precipitation end point, and the wax content of the liquid sample is output; and step 5, supernatant obtained after the standing and settling is completed is taken and reacted in a water bath to form a yellow complex.The application has the advantages of a short detection period, high separation efficiency, accurate signal recognition and objective and automatic calculation process.

Description

Technical Field

[0001] This invention relates to the field of intelligent sensor technology, and in particular to a method for real-time detection of crystallization precipitation in liquid samples based on intelligent sensors. Background Technology

[0002] With the rapid development of the Internet of Things (IoT) and intelligent manufacturing technologies, intelligent sensors, as core components of the sensing layer, are evolving towards integration, intelligence, and networking. Unlike traditional sensors, intelligent sensors integrate sensing elements, signal processing circuits, microprocessors, and communication interfaces, possessing functions such as signal acquisition, data processing, feature extraction, and intelligent judgment, enabling a leap from "sensing" to "cognition." The detection of crystallization precipitation of target components in liquid samples is of great significance in petrochemical, pharmaceutical, and food processing industries. Accurate and real-time monitoring of the crystallization process is crucial for controlling product quality, optimizing process parameters, and reducing production costs. Traditional methods for detecting crystallization precipitation mainly rely on manual observation, offline sampling analysis, or single-parameter monitoring, which suffer from low detection accuracy, slow response speed, and the inability to achieve continuous monitoring.

[0003] In recent years, optical detection technology has been widely used in the field of liquid analysis due to its advantages such as non-contact, fast response, and high precision. However, most existing optical detection equipment are independent analytical instruments, lacking intelligent data processing capabilities and unable to achieve multi-spectral information fusion and adaptive parameter adjustment. At the same time, traditional detection methods can usually only obtain information of a single wavelength or a limited spectral band. For crystallization processes in complex systems, it is difficult to accurately determine the precipitation endpoint and is easily affected by matrix interference and environmental factors. Existing technologies also have the following shortcomings: (1) The detection process requires manual intervention and has a low degree of automation; (2) The data processing algorithm is simple and lacks intelligent feature extraction and pattern recognition capabilities; (3) It cannot achieve real-time fusion analysis of multiple parameters, and the reliability of the detection results needs to be improved; (4) The equipment is large in size and expensive, making it difficult to achieve embedded integration and field deployment. Summary of the Invention

[0004] The purpose of this invention is to provide a real-time detection method for crystallization precipitation in liquid samples based on intelligent sensors. This method involves analyzing a target group induced by an external physical field, with the intelligent sensor collecting absorbance data at fixed time intervals. An embedded processor first performs Laplacian line subtraction on the absorbance data, then performs wavelet packet feature extraction, and intelligently determines the crystallization endpoint based on threshold rules. This method utilizes intelligent sensors to achieve sensitive detection of trace wax components in the supernatant and improves monitoring accuracy through dual-channel difference-integral technology. It solves the problems of poor wax control, long detection cycles, and strong colorimetric interference in traditional methods, offering advantages such as rapid response, ease of operation, and accurate results.

[0005] To address the aforementioned technical problems, this invention provides a real-time detection method for crystallization precipitation of liquid samples based on intelligent sensors, the method comprising:

[0006] Step 1: Combine the liquid sample with the sample pretreatment solvent system according to a predetermined volume ratio, and form a homogeneous mixture in a constant temperature preparation environment; the sample pretreatment solvent system is a solvent combination used to promote the crystallization and precipitation of the target component, and the target component is a solute with precipitation tendency;

[0007] Step 2: Place the mixed system into a constant temperature and pressure settling device and expose it to at least two external physical fields in sequence, namely ultrasonic field, mechanical shear field and magnetic field, to accelerate crystal nucleus formation and growth;

[0008] Step 3: After terminating the external physical field, allow the mixed system to settle. The intelligent sensor collects absorbance data at fixed time intervals. The embedded processor first performs Laplacian line subtraction on the absorbance data, then performs wavelet packet feature extraction, and determines the crystallization endpoint according to the threshold rule. The intelligent sensor includes a light source component, a detection cell, and an intelligent photoelectric sensor component.

[0009] Step 4: At the endpoint of crystallization, separate the lower layer containing sediment from the supernatant; optionally, perform mass and microcalorimetric measurements on the lower sediment, and calibrate the amount of crystallization using a dual-channel differential integral.

[0010] Step 5: Take the supernatant obtained after settling and react it with the colorimetric reagent solution to form a yellow complex. Place the reaction system in a cuvette and measure the absorbance at a wavelength of 420 nm or a preset working wavelength. Calculate the residual target component content in the supernatant based on the preset standard absorbance working curve. Combine the amount of crystallization obtained in Step 4 with the residual amount in the supernatant to calculate and output the target component monitoring results of the liquid sample.

[0011] Furthermore, step 1 specifically includes:

[0012] Step 1.1: Provide a constant temperature solution preparation tank with a double-layer cooling jacket and connected to a circulating refrigeration unit; add the components of the sample pretreatment solvent system to the constant temperature solution preparation tank in sequence. In this example, the components are toluene, acetone, and standard solvent oil, and the volume fractions of the three are set as 2-6 parts for toluene, 1-4 parts for acetone, and 3-7 parts for standard solvent oil, respectively; introduce a low-temperature coolant into the jacket of the constant temperature solution preparation tank to keep the temperature inside the tank constant at -20 degrees Celsius and maintain it for no less than 5 minutes to achieve temperature equilibrium;

[0013] Step 1.2: Take a liquid sample with a volume equal to the total volume of the sample pretreatment solvent system and slowly inject it into the same constant temperature preparation tank through the bottom feed tube under constant temperature conditions;

[0014] Step 1.3: Start the fixed-speed paddle stirrer and maintain stirring for 2 minutes to ensure that the sample pretreatment solvent system and the liquid sample are in full contact and form a uniform mixture.

[0015] Step 1.4: Immediately stop stirring and seal the container after stirring is complete. Let it stand for 30 seconds to eliminate turbulence. Confirm that the mixed system in the container has uniform color and turbidity and no stratification. The mixed system is now obtained.

[0016] Furthermore, in step 1.2, the time for injecting the liquid sample into the constant temperature mixing tank is controlled between 1 and 2 minutes to prevent temperature gradient fluctuations; in step 1.3, the rotation speed of the fixed speed paddle stirrer is set to 300 revolutions per minute.

[0017] Furthermore, in step 2, the mixed system is injected into a cryogenic, constant-pressure settling device via an insulated pipeline. The internal temperature is maintained at a constant -20 degrees Celsius through a refrigerant circulation system. Within the settling device, the system is exposed to at least two external physical fields: an ultrasonic field, a mechanical shear field, and a magnetic field, inducing the formation of wax crystal nuclei. When exposed to an ultrasonic field, a vertical ultrasonic transducer array is activated, exposing the mixed system to a vertical ultrasonic field with a frequency of 40 kHz and a power of 200 watts for 3 minutes. The irradiation direction of the ultrasonic field is parallel to the direction of gravity. When exposed to a mechanical shear field, a mechanical helical shear blade is activated, placing the mixed system in a mechanical shear field with a rotation speed of 300 revolutions per minute for 2 minutes. The rotation direction of the mechanical helical shear blade is aligned with the axis of the settling device. When exposed to a magnetic field, a direct current is passed through an external coil of the settling device, generating a constant magnetic field with a magnetic induction intensity of 0.1 Tesla and a magnetic field direction aligned with the axis of the settling device for 5 minutes.

[0018] Furthermore, step 3 specifically includes:

[0019] Step 3.1: After terminating the external physical field, allow the material to settle. The absorbance curve inside the settling cylinder is recorded by a smart sensor set at 20-second intervals. The data is continuously collected until a plateau appears on the curve.

[0020] Step 3.2: Cut the obtained absorbance curve into a fixed 5-minute time window, perform discrete Laplace operation on the absorbance sequence within each time window to remove the sedimentation background, and then perform three-level orthogonal wavelet packet decomposition on the Laplace residual sequence to extract the crystallization precipitation characteristic signal.

[0021] Step 3.3: Linearly fuse the wavelet coefficients of the wavelet packet signal stream according to the pre-stored coefficient table to obtain the fusion decision curve;

[0022] Step 3.4: When the fusion determination curve is lower than the predetermined threshold for the first three consecutive sampling points, it is considered as the end point of crystallization precipitation.

