Method and System for Constructing Autonomous Phase Mapping Based on Cellular Automata

Abstract

The present invention provides a method and system for constructing an autonomous phase map based on cellular automata, including: Step S1: preparing a compositionally continuous composite material sample based on material combination technology; Step S2: performing X-ray diffraction characterization on the characterization target points, and initially the characterization target points are the corner points of the sample; Step S3: performing real-time analysis on the X-ray diffraction data to achieve real-time phase identification; Step S4: feeding the phase identification result back into the cellular automata algorithm, and using the cellular automata algorithm to determine the next characterization target point; repeating Steps S2 to S4 until all phase boundaries are determined.

Description

Technical Field

[0001] The present invention relates to the technical field of closed-loop feedback autonomous experimental design. Specifically, it relates to a method and system for constructing an autonomous phase map based on cellular automata. More specifically, it relates to an autonomous phase diagram construction algorithm that combines combinatorial material technology and cellular automata algorithm. Background Art

[0002] Phase mapping is the basis for material research and also the blueprint for material design and process optimization. Currently, phase mapping is mainly determined through costly experiments or calculations. The traditional experimental method for constructing a phase distribution map uses a grid search method to traverse all sample points. However, its disadvantage is low efficiency, requiring a large amount of experimental resources and consuming a large amount of time. A promising approach to address this challenge is combinatorial material technology.

[0003] Combinatorial material technology involves high-throughput preparation, high-throughput characterization, and data processing technology based on machine learning. It has revolutionized the way of constructing phase maps and the construction of multi-parameter coupling relationships between composition, structure, and properties. In addition, a closed-loop autonomous system for material exploration and optimization provides a new paradigm for accelerating material screening. It involves an adaptive algorithm that can update experimental design in real time. The same method has been proven to be very effective in phase map construction. Currently, closed-loop autonomous systems are mainly implemented using machine learning algorithms such as Bayesian optimization, active learning, uncertainty sampling, and so on. Although machine learning algorithms provide powerful tools for phase map construction, they require a large amount of computer programming, data science expertise, and parameter adjustment. Therefore, developing a simple algorithm without parameters is an ideal goal.

[0004] Cellular automata is a dynamic evolution algorithm based on a discrete network. It evolves according to local rules and generates complex pattern behaviors from simple rules. Cellular automata have been used in fields such as recrystallization research, solidification simulation, fuzzy classification, and biological modeling. Since the cellular automata algorithm mainly relies on logical operations, it is much more efficient than machine learning in terms of computational time and cost, and is more conducive to autonomous experiments. One of the most attractive features of the cellular automata algorithm is that there are no parameters to be adjusted. In contrast, machine learning algorithms have many built-in parameters that need to be tuned. Due to the continuity of the discrete sampling point distribution, phase boundaries can be constructed based on local rules. Therefore, the cellular automata algorithm can be used as an effective tool for determining phase boundary points.

[0005] In the prior art, the journal "Science and Technology of Advanced Materials: Methods" published a high-throughput method for constructing phase diagrams based on uncertainty sampling on pages 153-161 of Volume 2, Issue 1, 2022; the journal "npj Computational Materials" published a method for reconstructing phase diagrams from local measurements using Gaussian processes in Article No. 23 of Volume 4, Issue 1, 2022; the journal "The Journal of Physical Chemistry B" published an efficient phase diagram sampling method through active learning on pages 1275-1284 of Issue 124, 2020; the journal "Materials & Design" published a phase diagram prediction classification method based on uncertainty sampling in Issue 215, 2022, Article No. 110497. Although the above machine learning methods have all achieved efficient construction of phase diagrams, they all involve a large amount of computer programming, a large amount of data science knowledge, and a parameter tuning process, which pose high requirements for the computing power of researchers and computers. In addition, machine learning algorithms often introduce uncertainties and sometimes lead to errors in phase diagram construction. In addition, when faced with complex systems and a large number of data points, the computational cost of machine learning often becomes extremely high. Summary of the Invention

[0006] Aiming at the defects in the prior art, the purpose of the present invention is to provide an autonomous phase mapping construction method and system based on cellular automata.

