Method and apparatus for xrd quantification of cement clinker

Abstract

This invention discloses a method and apparatus for quantitative correction of cement clinker using XRD, belonging to the field of mineral composition analysis of cement clinker. The method includes: acquiring preliminary Rietveld XRD analysis values, free calcium oxide chemical analysis values, loss on ignition (LOI), and petrographic analysis values ​​of multiple batches of cement clinker samples; constructing an input vector using the preliminary XRD analysis values, free calcium oxide chemical analysis values, and LOI, and constructing an output vector using the petrographic analysis values ​​as actual values, and training the constructed artificial neural network model; inputting the preliminary XRD analysis values, free calcium oxide chemical analysis values, and LOI of the cement clinker sample to be tested into the trained artificial neural network model for correction, and outputting the corrected mineral component content. Based on Rietveld quantitative analysis, this invention combines f-CaO and LOI parameters from chemical analysis, and performs secondary correction through an artificial neural network to achieve quantitative results closer to the true chemical composition, significantly improving the accuracy of quantitative analysis of cement clinker minerals.

Description

Technical Field

[0001] This invention relates to the field of mineral composition analysis of cement clinker, and in particular to a method and apparatus for quantitative correction of cement clinker by XRD. Background Technology

[0002] X-ray diffraction (XRD) analysis is widely used in the determination of phase composition in cement clinker. Among these methods, the Rietveld full-spectrum fitting method is internationally recognized as a relatively accurate quantitative method, which can, to some extent, address the errors caused by single-peak fitting. However, the Rietveld method still has certain limitations: it is highly dependent on a complete crystal structure model; it is not effective in handling amorphous phases, glassy phases, and trace components; for rapid on-site analysis, it has high computational complexity and data quality requirements; and in cases containing free lime and carbonate minerals with large fluctuations, systematic biases may still occur. Traditional testing methods can no longer achieve accurate results. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention provides a method and apparatus for quantitative correction of cement clinker XRD, which can significantly improve the accuracy of quantitative analysis of cement clinker minerals and provide an effective tool for production control and quality evaluation.

[0004] The technical solution provided by this invention is as follows:

[0005] A quantitative XRD correction method for cement clinker, the method comprising:

[0006] S1: Obtain preliminary Rietveld XRD analysis values, free calcium oxide chemical analysis values, loss on ignition and petrographic analysis values ​​for multiple batches of cement clinker samples;

[0007] S2: The input vector is constructed using preliminary XRD analysis values, free calcium oxide chemical analysis values, and loss on ignition, and the output vector is constructed using petrographic analysis values ​​as actual values. The constructed artificial neural network model is then trained.

[0008] S3: Input the preliminary XRD analysis value, free calcium oxide chemical analysis value and loss on ignition of the cement clinker sample to be tested into the trained artificial neural network model for correction, and output the corrected mineral component content.

[0009] Furthermore, the input vector is (XRD_C3S, XRD_C2S, XRD_C3A, XRD_C4AF, XRD_f-CaO, LOI, f-CaO, XRD_pergasse, XRD_calcite, XRD_dolomite).

[0010] Among them, XRD_C3S, XRD_C2S, XRD_C3A, XRD_C4AF, XRD_f-CaO, XRD_pericarpegite, XRD_calcite, and XRD_dolomite are the contents of tricalcium silicate, dicalcium silicate, tricalcium aluminate, tetracalcium aluminoferrite, free calcium oxide, periclase, calcite, and dolomite in cement clinker samples measured by the Rietveld method, respectively. LOI and f-CaO are the loss on ignition and chemical analysis values ​​of free calcium oxide, respectively.

[0011] Furthermore, the output vector is (True_C3S, True_C2S, True_C3A, True_C4AF, True_pericarpegite, True_calcite, True_dolomite).

[0012] Among them, True_C3S, True_C2S, True_C3A, True_C4AF, True_pericarpegite, True_calcite, and True_dolomite represent the actual contents of tricalcium silicate, dicalcium silicate, tricalcium aluminate, tetracalcium aluminoferrite, periclase, calcite, and dolomite after petrographic analysis correction, respectively.