[0023] Further, step 3.2 specifically includes: cutting the obtained absorbance curve sequentially according to a fixed time window rule of 5 minutes; each time window contains 15 absorbance sampling points; establishing a one-dimensional time series for the 15 sampling points in each time window according to the sampling point order; using discrete Laplace operation based on numerical table integration, selecting 20 equidistant operation nodes in the range of 0 to 5 of the real part coordinates of the complex plane, and using each operation node as a variable for one Laplace operation; using each operation node as a reference, performing a stepwise phase operation on each of the 15 sampling points according to the preset Laplace kernel function weight matrix. Multiply and sum to obtain 20 Laplace transform coefficients; use one-dimensional linear interpolation to reconstruct these equidistant coefficients into a smooth curve, which serves as the Laplace line for the current time window; subtract the Laplace line value at the corresponding time point from the original value of the absorbance sampling point to obtain a residual vector of length 15, which serves as the Laplace residual sequence; perform three-level orthogonal wavelet packet decomposition on the residual sequence to obtain four wavelet packet sub-band signal streams; calculate the energy density value for each of the four sub-band signal streams; select the sub-band signal stream with the highest energy density value that is greater than a preset threshold as the crystallization precipitation characteristic signal.

[0024] Furthermore, when performing three-level orthogonal wavelet packet decomposition on the Laplace residual sequence, a Dabache orthogonal basis with four vanishing moments is selected as the wavelet packet parent basis: the first level decomposes the residual sequence into one low-frequency sub-band and one high-frequency sub-band; the second level decomposes the low-frequency sub-band obtained in the first level into one new low-frequency sub-band and one new high-frequency sub-band, while keeping the high-frequency sub-band of the first level undecomposed; the third level continues to decompose the low-frequency sub-band obtained in the second level into one low-frequency sub-band and one high-frequency sub-band, while keeping the remaining sub-bands obtained previously undecomposed, thus forming a total of four wavelet packet sub-band signal streams.

[0025] Furthermore, step 4 specifically includes:

[0026] Step 4.1: Keep the internal temperature of the sedimentation device constant at -20 degrees Celsius, open the sampling valve 5 mm from the bottom on the side wall of the sedimentation device, and use the liquid level difference to discharge the upper supernatant until the lower sediment is exposed; close the valve, use a polytetrafluoroethylene coated scraper to scrape off the sediment, and transfer it to a pre-cooled stainless steel sample dish to complete the separation.

[0027] Step 4.2: Place the sediment sample on an electronic analytical balance, wipe off the free liquid on the surface with non-fibrous absorbent paper, and weigh the initial mass. Then, place the sample and the sample dish in a 60°C far-infrared constant temperature oven for constant weight drying. Take it out and weigh it every 5 minutes and record the mass value. When the relative difference between two consecutive weighing results does not exceed 0.1%, it is considered that the constant weight state has been reached. Arrange all mass values ​​in chronological order to form a mass measurement result sequence.

[0028] Step 4.3: Take another sediment sample of equal mass and put it into a sealed aluminum dish of differential scanning calorimeter. Under the condition of isothermal starting point of -10 degrees Celsius, heat it to 80 degrees Celsius at a heating rate of 2 degrees Celsius per minute. Record the heat absorption value in the range of 30 degrees Celsius to 70 degrees Celsius in real time. Arrange all heat absorption values ​​in time order to form a sequence of microcalorimetric measurement results.

[0029] Step 4.4: The mass measurement result sequence and the microcalorimetry result sequence are synthesized using the dual-channel difference-integral method. The amount of crystallization is calibrated and then used as the amount of crystallization of the target component.

[0030] Further, step 4.4 specifically includes: performing adjacent difference operations on the mass measurement result sequence to generate a primary difference sequence; performing the same adjacent difference operation on the microcalorimetry result sequence to generate a corresponding primary difference sequence; converting the two primary difference sequences into a list of unsigned integers magnified by a thousand times according to their time indices, and then performing a binary XOR combination with a cyclic right shift of 5 bits in sequence to obtain a fused sequence; performing stepwise accumulation on the fused sequence from the first term to obtain a monotonically increasing sequence; scanning the monotonically increasing sequence to locate the smooth segment with the longest continuous value increase, and using the end point of this segment as the deterministic pivot; using the sum of all accumulated values ​​before the deterministic pivot as the local integral, and the total value of the sequence as the global integral, and calculating the ratio between the two; multiplying the obtained ratio by 100 to obtain the calibrated target component mass percentage, retaining 3 significant figures.

[0031] The present invention provides a real-time detection method for liquid sample crystallization based on intelligent sensors, which offers the following advantages: This method is convenient to operate, highly sensitive, and provides accurate monitoring, especially in colorimetric analysis. By introducing a ternary dewaxing medium composed of toluene, acetone, and solvent oil, the precipitation behavior of waxes in crude oil samples is effectively controlled. Combined with external physical fields such as ultrasound, shearing, or magnetic fields to induce the formation of target components, the controllability and repeatability of the wax precipitation process are improved. After the wax precipitation endpoint, the supernatant is obtained through static separation. A specially designed colorimetric reagent is then introduced for colorimetric reaction, forming a stable yellow complex. The absorbance is further measured at a specific wavelength, thus establishing a monitoring and analysis method for residual wax in the supernatant. This colorimetric step greatly enhances the resolution capabilities that traditional gravimetric methods cannot provide, enabling sensitive identification of trace residual waxes. Furthermore, by integrating the difference between the data from the microcalorimetric and mass determination channels, multi-dimensional qualitative monitoring and fusion analysis of wax content is achieved. This method significantly reduces background absorption interference, improves the stability and linear response range of colorimetric results, and solves the problems of poor colorimetric selectivity, insufficient sensitivity, and strong background interference in existing technologies. Furthermore, the method requires simple equipment and is easily applicable to field and laboratory testing in oil fields, providing a rapid, accurate, and highly operable technical means for crude oil fluidity evaluation and wax deposition risk monitoring. Attached Figure Description

[0032] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0033] Figure 1 A schematic diagram of the method flow for real-time detection of liquid sample crystallization based on intelligent sensors provided in an embodiment of the present invention;

[0034] Figure 2 This is a schematic diagram illustrating the principle of wax nucleus formation in a toluene-acetone adjustable ratio ternary dewaxing medium provided in an embodiment of the present invention.

[0035] Figure 3 This is a schematic diagram illustrating the effect of toluene-acetone adjustable ratio ternary dewaxing medium on the uniformity of mixing with crude oil, provided in an embodiment of the present invention.

[0036] Figure 4 This is a schematic diagram of the optical density time history variation curves under different toluene-acetone volume ratios provided in an embodiment of the present invention. Detailed Implementation

[0037] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0038] refer to Figure 1 A real-time detection method for crystallization precipitation of liquid samples based on intelligent sensors, the method comprising:

[0039] Step 1: Combine the liquid sample with the toluene-acetone adjustable ratio ternary dewaxing medium composed of toluene, acetone and solvent oil according to a predetermined volume ratio, and form a homogeneous mixture in a constant temperature preparation environment;

[0040] The toluene-acetone tunable-ratio ternary dewaxing medium constructed in a constant-temperature solution preparation environment provides a controllable platform for adjusting the dissolution free energy barrier of the entire system. The toluene molecular skeleton exhibits a uniform π-electron cloud distribution and a combination of hydrophobicity and aromatization, allowing it to form a transient conjugated system with aromatic components in crude oil through π-π stacking, thus reducing the local solvation energy. Acetone molecules, due to the strong dielectric properties of their carbonyl groups, exhibit amphiphilicity at the polar-nonpolar interface, accelerating the depolymerization of polar micelles in crude oil and reducing viscous association. The standard solvent oil, with its narrow viscoelastic spectrum at ambient pressure and low temperature, can serve as a dilution layer for molecular diffusion channels, suppressing viscosity steps caused by competitive dissolution of aromatics and ketones. The resulting toluene-acetone tunable-ratio ternary dewaxing medium-crude oil mixture maintains low viscosity and a high diffusion coefficient macroscopically, while forming tunable-ratio hydrophobic-electrophilic structural domains microscopically, providing a unified thermodynamic potential field for subsequent supersaturation and abrupt nucleation of wax molecules.

[0041] Step 2: Place the mixed system into a settling device and expose it to at least two external physical fields in sequence, namely an ultrasonic field, a mechanical shear field, and a magnetic field, to induce the formation of wax crystal nuclei;

[0042] After the mixed system enters the cryogenic constant-pressure settling device, external ultrasonic fields, mechanical shear fields, and magnetic fields are applied in a predetermined sequence or combination, essentially causing graded disturbances to the kinetic path of crystal nucleus formation. Cavitation bubbles in the ultrasonic field generate extremely high instantaneous pressure and temperature micro-regions during rapid contraction and bursting. These micro-regions induce local destabilization of wax molecules, forming primary crystal nuclei with extremely small critical radii. Simultaneously, cavitation jetting and acoustic flow effects disperse the primary crystal nuclei at high speed, preventing aggregation due to local concentration gradients. The competition between the shear strain rate introduced by the mechanical shear field and the viscous resistance of the liquid causes the flow field in the micro-region containing the wax crystal nuclei to exhibit an unsteady vortex structure. The pressure difference between the shear inlet and outlet induces a rearrangement of molecular orientation on the crystal nucleus surface, increasing the free energy release efficiency of the crystal nucleus surface, shortening the induction period, and increasing the number of active sites on the crystal nucleus surface. The magnetic field utilizes the slight magnetic anisotropy of a small amount of magnetic impurities in the crude oil and the waxy hydrocarbon chains to guide the crystal nuclei to form a longitudinal alignment in the macroscopic gravitational direction, reducing the probability of lateral random collisions, thus exhibiting a higher axial settling velocity in the subsequent gravity settling stage. The three physical fields partially overlap in the time domain and are complementary and coupled in space, jointly establishing a low-barrier, high-gradient, multi-scale nucleation channel, enabling waxy crystal nuclei to be generated uniformly in the system in a quasi-synchronous manner, further ensuring the uniformity and decoupling of the optical scattering monitoring signal.