[0007] An autonomous phase mapping construction method based on cellular automata provided by the present invention includes:

[0008] Step S1: Prepare a compositionally continuous composite material sample based on the material combination technology;

[0009] Step S2: Perform X-ray diffraction characterization on the characterization target points, and initially the characterization target points are the corner points of the sample;

[0010] Step S3: Perform real-time analysis on the X-ray diffraction data to achieve real-time phase identification;

[0011] Step S4: Feed the phase identification result back into the cellular automata algorithm, and use the cellular automata algorithm to determine the next characterization target point; repeat Steps S2 to S4 until all phase boundaries are determined.

[0012] Preferably, Step S3 adopts:

[0013] Step S3.1: Convert the original two-dimensional diffraction spectrum obtained in the X-ray diffraction characterization into a one-dimensional intensity-2θ spectral line;

[0014] Step S3.2: Preprocess the one-dimensional intensity-2θ spectral line to obtain a smoothed curve with background interference eliminated;

[0015] Step S3.3: Use the built-in function of scipy in python to preliminarily lock the peak positions of the characteristic peaks from the smoothed curve;

[0016] Step S3.4: Based on the preliminarily locked characteristic peaks, use the combination of multiple Gaussian peaks for fitting according to the least squares optimization algorithm to obtain the accurate peak positions, peak heights and full-width at half-maximums of the characteristic peaks;

[0017] Step S3.5: Screen the characteristic peaks according to the preset thresholds for the peak heights and full-width at half-maximums of the obtained characteristic peaks, and retain the prominent and sharp crystallization peaks;

[0018] Step S3.6: Collect the PDF cards of the possible phases in the current system from the ICSD database, compare the peak position data of the obtained crystallization peaks with the PDF cards one by one, thereby determine the phase composition of the current sample point, and mark it with the corresponding phase label.

[0019] Preferably, the step S3.2 adopts:

[0020] Step S3.2.1: The one-dimensional intensity-2θ spectral line is smoothed by Gaussian convolution filtering with a 1x7 discrete Gaussian function kernel to remove the high-frequency noise caused by statistical fluctuations;

[0021] Step S3.2.2: Further process using the rolling ball algorithm to obtain a curve with background interference eliminated.

[0022] Preferably, the cellular automaton algorithm adopts:

[0023] Step S4.1: Use the dichotomy method to find a pair of adjacent points on the potential phase boundary based on the identified phase labels;

[0024] Step S4.2: Construct the phase boundary based on the pair of adjacent points obtained on the phase boundary and the principle of phase boundary continuity, and repeat steps S4.1 to S4.2 until all phase boundaries are found.

[0025] Preferably, the step S4.1 adopts: When the recently explored nearest neighbor sample points in the phase mapping edge region belong to different phase labels, there is a phase boundary between them; use the dichotomy method to find a pair of adjacent points on the potential phase boundary; when it is checked that there is no longer a potential phase boundary, end the process.

[0026] Preferably, in step S4.2, the following is adopted: define the sample points with different and adjacent phase class labels at the phase boundary as the growth base points; continuously extend the phase boundary based on the iteration rule of the growth base points until the boundary grows to the boundary of the phase diagram or the boundary forms a closed loop, and the iteration of the growth base points will stop; when the iteration of all growth base points stops, a continuous phase boundary is constructed.

[0027] An autonomous phase mapping construction system based on cellular automata provided by the present invention includes:

[0028] Module M1: Prepare a combined material sample with continuous composition based on material combination technology;

[0029] Module M2: Perform X-ray diffraction characterization on the characterization target points. Initially, the characterization target points are the corner points of the sample;

[0030] Module M3: Perform real-time analysis on the X-ray diffraction data to achieve real-time phase identification;

[0031] Module M4: Feed the phase identification result back into the cellular automata algorithm, and use the cellular automata algorithm to determine the next characterization target point; repeat Modules M2 to M4 until all phase boundaries are determined.