[0043] Step 3: After terminating the external physical field, allow the mixed system to settle. Collect optical density data at fixed time intervals. First, perform Laplace transform on the optical density data to remove the settling background, then perform wavelet analysis to extract the wax precipitation signal. Determine the wax precipitation endpoint based on the preset threshold.

[0044] The essence of the static sedimentation stage is the coupling effect of multi-component convection-diffusion driven by the continuous evolution of particle size distribution and concentration stratification. The optical density curve obtained by the laser scattering density recorder can be regarded as the comprehensive scattering intensity of the incident laser at different spatial-temporal cross-sections. The sedimentation background signal exhibits a slow linear or sublinear change, mainly reflecting the influence of the overall particle volume fraction shift in the liquid-solid suspension system on the scattering cross-section; the wax precipitation signal, however, shows a rapid oscillation of high-frequency noise troughs and peaks, originating from the local refractive index abrupt changes caused by the continuous rearrangement, aggregation, and fall of wax crystals. The two are superimposed on each other in the original curve and are difficult to distinguish intuitively. The discrete Laplace transform amplifies the phase difference of the low-frequency trend term by mapping the time-domain sequence to a small-window integral form in the complex frequency domain, thus separating the sedimentation background from the instantaneous scattering change on the phase-amplitude plane, so that the sedimentation contribution can be eliminated by simple baseline fitting. The Laplace residual sequence is then decomposed into four sub-band signal streams by three layers of orthogonal wavelet packet decomposition, covering four frequency bands from quasi-static concentration fluctuations to transient crystallization events. Orthogonal bases ensure that subbands are uncorrelated, avoiding information duplication. Weighted fusion of wavelet energy density is equivalent to multi-resolution weighted projection of the contributions of events at different time scales in the frequency domain. When the fusion judgment curve shows a continuous decline and crosses the threshold, it indicates that the high-frequency wax precipitation event tends to terminate and the low-frequency sedimentation trend dominates, thus objectively determining the wax precipitation endpoint.

[0045] Absorbance data is collected at fixed time intervals using a smart sensor. The light source component of the smart sensor employs a multi-wavelength LED array structure, including ultraviolet LEDs (wavelength 280-400nm), visible LEDs (wavelength 400-700nm), and near-infrared LEDs (wavelength 700-1000nm). Each LED unit is equipped with an independent current control circuit, enabling precise adjustment of light intensity and programmed switching of wavelengths via an embedded microcontroller. The light source component also includes an optical fiber coupler to converge multiple light sources and guide them into the detection cell via multimode fiber, ensuring optical path stability and light intensity uniformity.

[0046] The detection cell is made of quartz glass, which possesses excellent optical transmittance and chemical stability. It features a flow-through structure, including a sample inlet, a sample outlet, and an optical path control chamber. The optical path length is set at 10 mm, ensuring sufficient detection sensitivity while avoiding excessive light intensity attenuation. The inner wall of the detection cell undergoes special polishing, with a surface roughness of less than 0.1 μm, reducing light scattering interference. The cell is designed with a temperature control jacket, maintaining a constant detection temperature through circulating constant-temperature liquid, with a temperature control accuracy of ±0.1℃. A magnetic stirrer is installed at the bottom of the detection cell to ensure that the sample remains uniformly mixed during detection.

[0047] The intelligent photoelectric sensor assembly is the core component of this invention. It adopts a highly integrated modular design and mainly includes: a multi-element photodetector array: employing a silicon photodiode array, comprising 64 independent photodetector units, each corresponding to a specific wavelength range, achieving synchronous detection across the entire spectrum. A narrowband filter array is configured at the front end of the detector, with a center wavelength spacing of 5nm and a full width at half maximum (FWHM) of 3nm, ensuring spectral resolution. A signal conditioning circuit: each photoelectric signal is converted into a voltage signal by a transimpedance amplifier, and then amplified by a programmable gain amplifier (PGA). The amplification factor can be adjusted programmatically within the range of 1-1000 times. The signal conditioning circuit integrates a noise filtering module, employing an adaptive digital filtering algorithm to effectively suppress environmental noise interference. An embedded data processing unit: using a 32-bit ARM Cortex-M4 microprocessor with a main frequency of 120MHz, integrating a floating-point unit (FPU) and a digital signal processor (DSP). The processor has built-in 512KB Flash memory and 128KB BSRAM to store signal processing algorithms and standard databases.

[0048] Step 4: Separate the wax-containing layer at the wax precipitation endpoint, and perform mass determination and microcalorimetry determination respectively. Use a dual-channel difference subtraction integral method to synthesize the mass determination result sequence and the microcalorimetry result sequence, and output the wax content of the liquid sample.

[0049] After the wax layer is separated, mass determination and microcalorimetry characterize the "quantity" and "energy" of the wax sample in different dimensions. During mass determination, the sample mass decreases non-linearly over time. The early volatilization-driven phase corresponds to the evaporation of residual light solvents, the mid-term contraction phase reflects the migration of a small amount of high-boiling hydrocarbons from the target component, and the final constant-weight phase indicates the stability of the solid wax skeleton. The temperature-scanning phase of microcalorimetry records the multi-stage melting-rearrangement-endothermic process of the wax lattice in the range of -10°C to 80°C. The integral of its endothermic peak is equivalent to the enthalpy change of the molecules from order to disorder after overcoming the lattice potential. Amplifying, cyclically shifting to the right, and performing binary XOR on the two primary difference sequences essentially projects the two physical quantities of mass and energy into the same unsigned integer space, eliminating dimensional differences through bit operations and achieving data structure alignment. In the monotonically increasing sequence after stepwise accumulation, the segment with the smallest smooth increase represents the point where both mass and energy signals enter the dynamic saturation point. The ratio of the local integral to the global integral serves as a dimensionless index, encompassing both the absolute enrichment of wax within the framework of mass conservation and the relative energy contribution of melting enthalpy and structural order. This allows the measurement results to account for system drift caused by differences in residual solvent content, microcrystal orientation, and side chain isomerism.

[0050] Step 5: Take the supernatant obtained after settling and add the colorimetric reagent solution at a volume ratio of 1:1. React in a water bath to form a yellow complex. Transfer the reaction system to a cuvette and measure the absorbance at a wavelength of 420 nm. Calculate the residual wax content in the supernatant based on the pre-established standard absorbance curve. Subtract the residual wax content in the supernatant from the wax content obtained in Step 4 to obtain the colorimetric result.

[0051] Furthermore, in the entire rapid sedimentation-wax separation coupled detection process, the initial low-temperature solution preparation stage is not only the starting point for setting the physicochemical conditions, but also a crucial prerequisite for the effective implementation of subsequent induced nucleation and online optical monitoring. Firstly, the constant-temperature solution preparation tank adopts a double-layer cooling jacket structure and continuously circulates a low-temperature refrigerant within the jacket using a circulating refrigeration unit, thereby stabilizing the tank temperature at -20 degrees Celsius through forced heat exchange. The significance of the low-temperature environment for the ternary dewaxing medium lies not only in suppressing the saturated vapor pressure of toluene and acetone to prevent volatilization and resulting ratio drift, but also in simultaneously reducing the overall Gibbs free energy of the system, weakening the driving force for mutual diffusion between toluene-acetone and the standard solvent oil, thus reducing the macroscopic concentration gradient. When toluene (2-6%), acetone (1-4%), and standard solvent oil (3-7%) are injected into the tank in a sequential volume fraction range, toluene preferentially encapsulates short-chain and medium-chain n-alkane segments due to its aromatic π-electron shielding effect. The carbonyl dipole of acetone penetrates into the originally tightly packed high-cycloalkane and microcolloid clusters, causing them to undergo polar rectification in a short time. The standard solvent oil, existing as a mixture of low-volatility, high-flash-point alkanes, acts as a molecular diffusion buffer layer in the viscosity-temperature sensitive range, limiting the heat dissipation of instantaneous miscibility between toluene and acetone. Maintaining a constant low temperature for more than 5 minutes indicates that the ternary dewaxing medium has completed molecular rearrangement at the microscopic level and is approaching thermodynamic equilibrium. At this point, there is no significant density convection or concentration stratification, creating a level playing field for subsequent mechanical perturbations.

[0052] Next, a liquid sample with a total volume equal to that of the aforementioned ternary dewaxing medium is slowly injected into the same tank through the bottom feed pipe. The physical principle behind bottom feeding at -20 degrees Celsius lies in utilizing the positive density difference to construct a laminar flow-like upward channel, allowing the crude oil sample to rise controllably along the central axis, while the original ternary dewaxing medium remains relatively stationary in the peripheral region, thus avoiding violent convection or even fractal fingering phenomena. An injection window of one to two minutes ensures that the axial Reynolds number is maintained in the low-to-medium transition range, preventing excessive turbulence at the crude oil-medium mixing interface that could lead to localized instantaneous high temperatures, and also suppressing shear heating caused by the high viscosity of the crude oil itself. As the crude oil slowly fills, the solvated envelope layer at the local micro-interface begins to reorganize: toluene preferentially penetrates into the central unit volume of the crude oil, acetone forms multipolar bridges around organic acids, colloids, and trace metal chelates, and the long-chain skeleton of the standard solvent oil fills the free volume in the three-dimensional network, thereby establishing a supersaturated mixture characterized by a hydrophobic core-electrophilic shell dual-domain structure.

[0053] A fixed-speed paddle agitator inside the tank was then activated and maintained for 2 minutes. The paddles cut the fluid at a constant speed under -20°C conditions, inducing a complex flow field where macroscopic circulation and microscale turbulent motion coexist. The high-speed section propelled the mixture down the tank wall and then back up, while the low-speed section formed a vortex street at the paddle tail, enhancing axial-radial mixing. During this process, long-chain alkanes, n-fatty acids, and a small amount of asphaltenes in the crude oil were rapidly distributed into the toluene-acetone solvation network, and the system viscosity decreased significantly with the inter-dispersion of components. More importantly, the stirring shear provided a uniform precursor for subsequent wax supersaturation: within any microcell in the flow field, the dissolution-re-collision timescale experienced by wax molecules tended to be consistent, preventing premature formation of hidden micronuclei that would be sheared and destroyed, and also preventing some areas from failing to reach supersaturation due to a lack of wax. Because the entire tank was maintained at -20°C, the turbulent heat and viscous heat generated by the stirring were rapidly carried away by the jacket, ensuring that macroscopic temperature fluctuations remained within ±0.2°C.

[0054] The machine is immediately stopped and the tank lid is sealed after 2 minutes. Within 30 seconds, the fluid inertia rapidly decays to near-stationary state through wall friction and viscosity dissipation. Turbulent kinetic energy is converted into internal energy at the molecular level and absorbed by the low-temperature environment. The main function of this settling window is to smooth out the flow field memory effect: if the vortex structure persists, it will cause uneven relative migration speeds of wax molecules, triggering fluctuations in spatial nucleation conditions again; while prolonged settling will cause slight temperature stratification between the tank bottom and top, thus affecting subsequent sedimentation dynamics. After 30 seconds of turbulent decay observation, if the color and turbidity of the mixed system are in a monochromatic continuous distribution without interface stratification or local turbidity spots, it can be considered that a truly homogeneous toluene-acetone adjustable ratio ternary dewaxing medium-crude oil mixed system has been formed. At this moment, although the concentration of wax molecules inside the system is supersaturated, the nucleation barrier is still suppressed in a metastable state by the low temperature and sufficient solvation, meaning that the system has stored potential wax precipitation energy that can be released with external physical stimulation.

[0055] Furthermore, the injection time of the liquid sample into the isothermal mixing tank is controlled between 1 and 2 minutes. This narrow time window utilizes slow axial upward flow to suppress instantaneous convection, ensuring that the uniform temperature field of -20 degrees Celsius in the tank is not broken. If the injection is too fast, the initial temperature carried by the crude oil will form a "hot tongue" along the central axis, which may induce local wax precipitation before radial diffusion, disrupting the unified starting point for subsequent controllable nucleation. If the injection is too slow, the cooling of the tank wall will cause a sudden increase in the viscosity of the crude oil-medium interface, leading to laminar slip and fingering, generating micro-concentration hotspots, which will also weaken the ability of the subsequent external physical field to synchronously trigger the crystal nuclei. Therefore, the injection rhythm of 1 to 2 minutes keeps the system operating on an acceptable equilibrium line between density difference and shear rate, with the temperature gradient always below 0.2 degrees Celsius, and the concentration distribution maintained in a uniform supersaturated state.

[0056] A fixed-speed paddle agitator operates at 300 rpm for 2 minutes. This speed is higher than the laminar-turbulent flow criticality but lower than the cavitation sound velocity threshold. 300 rpm provides sufficient shear rate in a high-viscosity environment at -20°C, enabling rapid inter-diffusion of toluene-acetone and long-chain alkanes in crude oil, smoothing out micro-concentration fluctuations, and preventing solvent evaporation or microbubble generation caused by localized negative pressure at the blade tips. During the shearing process, the energy generated by the paddles is instantaneously carried away by the forced heat exchange of the tank cooling jacket, resulting in negligible overall system heat accumulation and maintaining a continuous thermodynamic path from dissolution to supersaturation to metastable state. Ultimately, the 1-2 minute injection window and the 300 rpm agitation intensity are coupled in the time-kinetic dimension to construct a uniform, low-viscosity ternary dewaxing medium-crude oil mixture, laying a repeatable and predictable nucleation benchmark for subsequent synergistic induction by ultrasonic, mechanical shear, or magnetic fields.

[0057] Furthermore, after the isothermal liquid preparation is completed, the mixed system is sent to a cryogenic constant-pressure settling device via insulated pipelines. The pipeline walls are reinforced with a vacuum insulation layer and a polytetrafluoroethylene lining to suppress external heat flux, ensuring that the system temperature remains at -20 degrees Celsius during flow and preventing premature release of supersaturated energy. The settling device itself adopts a jacketed refrigerant circulation structure, using a proportional valve and a variable frequency pump to match the inlet and outlet flow rates in real time, keeping the chamber pressure stable and slightly higher than atmospheric pressure. This prevents cryogenic boiling and eliminates nucleation noise caused by oscillations. After entering the device, the mixed system is initially in a static stable region, and then exposed to at least two external physical fields—an ultrasonic field, a mechanical shear field, and a magnetic field—in a preset sequence. Under the excitation of a vertical ultrasonic transducer array, a 40 kHz, 200 W sound wave perfectly coincides with the direction of gravity along the sedimentation axis. Acoustic cavitation microbubbles burst open in a very short local time. The instantaneous high pressure and high temperature pulse destroy the weak van der Waals bond between the wax molecules and the solvent shell, lowering the nucleation barrier, and uniformly dispersing the nascent crystal nuclei throughout the volume via microjets. Acoustic irradiation for 3 minutes ensures that the cavitation number density reaches a steady state without causing large-scale acoustic heating of the system. Subsequently, a mechanical helical shear blade rotates in the same direction at 300 rpm for 2 minutes. The axial thrust of the propeller and the high radial shear work together to rapidly flatten the microscale pressure distribution left by the ultrasonic field, preventing the crystal nuclei from agglomerating and growing, ensuring a narrow crystal nucleus size distribution, and reducing the difference in the sinking velocity relative to the direction of gravity, thus constructing a clear scattering cross-section gradient for subsequent optical density monitoring. If a magnetic field is chosen as the synergistic field, a constant 0.1 Tesla magnetic field is applied to the external coil of the device, with the magnetic flux direction aligned with the sedimentation axis and lasting for 5 minutes. This magnetic induction intensity generates an orientation moment for the target component nuclei containing trace amounts of paramagnetic components or metallic complex groups, prompting them to preferentially align along the axial direction, further reducing the surface free energy, thereby stabilizing the initial crystal lattice and inhibiting re-dissolution. Two or three of the three physical fields can be selected and superimposed according to the operating conditions. Their respective durations are controlled through experimental optimization to ensure that they do not interfere with each other while producing significant synergy: ultrasound preferentially triggers nucleation, mechanical shearing locks the particle size, and the magnetic field reduces the surface energy of the crystal nuclei through anisotropic orientation, thus achieving a balance between the three parameters of nucleation rate, crystal nucleus uniformity, and subsequent sedimentation readability.

[0058] Furthermore, after the external physical field terminates, the mixed system immediately enters a static settling state. The key to this process lies in separating the macroscopic settling background from the microscopic wax precipitation signal. A laser scattering density recorder scans the internal cross-section of the settling cylinder at 20-second intervals, continuously obtaining optical density curves. This curve is essentially the coupling result of particle volume fraction, liquid refractive index, and optical path, encompassing both the slow, monotonic changes caused by the overall particle settling and the transient fluctuations caused by wax nuclei aggregation, desorption, and redistribution. To prevent thermal drift and light source attenuation from masking subtle signals during long-term acquisition, the system employs a high-stability temperature-controlled optical platform and digital feedback laser drive, ensuring that the light source power and detector sensitivity are always locked within preset ranges. When the curve gradually becomes horizontal, it indicates that the concentration of scatterable particles has entered a relatively stable range; at this point, acquisition terminates, and the data enters the subsequent processing chain.