[0032] Preferably, the following is adopted for Module M3:

[0033] Module M3.1: Convert the original two-dimensional diffraction spectrum obtained in the X-ray diffraction characterization into a one-dimensional intensity - 2θ spectrum line;

[0034] Module M3.2: Preprocess the one-dimensional intensity - 2θ spectrum line to obtain a smoothed curve with background interference eliminated;

[0035] Module M3.3: Use the built-in function of scipy in python to initially lock the peak positions of the characteristic peaks from the smoothed curve;

[0036] Module M3.4: Based on the initially locked characteristic peaks, use the combination of multiple Gaussian peaks for fitting according to the least squares optimization algorithm to obtain the accurate peak positions, peak heights, and full-width at half-maximum of the characteristic peaks;

[0037] Module M3.5: Screen the characteristic peaks according to the preset threshold for the peak heights and full-width at half-maximum of the obtained characteristic peaks, and retain the prominent and sharp crystallization peaks;

[0038] Module M3.6: Collect the PDF cards of the possible phases in the current system from the ICSD database, compare the obtained peak position data of the crystallization peaks with the PDF cards one by one, thereby determine the phase composition of the current sample point, and mark it with the corresponding phase class label.

[0039] Preferably, the following is adopted for Module M3.2:

[0040] Module M3.2.1: The one-dimensional intensity-2θ spectrum is smoothed by Gaussian convolution filtering with a 1x7 discrete Gaussian function kernel to remove high-frequency noise caused by statistical fluctuations.

[0041] Module M3.2.2: Further process using the rolling ball algorithm to obtain a curve with background interference removed.

[0042] Preferably, the cellular automaton algorithm adopts:

[0043] Module M4.1: Based on the identified phase class labels, use the bisection method to find a pair of adjacent points on the potential phase boundary.

[0044] Module M4.2: Based on a pair of adjacent points on the obtained phase boundary and the principle of phase boundary continuity, construct the phase boundary. Repeat steps S4.1 to S4.2 until all phase boundaries are found.

[0045] Module M4.1 adopts: When the recently explored nearest neighbor sample points in the phase mapping edge region belong to different phase class labels, there is a phase boundary between them; use the bisection method to find a pair of adjacent points on the potential phase boundary; when it is checked that there is no longer a potential phase boundary, end the process.

[0046] Module M4.2 adopts: Define the sample points with different and adjacent phase class labels at the phase boundary as the growth base points; continuously extend the phase boundary based on the iterative rules of the growth base points until the boundary grows to the boundary of the phase diagram or the boundary forms a closed loop, and the iteration of the growth base points will stop; when the iteration of all growth base points stops, a continuous phase boundary is constructed.

[0047] Compared with the prior art, the present invention has the following beneficial effects: By adopting a closed-loop feedback experimental workflow constructed by combining the combinatorial materials technology and the cellular automaton algorithm, the present invention solves the problems of time-consuming, laborious and high cost of traditional phase mapping methods, and achieves the effect of efficiently and autonomously constructing phase mapping. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] By reading the detailed description of the non-limiting embodiments with reference to the following drawings, other features, objectives and advantages of the present invention will become more obvious:

[0049] Figure 1 It is a schematic diagram of the overall workflow for autonomous construction of phase mapping.

[0050] Figure 2 It is a schematic diagram of the adjacent relationship in the cellular automaton, which defines the adjacent relationship in the algorithm.

[0051] Figure 3They are three situations in the iterative process of Mode II, including that C and A belong to the same phase, C and B belong to the same phase, and C belongs to neither the same phase as A nor the same phase as B.

[0052] Figure 4 It is the overall process framework diagram of the cellular automaton algorithm. Specific implementation manners

[0053] The present invention will be described in detail below in conjunction with specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but do not limit the present invention in any form. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several changes and improvements can still be made. These all belong to the protection scope of the present invention.

[0054] Example 1

[0055] An autonomous phase mapping construction method and system based on cellular automaton provided by the present invention can quickly identify phase boundaries through the cellular automaton algorithm, and realize efficient autonomous construction of phase mapping in combination with combinatorial materials technology. Therefore, only by determining the phase boundary points on the combinatorial materials chip based on known data as the characterization targets can the efficiency be improved while ensuring the construction of the phase diagram. This workflow is based on combinatorial materials technology, and finds phase boundaries through a self-developed cellular automaton algorithm to realize autonomous construction of phase mapping. Combining this method with real-time data analysis can realize a true dynamic materials development model.

[0056] The autonomous phase mapping construction method based on cellular automaton, as Figure 1 shown, includes:

[0057] Step S1: Prepare a combinatorial materials sample with continuous composition based on materials combinatorial technology;

[0058] Step S2: Perform X-ray diffraction characterization on the characterization target points. Initially, the characterization target points are the corner points of the sample;

[0059] Step S3: Perform real-time analysis on the X-ray diffraction data to achieve real-time phase identification;

[0060] Step S4: Feed the phase identification result back into the cellular automaton algorithm, and use the cellular automaton algorithm to determine the next characterization target point; repeat steps S2 to S4 until all phase boundaries are determined.