[0059] The obtained optical density curve is segmented into fixed five-minute time windows, essentially constructing a sliding observation frame in the time domain to maintain statistical homogeneity within each segment. Discrete Laplace transform is used to remove sedimentation background. The idea is to use a complex plane kernel function to map the original time series to a new domain, systematically suppressing first- and second-order time correlations, folding the monotonic sedimentation term back to its zero nearest neighbor, and leaving a residual sequence dominated by high-order fluctuations. This residual still contains noise and effective wax precipitation signals, thus requiring further decomposition using multi-resolution tools. Three-layer orthogonal wavelet packet decomposition, while maintaining energy integrity, divides the residual sequence into a low-frequency smooth component and multiple high-frequency sub-band signal streams. The growth and aggregation of wax crystal nuclei instantaneously change the scattering cross-section, exhibiting a burst-like characteristic of highly concentrated energy density but limited time span, falling precisely within a specific wavelet sub-band, while random noise shows a sparse energy distribution. By calculating the energy density of all sub-bands and comparing with predetermined thresholds, the most representative wax precipitation signal stream can be quickly located.

[0060] The wavelet coefficients of the selected sub-bands were then linearly fused according to a pre-stored weight table to obtain a fusion decision curve. The weight table was derived from training data of a large number of different crude oil samples, comprehensively considering the influence of viscosity, the proportion of resinous material, and potential metal complexes on the scattering response. The significance of the fusion operation is to converge multi-scale information into a single dimension, preserving the sensitivity to abrupt changes in wax precipitation while eliminating false triggers caused by individual fluctuations. When the fusion decision curve shows three consecutive sampling points simultaneously below the threshold for the first time, it indicates that the number of wax nuclei and the scattering ability in the system no longer change significantly over time, the sedimentation has entered a stable region, the wax precipitation kinetics process has terminated, and this is monitored and regarded as the endpoint of wax precipitation.

[0061] Furthermore, the system segments the entire optical density curve according to a fixed time window rule of "5 minutes per segment," with each time window strictly containing 15 optical density sampling points. This ensures both a constant sampling interval within each segment and separability between the window span and the macroscopic timescale of particle descent during sedimentation. Subsequently, a one-dimensional time series vector is established for the 15 optical density sampling points within each time window in chronological order of appearance, and a discrete Laplace operation based on numerical table integration is immediately invoked. This process selects 20 equidistant operation nodes between the real part coordinates 0 and 5 in the complex plane, treating each node as a discrete Laplace variable. The system calls a pre-stored Laplace kernel function weight matrix, multiplying each of the 15 sampling points sequentially with the corresponding weights and accumulating the results to obtain 20 Laplace transform coefficients. Next, a one-dimensional linear interpolation method is used to reconstruct these 20 coefficients into a smooth curve, which is defined as the sedimentation Laplace line of the current time window, representing the slow contribution of macroscopic gravitational sedimentation to optical density. Next, the program subtracts the corresponding Laplacian line values ​​from the values ​​of the original 15 optical density sampling points, forming a residual vector of length 15, i.e., the Laplacian residual sequence. At this point, the macroscopic sedimentation background is effectively removed, leaving only the net signal containing noise and potential wax precipitation disturbances.

[0062] To further identify wax precipitation disturbances, the system performs three-level orthogonal wavelet packet decomposition on the Laplace residual sequence. Using a Dabéchian orthogonal basis with four vanishing moments, and under energy completeness constraints, the residual sequence is expanded layer by layer, ultimately yielding four wavelet packet sub-band signal streams. Each sub-band signal stream corresponds to a different frequency band: the low-frequency sub-band retains the trend term, while the mid-to-high-frequency sub-band highlights local burst changes. Wax nuclei instantaneously change their local scattering cross-section during formation, aggregation, or deagglomeration; these abrupt changes often concentrate in specific mid-to-high frequency ranges. The system calculates the energy density values ​​for each of the four sub-band signal streams and compares them with preset thresholds, selecting the sub-band signal stream with the highest energy density exceeding the threshold as the wax precipitation signal for the current time window. In this way, low-frequency fluctuations caused by macroscopic sedimentation are shielded, and random noise is weakened due to energy dispersion, leaving only a net signal highly sensitive to the behavior of wax nuclei.

[0063] Furthermore, when the Laplace residual sequence enters the three-level orthogonal wavelet packet decomposition, the system first invokes the Dabéchian orthogonal basis with four vanishing moments. The fourth-order vanishing moment of the Dabéchian orthogonal basis means that its scaling function can completely eliminate the polynomial tendency up to the third order, allowing the energy of slowly changing terms to be naturally folded into the low-frequency channel without contaminating the high-frequency components. At the same time, its symmetric support and compact support properties ensure that the filtering convolution is performed only within a finite data segment and does not introduce boundary spread. Following the orthogonal wavelet packet idea, the algorithm sequentially applies orthogonal low-pass and high-pass filter banks to the signal, and then performs binary decimation. After the first layer of operation, the Laplace residual sequence is decomposed into a low-frequency subband and a high-frequency subband. The low-frequency subband carries the main energy in the optical scattering residual, while the high-frequency subband records local abrupt changes and higher-order oscillations. To further highlight the transient information of the wax crystal nuclei in subsequent processing, the system performs isomorphic filtering and extraction on the low-frequency subband of the first layer only in the second layer, subdividing it into one new low-frequency subband and one high-frequency subband, while retaining the high-frequency subband of the first layer without further decomposition. This avoids excessive splitting of high-frequency noise, which would dilute the energy, and also allows the true wax analysis signal to remain intact in the subbands that are still in the mid-to-high frequency range. In the third layer, the algorithm performs the same filtering and extraction on the low-frequency subband obtained in the second layer again, generating one low-frequency subband and one high-frequency subband for the third layer. At this point, all previously generated high-frequency subbands are fixed and no longer drilled down, ultimately forming a total of four wavelet packet subband signal streams at the nodes of the time-frequency plane: the third-layer low-frequency subband, the third-layer high-frequency subband, the second-layer high-frequency subband, and the first-layer high-frequency subband. These four subbands maintain energy conservation under orthogonality constraints, are mutually exclusive, and their sum equals the original energy of the Laplace residual sequence, allowing subsequent energy density comparisons to be performed on a unified benchmark. By employing a strategy of recursively decomposing only low-frequency branches, the system maximizes the compaction of wax analysis correlation signals and reduces the diffusion of high-frequency noise in the tree structure. This not only improves the resolution of the energy density criterion but also reduces the computational load and storage overhead during runtime, thus achieving a balance between limited sample length and real-time computing requirements.

[0064] Furthermore, once the fusion determination curve inside the settling cylinder confirms the wax precipitation endpoint, the settling device is kept at a constant temperature of -20 degrees Celsius to prevent any temperature rise from causing the wax to redissolve or recrystallize. The operator first opens the sampling valve located only 5 mm from the bottom on the side wall, allowing the upper clear liquid to flow out by gravity using the liquid level difference, until the fluid at the valve outlet changes from clear and transparent to a milky white semi-solid state, indicating that the wax-containing layer at the bottom has been exposed. The valve is then closed to prevent backflow disturbance from the residual clear liquid. At this point, a PTFE-coated scraper is slowly inserted into the bottom of the settling cylinder from the side wall, using a unidirectional scraping motion along the axial direction to the outer edge to peel off the entire wax-containing layer and push it to the leading edge of the scraper. It is then quickly transferred to a low-temperature stainless steel sample dish pre-placed in a dry ice bath. The entire operation takes no more than 30 seconds, allowing the wax to complete physical separation from the liquid phase and be solidified and locked in at a -20 degree Celsius isothermal path, preventing the volatilization and re-adsorption of surface-adsorbed light components.

[0065] After separation, the sample dish was placed on an electronic analytical balance. The operator used fiber-free absorbent paper to gently touch the surface of the wax block to remove residual free liquid, and immediately read and recorded the initial mass. The sample, along with the sample dish, was then placed in a 60°C far-infrared oven for constant weight drying. Far-infrared radiation penetrates the microporous structure of the wax, avoiding the hard-shell effect of traditional hot air ovens that prevents internal moisture or light oils from escaping. The sample was weighed and its mass recorded every 5 minutes, with the exposure time to room temperature controlled within 15 seconds to ensure weighing stability. When the relative difference between two consecutive weighings does not exceed 0.1%, constant weight was considered achieved. At this point, all weighing data were arranged chronologically to form a mass determination result sequence. This sequence typically exhibits a two-stage characteristic: the initial rapid descent corresponds to the volatilization of free liquid and surface low-molecular-weight hydrocarbons, while the subsequent slow stabilization reflects the trace mass loss associated with the redistribution of target components and micropore shrinkage. The sequence morphology provides sufficient texture for subsequent differential processing.

[0066] Simultaneously, an equal mass sample was taken from the same wax-containing layer and placed in a sealed aluminum dish of the differential scanning calorimeter. The instrument stabilized at an isothermal starting point of -10°C for 3 minutes to ensure internal and external temperature equilibrium, and then heated to 80°C at a linear heating rate of 2°C per minute, recording the endothermic values ​​in the range of 30 to 70°C in real time. This temperature range covers the endothermic peaks of the wax lattice transitioning from an ordered orthogonal or monoclinic structure to a disordered liquid phase. The peak area of ​​the endothermic curve is proportional to the content of the target component, while the peak position is related to the average carbon number and branching degree of the wax molecules. During the recording process, the volatile components in the sealed aluminum dish were effectively tamed by the airtightness of the equipment, and the heat flux baseline was updated automatically with zero-point correction, thus generating a drift-free microcalorimetric measurement result sequence. The sequence usually exhibits a single peak or multiple peak superposition, and the appearance, width, and symmetry of each peak are closely related to the distribution of the wax population and their interactions.

[0067] Next, the dual-channel difference-subtraction integral calculation begins. The software first performs adjacent differences on the mass measurement result sequence, generating a data sequence with a length one less than the mass measurement result sequence. The same operation is performed on the microcalorimetry result sequence to obtain the corresponding first-order difference sequence. To enhance the numerical resolution of integer logic operations, both difference sequences are multiplied by 1000 and their absolute values ​​are taken, converting them into a list of unsigned integers, thus eliminating the bias caused by dimensional differences. The system then right-shifts one sequence by 5 bits and performs a cyclic XOR operation, superimposing it with the corresponding bits of the other sequence to output a fused sequence. This XOR operation generates interference patterns at the binary level for the local slope information of the mass and energy dimensions, strengthening the energy of the resonance segment and canceling out the energy of the non-common segment, thereby improving the signal-to-noise ratio of the waxy characteristic segment. The fused sequence is then accumulated stepwise from the first element to obtain a strictly monotonically increasing sequence. The algorithm scans this sequence, calculates the continuous stationary length of each adjacent difference segment, and finds the endpoint of the segment with the smallest increase and the longest continuous time, defining it as the deterministic pivot. The sum of all accumulated values ​​before the pivot is used as the local integral, and the sum of the entire sequence is used as the global integral. The ratio of the two is multiplied by 100 to obtain the mass percentage of wax content, and 3 significant figures are retained in the output.

[0068] The underlying logic of this dual-channel difference-subtraction integration strategy lies in mapping the conservation of matter and energy onto the same digital framework. The mass sequence reveals the duration of the wax block's own volatilization and shrinkage, while the thermal sequence reflects the latent heat release from lattice melting. The difference between the two captures the synchronous rate of change of the same physical phenomenon in different measurement domains. By right-shifting the sequences and combining them with XOR, the system artificially creates a phase difference, allowing the real wax signals with the same trend to be superimposed and enhanced, while inconsistencies arising from accidental noise or baseline drift are mostly canceled out. The final accumulation and pivot positioning are equivalent to finding the window of maximum contribution of the wax characteristic signal in the global context. The ratio of the local integral to the global integral integrates the mass and thermal characteristics of the wax in a dimensionless form, and the corresponding mass fraction can be obtained after percentage factorization. This result is traceable to weighing accuracy, heat flux sensitivity, and digital calculation error. Laboratory calibration shows repeatability better than ±0.15 percentage points, significantly better than traditional single-path weighing methods or single-channel thermography.

[0069] Below is a complete example of a laboratory-scale implementation of a real-time detection method for liquid sample crystallization based on smart sensors. All volumes are in milliliters (mL), masses in grams (g), times in seconds (s) or minutes (min), energy in joules (J) or watts (W), and temperatures in degrees Celsius (°C).

[0070] Step 1: Low-Temperature Solution Preparation and Homogenization: In the solution preparation room, a stainless steel thermostatic solution preparation tank with an inner diameter of 120 mm and an effective height of 300 mm is pre-cooled to -20°C. The tank adopts a double-jacketed structure and is connected to a circulating refrigeration unit, with a flow rate set to 3.5 L·min. The refrigerant was a 20% ethylene glycol-water mixture. After the temperature control system stabilized for 5 minutes, the temperature fluctuation inside the tank remained within ±0.2°C. Toluene (40 mL), acetone (20 mL), and standard solvent oil (40 mL) were added sequentially to the tank, totaling 100 mL, with volume fractions of 0.4% for toluene, 0.2% for acetone, and 0.4% for solvent oil. This ratio falls within the permissible range of "2–6 parts toluene, 1–4 parts acetone, and 3–7 parts solvent oil" as specified in the claims. Another 100 mL sample of the crude oil to be tested was pre-cooled to -20°C. It was continuously injected into the tank through the bottom feed pipe over 90 seconds (meeting the "1 min–2 min" window). Axial average flow rate... Reynolds number At the lower limit of the laminar-turbulent critical flow, the temperature gradient surge can be effectively suppressed. Immediately start the in-tank paddle agitator, setting the speed to 300 r·min. The duration is 120 seconds. The blade diameter is 80mm, corresponding to a tip linear velocity of... The stirring power is determined by the impeller power coefficient. Estimate:

[0071] .

[0072] Stop the machine immediately after stirring, cover and let stand for 30 seconds to eliminate turbulence memory. A homogeneous toluene-acetone adjustable ratio ternary dewaxing medium-crude oil mixture system is thus obtained, with a total volume of 200 mL.

[0073] Step 2: Multiphysics-induced crystallization:

[0074] The mixed system was transported to a cryogenic constant-pressure settling device via a vacuum-insulated PTFE pipe with an inner diameter of 8 mm. The pipe was 0.6 m long, and the flow time was 12 s. Environmental heat release was negligible. The settling cylinder had an effective height of 400 mm and an inner diameter of 60 mm. The jacket was kept at a constant temperature using a -20°C refrigerant. The vertical ultrasonic transducer array was first activated with a center frequency of 40 kHz, an input power of 200 W, and irradiation for 3 minutes. The average bubble radius of acoustic cavitation was approximately 2 µm. The cavitation energy density was estimated. Immediately activate the mechanical spiral shear blades, coaxial and in the same direction, at a speed of 300 r / min. The shearing process lasted for 120 seconds. Because the system already contained a large number of microcrystalline nuclei, this shearing primarily aimed to homogenize the particle size and suppress agglomeration. Subsequently, a DC current of 3.2A was applied to an external electromagnetic coil with 800 turns, and the central magnetic induction intensity was measured. The magnetic field direction was aligned with the axis of the settling cylinder and lasted for 300 seconds. The magnetic energy was relatively small compared to the ultrasonic-shear energy, but it could generate an orientation moment for trace amounts of iron-containing colloidal-wax eutectic nuclei, reducing the surface free energy by approximately 6%–8%.

[0075] Step 3: Settling and Signal Processing: After terminating all external field operations, the settling system was allowed to settle. A laser scattering density recorder sampled every 20 seconds, with a laser wavelength of 635nm and a detection path length of 50mm. The entire recording period lasted 60 minutes, acquiring a total of 180 optical density (OD) data points. The curve plateaued around 57 minutes, at which point acquisition ceased. The OD curve was divided according to the rule of "5 minutes = 15 points," resulting in 12 time windows. The following example uses window 7 (35min–40min): Original OD vector .

[0076] In the real part of the complex plane Select 20 nodes Call the pre-stored Laplacian kernel weight matrix. right Perform a discrete Laplace function to generate coefficients. Linear interpolation reconstructs a smooth baseline. Residual vector Shaped like .right Performing a three-level orthogonal wavelet packet decomposition (Daubechies-4) yields four subbands. Calculate energy density Assume this window receives... , , , Set a threshold. Therefore, we chose This serves as the wax precipitation signal flow for that window. The signal flow coefficients selected from each of the 12 windows are weighted according to a weighting table. Fusion, resulting in a fusion decision curve. .when The first occurrence of three consecutive sampling points satisfying The corresponding time was 46.7 minutes, which was considered the endpoint of the wax precipitation. .

[0077] Step 4: Separation and Mass-Heat Dual-Channel Differential Subtraction: Within the last 2 minutes, the liquid was drained and the wax-containing layer was scraped off, yielding a total of 18.304g of wax blocks and residual liquid. The initial mass was then absorbed using fiber-free absorbent paper to remove surface oil. Place in a 60°C far-infrared constant temperature chamber and weigh at regular intervals:

[0078] ,when Time stability. Quality measurement result sequence. Take another wax block of equal mass, 17.986g, and place it into a DSC aluminum dish. Incubate at –10°C → 80°C at a rate of 2°C / min. Run DSC to record the endothermic power in the 30°C–70°C range. Integrating yields the heat absorption sequence. .right , Take the first-order difference and amplify it. , and then After the sequence is circularly shifted right by 5 bits and then... Perform bit-XOR combination on the difference sequences to generate the fused sequence. .right The monotonically increasing sequence is obtained by performing step-by-step accumulation. The deterministic pivot index is obtained by scanning. Let the wax content be the mass fraction. This example calculates to obtain... , ,then Retaining three significant figures, the final reported wax content of the crude oil sample is 7.34%.