[0061] Specifically, the step S3 adopts:

[0062] Using custom software from the "xrayutilities" library in Python, the original two-dimensional (2D) diffraction spectra obtained in XRD characterization are converted into one-dimensional intensity-2θ (2θ) spectral lines. On the premise of certain conditions, first, the diffraction profile corresponding to the amorphous silicon substrate measured in the same experiment is subtracted from the one-dimensional profile, and then the rolling ball algorithm is used for further processing to obtain a curve with background interference removed. Then, Gaussian convolution filtering with a 1×7 discrete Gaussian function kernel is used to smooth the curve to remove high-frequency noise caused by statistical fluctuations. By using the built-in functions of scipy in Python, the peak positions (2θ) and other characteristics, including peak height and full width at half maximum, are extracted from the smoothed curve. Then, a better curve fitting method is used to locate those peaks that are very close and difficult to distinguish: at the position where two characteristic peaks almost overlap, the curve within a small range of 2θ is fitted with a combination of multiple Gaussian peaks based on the least squares optimization algorithm. According to the optimized fitting parameters, the positions, peak heights, and full widths at half maximum of the peaks are obtained. The characteristic peaks are screened according to the preset threshold for the peak height and full width at half maximum of the obtained characteristic peaks, and prominent and sharp crystalline peaks are retained; then, PDF cards of the possible phases in this system are collected from the ICSD database, and the obtained peak position data are compared with the PDF cards one by one to determine the phase composition of the sample point and mark it with the corresponding label.

[0063] Specifically, the cellular automaton algorithm includes: when performing cellular automaton analysis, a grid needs to be created to divide the combined chip into multiple cells to establish the adjacent relationship between sample points. Since the distribution of any sample points can be divided into a triangular network using the Delaunay triangulation method, a cellular automaton algorithm based on triangular cells is designed. In the triangular grid cells, if two points belong to the same cell, they are considered adjacent. As Figure 2 shown, all the pentagrams in the figure are adjacent points of the dot.

[0064] The process framework of this cellular automaton algorithm is as Figure 4 shown. This algorithm consists of two modes, each mode contains a set of evolution (iteration) rules, and then different evolution behaviors are generated respectively.

[0065] Mode Ⅰ: Search for potential phase boundaries based on the dichotomy method. Mode Ⅰ only examines the edge region of the phase map, that is, the connecting line region between the vertices of the phase diagram. When the explored nearest neighbor sample points in this region belong to different phase class labels, there must be a phase boundary between the two. The dichotomy method is used to find a pair of adjacent points on the potential phase boundary, that is, each time their midpoint is explored until the two points are adjacent. When a pair of adjacent points on the potential phase boundary are found, switch to Mode Ⅱ; if it is checked that there are no longer potential boundaries, the edge process ends, and the data points with default phase class labels are filled to complete the construction of the phase map.

[0066] Mode II: Construct the phase boundary based on the continuity of the phase boundary. Define two adjacent sample points with different phase class labels as the growth base points, which are actually part of the phase boundary. There must be 1 - 2 common neighboring sample points for these two sample points, and this condition enables the iteration to continue. As Figure 3 shown, points A and B are two adjacent sample points with different phase class labels, and point C is their common neighboring point. Then there are three cases for C: a. C belongs to the same phase as A, then B and C form a new growth base point, and E becomes the new point to be characterized; b. C belongs to the same phase as B, then A and C form a new growth base point, and D becomes the new point to be characterized; c. C does not belong to the same phase as either A or B, then A and C, B and C respectively form new growth base points, and both D and E become new points to be characterized. Because the phase region is continuous, the phase boundary is also continuous. Therefore, a complete phase boundary can be obtained by continuously extending this iteration, similar to the growth of a vine. When the boundary grows to the boundary of the phase diagram or the boundary forms a closed loop, the iteration of the growth base points will stop. When the iteration of all growth base points stops, the growth of a continuous phase boundary is completed, and then switch to Mode I to continue searching for potential boundaries.