[0079] After settling, the supernatant was added to a colorimetric reagent solution at a volume ratio of 1:1, and the mixture was reacted in a 40°C water bath for 10 min to form a yellow complex. The reaction system was transferred to a cuvette, and the absorbance was measured at a wavelength of 420 nm. The residual wax content in the supernatant was calculated based on the pre-established standard absorbance curve. The colorimetric result was obtained by subtracting the residual wax content in the supernatant from the wax content obtained in step 4. The colorimetric reagent consisted of p-nitrobenzoic acid and ethanol at a concentration of 2.0 mol·L⁻¹.

[0080] Figure 2 This diagram illustrates the principle of wax nucleus formation in a toluene-acetone tunable-ratio ternary dewaxing medium. It details the physicochemical mechanism of the mixed system formation under constant-temperature solution preparation conditions in step 1. Figure 2As shown, the constant-temperature mixing tank adopts a double-layer cooling jacket structure. A low-temperature refrigerant is introduced into the outer jacket, and a circulating refrigeration unit maintains the tank temperature at a constant -20 degrees Celsius. Inside the tank, a toluene-acetone adjustable-ratio ternary dewaxing medium is prepared according to a predetermined volume ratio, with toluene comprising 2-6 parts, acetone 1-4 parts, and standard solvent oil 3-7 parts. The figure uses different geometric shapes to represent the molecules of each component: circles represent toluene molecules, triangles represent acetone molecules, squares represent solvent oil molecules, and ellipses represent crude oil wax molecules. Under constant-temperature conditions, the liquid sample is slowly injected through the bottom feed pipe, with the injection time controlled between 1-2 minutes to prevent temperature gradient fluctuations. Subsequently, a fixed-speed paddle stirrer inside the tank is started, with the speed set to 300 rpm, and stirring is maintained for 2 minutes. Under the action of stirring, the ternary dewaxing medium formed by toluene, acetone, and solvent oil comes into full contact with and mixes uniformly with the liquid sample. From a molecular level perspective, toluene, as an aromatic hydrocarbon solvent, exhibits excellent solubility, while acetone, as a polar solvent, can adjust the polarity of the solvent system. Solvent oil serves to dilute and regulate viscosity. At a low temperature of -20 degrees Celsius, the solubility of wax molecules in crude oil significantly decreases, creating favorable conditions for the subsequent wax precipitation process. After stirring, stirring was immediately stopped and the tank lid was sealed. The mixture was allowed to stand for 30 seconds to eliminate turbulence, confirming that the color and turbidity of the mixture remained uniform and without stratification. This resulted in a stable toluene-acetone adjustable-ratio ternary dewaxing medium-crude oil mixture, preparing for the next step of the physical field-induced wax precipitation process.

[0081] Figure 3This paper demonstrates the mixing process and homogenization effect of a toluene-acetone adjustable-ratio ternary dewaxing medium and a liquid sample in a thermostatic mixing tank. The figure is divided into three stages, clearly depicting the complete transformation from a stratified state to a homogeneous mixture. The pre-mixing state shows the initial configuration within the thermostatic mixing tank, where the toluene-acetone adjustable-ratio ternary dewaxing medium is in the upper layer, appearing as a relatively transparent liquid phase, while the liquid sample is in the lower layer, with a clear interface between the two phases. At this point, the system temperature is maintained at a constant -20°C, with temperature stability ensured by the synergistic effect of a double-layer cooling jacket and a circulating refrigeration unit. The presence of stratification indicates that the density difference and mutual solubility of the two liquids are limited, requiring mechanical stirring to achieve homogeneous mixing. The stirring and mixing stage demonstrates the hydrodynamic effects of a fixed-speed paddle mixer. The paddle operates at a constant speed of 300 rpm for 2 minutes, creating a complex flow field distribution within the tank. The curved streamlines in the figure clearly show the liquid's trajectory, including radial flow and axial circulation. This three-dimensional flow pattern effectively promotes mass transfer and mixing between liquids of different phases. During stirring, shear forces disrupt the phase interface, causing the toluene-acetone adjustable ratio ternary dewaxing medium to form a smaller-scale dispersed phase with the crude oil, gradually leading to homogenization. The post-mixing state shows the homogeneous mixed system formed after thorough stirring. At this point, the liquid in the tank exhibits a completely uniform color and turbidity, with no stratification, indicating that molecular-level homogeneous dispersion has been achieved. The uniformly distributed tiny black dots in the figure represent the waxy components precipitated from the crude oil. These waxes begin to crystallize and precipitate under low temperature and solvent conditions, laying the foundation for subsequent sedimentation-wax precipitation detection. The formation of the mixed system marks the successful completion of the first step, creating ideal initial conditions for subsequent external physical field induction in the sedimentation unit.

[0082] Figure 4Four curves with different characteristics illustrate the significant impact of the toluene-acetone volume ratio on the optical density variation during the detection of crude oil wax content. The horizontal axis represents time in minutes, ranging from 0 to 50 minutes; the vertical axis represents optical density in OD, ranging from -0.6 to +1.0, comprehensively recording the evolution of optical properties throughout the entire sedimentation-wax precipitation process. At a toluene-acetone volume ratio of 2:1, the optical density curve exhibits the steepest downward trend. Starting from an initial value of 0.95 OD, the curve rapidly decreases to around 0.15 OD within the first 25 minutes, then levels off, eventually stabilizing at approximately 0.07 OD. This rapid change indicates that the wax precipitation rate is fastest at this ratio, and the wax precipitation endpoint appears earliest, reaching its peak around 25 minutes. The rapid decrease in optical density reflects the rapid formation and aggregation of wax crystal nuclei, leading to a significant improvement in solution transmittance. When the toluene-acetone volume ratio is adjusted to 3:2, the optical density variation curve shows a relatively gentle downward pattern. The initial OD value was approximately 0.96, which gradually decreased over 35 minutes before stabilizing at 0.11 OD. Compared to a 2:1 ratio, the slope of this curve was significantly lower, indicating a more gradual wax precipitation process. This suggests that a moderate acetone content is beneficial for controlling the wax precipitation rate and providing a longer observation window. With a toluene-acetone volume ratio of 4:3, the optical density curve showed a smoother change. Throughout the process, the OD value slowly decreased from 0.97 to 0.13, with the wax precipitation endpoint delayed to approximately 35 minutes. This gentler trend facilitates accurate capture of the wax precipitation signal and reduces measurement errors. When the volume ratio reached 6:4, the optical density change was the slowest, requiring approximately 40 minutes to decrease from 0.98 OD to 0.15 OD. While an excessively high acetone ratio extended the detection time window, it could also affect the wax precipitation efficiency. The comparison of the four curves clearly demonstrates that optimizing the toluene-acetone volume ratio is crucial for achieving ideal detection results, providing an important theoretical basis for parameter selection in practical applications.

[0083] The present invention has been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of the invention. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and core ideas of the present invention. It should be noted that those skilled in the art can make various improvements and modifications to the present invention without departing from its principles, and these improvements and modifications also fall within the protection scope of the claims of the present invention.

Claims

1. A method for real-time detection of crystallization precipitation in liquid samples based on intelligent sensors, characterized in that, The method includes: Step 1: Combine the crude oil sample with a toluene-acetone adjustable ratio ternary dewaxing medium composed of toluene, acetone and solvent oil according to a predetermined volume ratio, and form a homogeneous mixture system in a constant temperature solution preparation environment; Step 2: Place the mixed system into a constant temperature and pressure settling device and expose it to at least two external physical fields in sequence, namely ultrasonic field, mechanical shear field and magnetic field, to accelerate crystal nucleus formation and growth; Step 3: After terminating the external physical field, the mixed system is allowed to settle. The intelligent sensor collects absorbance data at fixed time intervals. The embedded processor first divides the absorbance data into fixed time windows, performs discrete Laplace operation on the absorbance sequence within each time window to remove the settling background, and then performs three-level orthogonal wavelet packet decomposition on the Laplace residual sequence to extract the crystallization precipitation feature signal. The crystallization precipitation endpoint is determined according to the threshold rule. The intelligent sensor includes a light source component, a detection cell, and an intelligent photoelectric sensor component. Step 4: At the end of crystallization, separate the lower layer containing sediment from the supernatant; perform mass and microcalorimetry on the lower sediment, assemble all mass values ​​into a mass measurement result sequence in chronological order, and assemble all endothermic values ​​into a microcalorimetry result sequence in chronological order. Use a dual-channel difference-subtraction integral method to synthesize the mass measurement result sequence and the microcalorimetry result sequence, and use the crystallization precipitation amount as the target component crystallization precipitation amount after calibration; Step 5: Take the supernatant obtained after settling and react it with the colorimetric reagent solution to form a yellow complex. Place the reaction system in a cuvette and measure the absorbance at a wavelength of 420 nm. Calculate the residual target component content in the supernatant based on the preset standard absorbance curve. Combine the amount of crystallization obtained in Step 4 with the residual amount in the supernatant to calculate and output the total target component content of the liquid sample.