[0067] The autonomous phase mapping construction system based on the cellular automaton includes:

[0068] Module M1: Prepare a composition - continuous composite material sample based on the material combination technology;

[0069] Module M2: Perform X - ray diffraction characterization on the target point to be characterized. Initially, the target point to be characterized is the corner point of the sample;

[0070] Module M3: Analyze the X - ray diffraction data in real - time to achieve real - time phase identification;

[0071] Module M4: Feed the phase identification result back into the cellular automaton algorithm, and use the cellular automaton algorithm to determine the next target point to be characterized; repeat steps S2 to S4 until all phase boundaries are determined.

[0072] Specifically, the module M3 adopts:

[0073] Using custom software from the "xrayutilities" library in Python, the original two-dimensional (2D) diffraction spectra obtained in XRD characterization are converted into one-dimensional intensity-2θ (2θ) spectra. First, the one-dimensional profile is subtracted by the diffraction profile corresponding to the amorphous silicon substrate measured in the same experiment, and then further processed using the rolling ball algorithm to obtain a curve with background interference removed. Then, Gaussian convolution filtering with a 1×7 discrete Gaussian function kernel is used to smooth the curve and remove high-frequency noise caused by statistical fluctuations. By using the built-in functions of scipy in Python, the peak positions (2θ) and other characteristics, including height and width, are extracted from the smoothed curve. Then, a better curve fitting method is used to locate the peaks that are very close and difficult to distinguish: at the position where two characteristic peaks almost overlap, the curve within a small range of 2θ is fitted with a combination of multiple Gaussian peaks based on the least squares optimization algorithm. According to the optimized fitting parameters, the positions, heights, and widths of the peaks are obtained. The peaks identified in the above two steps are merged and selected according to the thresholds of height and weight, and only prominent and sharp peaks are retained. Then, the PDF cards of the possible phases in this system are collected from the ICSD database, and the obtained peak position data are compared with the PDF cards one by one to determine the phase composition of the sample point and mark it with the corresponding label.

[0074] Specifically, the cellular automaton algorithm includes: when performing cellular automaton analysis, a grid needs to be created to divide the combined chip into multiple cells to establish the adjacent relationship between sample points. Since the distribution of any sample points can be divided into a triangular network using the Delaunay triangulation method, a cellular automaton algorithm based on triangular cells is designed. In the triangular grid cell, if two points belong to the same cell, they are considered adjacent. As Figure 2 shown, all the pentagrams in the figure are adjacent points of the dot.

[0075] The flow framework of this cellular automaton algorithm is as Figure 4 shown. The algorithm consists of two modes, each mode contains a set of evolution (iteration) rules, and then different evolution behaviors are generated respectively.

[0076] Mode Ⅰ: Search for potential phase boundaries based on the dichotomy method. Mode Ⅰ only checks the edge region of the phase map, that is, the connection region between the vertices of the phase diagram. When the explored nearest neighbor sample points in this region belong to different phase class labels, there must be a phase boundary between them. The dichotomy method is used to find a pair of adjacent points on the potential phase boundary, that is, each time their midpoint is explored until the two points are adjacent. When a pair of adjacent points on the potential phase boundary is found, switch to Mode Ⅱ; if it is checked that there are no longer potential boundaries, the process ends, and the data points with missing phase class labels are filled to complete the construction of the phase map.

[0077] Mode II: Construct the phase boundary based on the continuity of the phase boundary. Define two adjacent sample points with different phase class labels as the growth base points, which are actually part of the phase boundary. There must be 1-2 common neighboring sample points for these two sample points, which enables the iteration to continue. As Figure 3 shown, points A and B are two adjacent sample points with different phase class labels, and point C is their common neighboring point. Then there are three cases for C: a. C and A belong to the same phase class, then B and C form a new growth base point, and E becomes the new point to be characterized; b. C and B belong to the same phase class, then A and C form a new growth base point, and D becomes the new point to be characterized; c. C does not belong to the same phase class as A and B, then A and C, B and C respectively form new growth base points, and both D and E become new points to be characterized. Since the phase region is continuous, the phase boundary is also continuous. Therefore, a complete phase boundary can be obtained by continuously extending this iteration, similar to the growth of a vine. When the boundary grows to the boundary of the phase diagram or the boundary forms a closed loop, the iteration of the growth base points will stop. When the iteration of all growth base points stops, a continuous phase boundary growth is completed, and it switches to Mode I to continue searching for potential boundaries.