2. The real-time detection method for liquid sample crystallization precipitation based on intelligent sensors as described in claim 1, characterized in that, Step 1 specifically includes: Step 1.1: Provide a constant temperature solution preparation tank with a double-layer cooling jacket and connected to a circulating refrigeration unit; add the components of the sample pretreatment solvent system to the constant temperature solution preparation tank in sequence. In this example, the components are toluene, acetone, and standard solvent oil, and the volume fractions of the three are set as 2-6 parts for toluene, 1-4 parts for acetone, and 3-7 parts for standard solvent oil, respectively; introduce a low-temperature coolant into the jacket of the constant temperature solution preparation tank to keep the temperature inside the tank constant at -20 degrees Celsius and maintain it for no less than 5 minutes to achieve temperature equilibrium; Step 1.2: Take a liquid sample with a volume equal to the total volume of the sample pretreatment solvent system and slowly inject it into the same constant temperature preparation tank through the bottom feed tube under constant temperature conditions; Step 1.3: Start the fixed-speed paddle stirrer and maintain stirring for 2 minutes to ensure that the sample pretreatment solvent system and the liquid sample are in full contact and form a uniform mixture. Step 1.4: Immediately stop stirring and seal the container after stirring is complete. Let it stand for 30 seconds to eliminate turbulence. Confirm that the mixed system in the container has uniform color and turbidity and no stratification. The mixed system is now obtained.

3. The real-time detection method for liquid sample crystallization precipitation based on intelligent sensors as described in claim 2, characterized in that, In step 1.2, the time for injecting the liquid sample into the constant temperature mixing tank is controlled between 1 and 2 minutes to prevent temperature gradient fluctuations; in step 1.3, the rotation speed of the fixed speed paddle stirrer is set to 300 revolutions per minute.

4. The real-time detection method for crystallization precipitation of liquid samples based on intelligent sensors as described in claim 3, characterized in that, In step 2, the mixture is injected into a cryogenic, constant-pressure settling device via an insulated pipeline. The internal temperature is maintained at -20 degrees Celsius through a refrigerant circulation system. Within the settling device, the mixture is exposed to at least two external physical fields: an ultrasonic field, a mechanical shear field, and a magnetic field, inducing the formation of wax crystal nuclei. When exposed to an ultrasonic field, a vertical ultrasonic transducer array is activated, exposing the mixture to a vertical ultrasonic field with a frequency of 40 kHz and a power of 200 watts for 3 minutes. The irradiation direction of the ultrasonic field is parallel to the direction of gravity. When exposed to a mechanical shear field, a mechanical helical shear blade is activated, placing the mixture in a mechanical shear field with a rotation speed of 300 revolutions per minute for 2 minutes. The rotation direction of the mechanical helical shear blade is the same as the axis of the settling device. When exposed to a magnetic field, a direct current is passed through an external coil of the settling device, generating a constant magnetic field with a magnetic induction intensity of 0.1 Tesla and a magnetic field direction in the same direction as the axis of the settling device for 5 minutes.

5. The real-time detection method for liquid sample crystallization precipitation based on intelligent sensors as described in claim 4, characterized in that, Step 3 specifically includes: Step 3.1: After terminating the external physical field, allow the material to settle. The absorbance curve inside the settling cylinder is recorded by a smart sensor set at 20-second intervals. The data is continuously collected until a plateau appears on the curve. Step 3.2: Cut the obtained absorbance curve into a fixed 5-minute time window, perform discrete Laplace operation on the absorbance sequence within each time window to remove the sedimentation background, and then perform three-level orthogonal wavelet packet decomposition on the Laplace residual sequence to extract the crystallization precipitation characteristic signal. Step 3.3: Linearly fuse the wavelet coefficients of the wavelet packet signal stream according to the pre-stored coefficient table to obtain the fusion decision curve; Step 3.4: When the fusion determination curve is lower than the predetermined threshold for the first three consecutive sampling points, it is considered as the end point of crystallization precipitation.

6. The real-time detection method for liquid sample crystallization precipitation based on intelligent sensors as described in claim 5, characterized in that, Step 3.2 specifically includes: cutting the obtained absorbance curve sequentially according to a fixed time window rule of 5 minutes; each time window contains 15 absorbance sampling points; establishing a one-dimensional time series for the 15 sampling points in each time window according to the sampling point order; using discrete Laplace operation based on numerical table integration, selecting 20 equidistant operation nodes in the range of 0 to 5 of the real part coordinates of the complex plane, and using each operation node as a variable for one Laplace operation; using each operation node as a reference, multiplying each of the 15 sampling points item by item according to the preset Laplace kernel function weight matrix and... Accumulated, 20 Laplace transform coefficients are obtained; one-dimensional linear interpolation is used to reconstruct these equidistant coefficients into a smooth curve, which serves as the Laplace line for the current time window; the original absorbance sampling point value is subtracted from the corresponding Laplace line value at each time point to obtain a residual vector of length 15, which serves as the Laplace residual sequence; three-level orthogonal wavelet packet decomposition is performed on the residual sequence to obtain four wavelet packet sub-band signal streams; the energy density value is calculated for each of the four sub-band signal streams; the sub-band signal stream with the highest energy density value that is greater than a preset threshold is selected as the crystallization precipitation characteristic signal.

7. The real-time detection method for crystallization precipitation of liquid samples based on intelligent sensors as described in claim 6, characterized in that, When performing three-level orthogonal wavelet packet decomposition on the Laplace residual sequence, a Darbage orthogonal basis with four vanishing moments is selected as the wavelet packet mother basis: the first level decomposes the residual sequence into one low-frequency sub-band and one high-frequency sub-band; the second level decomposes the low-frequency sub-band obtained in the first level into one new low-frequency sub-band and one new high-frequency sub-band, while keeping the high-frequency sub-band of the first level undecomposed; the third level continues to decompose the low-frequency sub-band obtained in the second level into one low-frequency sub-band and one high-frequency sub-band, while keeping the remaining sub-bands obtained previously undecomposed, thus forming a total of four wavelet packet sub-band signal streams.

8. The real-time detection method for crystallization precipitation of liquid samples based on intelligent sensors as described in claim 6, characterized in that, Step 4 specifically includes: Step 4.1: Keep the internal temperature of the sedimentation device constant at -20 degrees Celsius, open the sampling valve 5 mm from the bottom on the side wall of the sedimentation device, and use the liquid level difference to discharge the upper supernatant until the lower sediment is exposed; close the valve, use a polytetrafluoroethylene coated scraper to scrape off the sediment, and transfer it to a pre-cooled stainless steel sample dish to complete the separation. Step 4.2: Place the sediment sample on an electronic analytical balance, wipe off the free liquid on the surface with non-fibrous absorbent paper, and weigh the initial mass. Then, place the sample and the sample dish in a 60°C far-infrared constant temperature oven for constant weight drying. Take it out and weigh it every 5 minutes and record the mass value. When the relative difference between two consecutive weighing results does not exceed 0.1%, it is considered that the constant weight state has been reached. Arrange all mass values ​​in chronological order to form a mass measurement result sequence. Step 4.3: Take another sediment sample of equal mass and put it into a sealed aluminum dish of differential scanning calorimeter. Under the condition of isothermal starting point of -10 degrees Celsius, heat it to 80 degrees Celsius at a heating rate of 2 degrees Celsius per minute. Record the heat absorption value in the range of 30 degrees Celsius to 70 degrees Celsius in real time. Arrange all heat absorption values ​​in time order to form a sequence of microcalorimetric measurement results. Step 4.4: The mass measurement result sequence and the microcalorimetry result sequence are synthesized using the dual-channel difference-integral method. The amount of crystallization is calibrated and then used as the amount of crystallization of the target component.

9. The real-time detection method for crystallization precipitation of liquid samples based on intelligent sensors as described in claim 8, characterized in that, Step 4.4 specifically includes: performing adjacent difference operations on the mass measurement result sequence to generate a primary difference sequence; performing the same adjacent difference operation on the microcalorimetry result sequence to generate a corresponding primary difference sequence; converting the two primary difference sequences into a list of unsigned integers magnified by a thousand times according to their time indices, and then performing a binary XOR combination of cyclic right shift by 5 bits in sequence to obtain a fused sequence; performing stepwise accumulation on the fused sequence from the first term to obtain a monotonically increasing sequence; scanning the monotonically increasing sequence to locate the smooth segment with the longest continuous value increase, and using the end point of this segment as the deterministic pivot; using the sum of all accumulated values ​​before the deterministic pivot as the local integral and the total value of the sequence as the global integral, and calculating the ratio between the two; multiplying the obtained ratio by 100 to obtain the calibrated target component mass percentage, retaining 3 significant figures.

Images