[0078] The present invention tests on a large number of isothermal section phase mapping data sets such as Cu-Cr-Co, Fe-Cr-Ni, Fe-Co-Ni, Ag-Ir-Pt-Pd-Ru, etc., and finds that this method greatly reduces the number of points to be characterized, and the accuracy rate reaches 100%. The characterized points under the guidance of the cellular automaton are almost 100% at the phase boundary. Therefore, the efficiency optimization is already very close to the theoretical limit. At the same time, this method is also outstanding in terms of generalization ability and can be directly extended to multi-component systems and various sample distribution conditions. Another advantage of this method is its simplicity - most of its calculations are logical operations, so the calculation cost is relatively low. The cellular automaton algorithm can immediately determine the next point to be characterized after obtaining the phase recognition result. In contrast, machine learning-based methods usually require a large amount of calculation time for decision-making and parameter adjustment.

[0079] Considering the universality of boundary continuity, the present invention can be combined with other data analysis modules. Therefore, in principle, it can be adapted to the autonomous construction of all property-composition mapping relationships. On the other hand, by combining this method with real-time data analysis, a real dynamic material development model can be realized.

[0080] Those skilled in the art know that, in addition to implementing the system and its various devices, modules, and units provided by the present invention in the form of pure computer-readable program code, it is entirely possible to achieve the same functions by logically programming the method steps so that the system and its various devices, modules, and units provided by the present invention are implemented in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers, etc. Therefore, the system and its various devices, modules, and units provided by the present invention can be regarded as a kind of hardware component, and the devices, modules, and units included therein for implementing various functions can also be regarded as the structures within the hardware component; it can also be considered that the devices, modules, and units for implementing various functions are both software modules for implementing the method and the structures within the hardware component.

[0081] The specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Without conflict, the embodiments of the present application and the features in the embodiments can be combined with each other arbitrarily.

Claims

1. An autonomous phase mapping construction method based on cellular automata, characterized in that Including: Step S1: Prepare a combined material sample with continuous composition based on the material combination technology; Step S2: Perform X-ray diffraction characterization on the characterization target points. Initially, the characterization target points are the corner points of the sample; Step S3: Perform real-time analysis on the X-ray diffraction data to achieve real-time phase identification; Step S4: Feed the phase identification result back into the cellular automaton algorithm, and use the cellular automaton algorithm to determine the next characterization target point; Repeat steps S2 to S4 until all phase boundaries are determined; The said step S3 adopts: Step S3.1: Convert the original two-dimensional diffraction spectrum obtained in the X-ray diffraction characterization into a one-dimensional intensity-2θ spectrum line; Step S3.2: Preprocess the one-dimensional intensity-2θ spectrum line to obtain a smoothed curve with background interference eliminated; Step S3.3: Use the built-in function of scipy in python to preliminarily lock the peak positions of the characteristic peaks from the smoothed curve; Step S3.4: Based on the preliminarily locked characteristic peaks, use the combination of multiple Gaussian peaks for fitting according to the least squares optimization algorithm to obtain the accurate peak positions, peak heights, and full widths at half maximum of the characteristic peaks; Step S3.5: Screen the characteristic peaks according to the preset threshold based on the peak heights and full widths at half maximum of the obtained characteristic peaks, and retain the prominent and sharp crystallization peaks; Step S3.6: Collect the PDF cards of the possible phases in the current system from the ICSD database, and compare the peak position data of the obtained crystallization peaks with the PDF cards one by one to determine the phase composition of the current sample point and mark it with the corresponding phase label; The said step S3.2 adopts: Step S3.2.1: The one-dimensional intensity-2θ spectrum line is smoothed by Gaussian convolution filtering with a 1x7 discrete Gaussian function kernel to remove the high-frequency noise caused by statistical fluctuations; Step S3.2.2: Further process using the rolling ball algorithm to obtain a curve with background interference eliminated.

2. The method for constructing an autonomous phase mapping based on a cellular automaton according to claim 1, wherein The said cellular automaton algorithm adopts: Step S4.1: Use the bisection method to find a pair of adjacent points on the potential phase boundary based on the identified phase labels; Step S4.2: Construct a phase boundary based on the pair of adjacent points on the obtained phase boundary and the principle of phase boundary continuity. Repeat steps S4.1 to S4.2 until all phase boundaries are found.

3. The method for constructing an autonomous phase mapping based on a cellular automaton according to claim 2, wherein The said step S4.1 adopts: When the explored nearest neighbor sample points in the phase mapping edge region belong to different phase labels, there is a phase boundary between them; Use the bisection method to find a pair of adjacent points on the potential phase boundary among them; When it is checked that there is no longer a potential phase boundary, the process ends.

4. The method for constructing an autonomous phase mapping based on a cellular automaton according to claim 2, wherein The said step S4.2 adopts: Define the sample points with different and adjacent phase labels at the phase boundary as the growth base points; Continuously extend the phase boundary based on the iterative rule of the growth base points until the boundary grows to the boundary of the phase diagram or the boundary forms a closed loop, and the iteration of the growth base points will stop; When the iteration of all growth base points stops, a continuous phase boundary growth is completed.

5. An autonomous phase mapping construction system based on cellular automata, characterized in that, Using the autonomous phase mapping construction method based on cellular automata described in any one of claims 1 to 4, including: Module M1: Prepare a combined material sample with continuous composition based on the material combination technology; Module M2: Perform X-ray diffraction characterization on the characterization target points. Initially, the characterization target points are the corner points of the sample; Module M3: Perform real-time analysis on the X-ray diffraction data to achieve real-time phase identification; Module M4: Feed the phase identification results back into the cellular automaton algorithm, and use the cellular automaton algorithm to determine the next characterization target point; Repeat steps S2 to S4 until all phase boundaries are determined.

6. The autonomous phase mapping construction system based on cellular automata according to claim 5, characterized in that The module M3 adopts: Module M3.1: Convert the original two-dimensional diffraction spectrum obtained in the X-ray diffraction characterization into a one-dimensional intensity-2θ spectrum line; Module M3.2: Preprocess the one-dimensional intensity-2θ spectrum line to obtain a smoothed curve with background interference eliminated; Module M3.3: Use the built-in function of scipy in python to initially lock the peak positions of the characteristic peaks from the smoothed curve; Module M3.4: Based on the initially locked characteristic peaks, use the combination of multiple Gaussian peaks for fitting based on the least squares optimization algorithm to obtain the accurate peak positions, peak heights, and full-width at half-maximum of the characteristic peaks; Module M3.5: Screen the characteristic peaks according to the preset threshold for the peak heights and full-width at half-maximum of the obtained characteristic peaks, and retain the prominent and sharp crystallization peaks; Module M3.6: Collect the PDF cards of the possible phases in the current system from the ICSD database, compare the peak position data of the obtained crystallization peaks with the PDF cards one by one, so as to determine the phase composition of the current sample point, and mark it with the corresponding phase label.

7. The autonomous phase mapping construction system based on cellular automata according to claim 6, wherein The module M3.2 adopts: Module M3.2.1: Use Gaussian convolution filtering with a 1x7 discrete Gaussian function kernel for the one-dimensional intensity-2θ spectrum line to smooth the curve and remove the high-frequency noise caused by statistical fluctuations; Module M3.2.2: Further process using the rolling ball algorithm to obtain a curve with background interference eliminated.

8. The autonomous phase mapping construction system based on cellular automata according to claim 6, characterized in that The cellular automaton algorithm adopts: Module M4.1: Use the bisection method to find a pair of adjacent points on the potential phase boundary based on the identified phase labels; Module M4.2: Based on the pair of adjacent points on the obtained phase boundary and the principle of phase boundary continuity, construct the phase boundary. Repeat steps S4.1 to S4.2 until all phase boundaries are found; The module M4.1 adopts: When the nearest neighbor sample points explored in the phase mapping edge region belong to different phase labels, there is a phase boundary between them; Use the bisection method to find a pair of adjacent points on the potential phase boundary among them; When it is checked that there are no longer potential phase boundaries, the process ends; The module M4.2 adopts: Define the sample points with different and adjacent phase labels at the phase boundary as the growth base points; Continuously extend the phase boundary based on the iterative rules of the growth base points until the boundary grows to the boundary of the phase diagram or the boundary forms a closed loop, and the iteration of the growth base points will stop; When the iteration of all growth base points stops, a continuous phase boundary growth is completed.

Images