[0013] Furthermore, the artificial neural network model is a three-layer feedforward artificial neural network structure, with an input layer containing 10 neurons, a hidden layer containing no fewer than 14 neurons, an activation function being the Sigmoid function, and an output layer containing 7 neurons.

[0014] A cement clinker XRD quantitative correction device, the device comprising:

[0015] The sample data acquisition module is used to acquire the preliminary Rietveld XRD analysis values, free calcium oxide chemical analysis values, loss on ignition and petrographic analysis values ​​of multiple batches of cement clinker samples;

[0016] The model training module is used to construct an input vector with preliminary XRD analysis values, free calcium oxide chemical analysis values, and loss on ignition, and to construct an output vector with petrographic analysis values ​​as actual values, in order to train the constructed artificial neural network model.

[0017] The calibration module is used to input the preliminary XRD analysis value, free calcium oxide chemical analysis value and loss on ignition of the cement clinker sample to be tested into the trained artificial neural network model for calibration, and output the corrected mineral component content.

[0018] Furthermore, the input vector is (XRD_C3S, XRD_C2S, XRD_C3A, XRD_C4AF, XRD_f-CaO, LOI, f-CaO, XRD_pergasse, XRD_calcite, XRD_dolomite).

[0019] Among them, XRD_C3S, XRD_C2S, XRD_C3A, XRD_C4AF, XRD_f-CaO, XRD_pericarpegite, XRD_calcite, and XRD_dolomite are the contents of tricalcium silicate, dicalcium silicate, tricalcium aluminate, tetracalcium aluminoferrite, free calcium oxide, periclase, calcite, and dolomite in cement clinker samples measured by the Rietveld method, respectively. LOI and f-CaO are the loss on ignition and chemical analysis values ​​of free calcium oxide, respectively.

[0020] Furthermore, the output vector is (True_C3S, True_C2S, True_C3A, True_C4AF, True_pericarpegite, True_calcite, True_dolomite).

[0021] Among them, True_C3S, True_C2S, True_C3A, True_C4AF, True_pericarpegite, True_calcite, and True_dolomite represent the actual contents of tricalcium silicate, dicalcium silicate, tricalcium aluminate, tetracalcium aluminoferrite, periclase, calcite, and dolomite after petrographic analysis correction, respectively.

[0022] Furthermore, the artificial neural network model is a three-layer feedforward artificial neural network structure, with an input layer containing 10 neurons, a hidden layer containing no fewer than 14 neurons, an activation function being the Sigmoid function, and an output layer containing 7 neurons.

[0023] The present invention has the following beneficial effects:

[0024] This invention aims to provide a quantitative correction method for XRD analysis of cement clinker based on the detection results of free calcium oxide and loss on ignition, and an artificial neural network model. This method, building upon Rietveld quantitative analysis, combines f-CaO and LOI parameters from chemical analysis with a secondary correction using an artificial neural network to achieve quantitative results that more closely approximate the true chemical composition. This invention can significantly improve the accuracy of quantitative analysis of minerals in cement clinker, providing an effective tool for production control and quality evaluation. Attached Figure Description

[0025] Figure 1 This is a flowchart of the XRD quantitative correction method for cement clinker of the present invention;

[0026] Figure 2 This is a schematic diagram of the cement clinker XRD quantitative correction device of the present invention. Detailed Implementation

[0027] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0028] Example 1:

[0029] This invention provides a quantitative correction method for XRD analysis of cement clinker, aiming to improve the accuracy and stability of quantitative analysis results using the Rietveld method of X-ray diffraction (XRD). Figure 1 As shown, the method includes:

[0030] S1: Obtain preliminary Rietveld XRD analysis values, free calcium oxide chemical analysis values ​​(f-CaO), loss on ignition (LOI), and petrographic analysis values ​​from multiple batches (e.g., no less than 300 batches) of cement clinker samples.

[0031] The preliminary XRD analysis values ​​by the Rietveld method include XRD_C3S, XRD_C2S, XRD_C3A, XRD_C4AF, XRD_f-CaO, XRD_pericarpegite, XRD_calcite, and XRD_dolomite, which represent the contents of tricalcium silicate, dicalcium silicate, tricalcium aluminate, tetracalcium aluminoferrite, free calcium oxide, periclase, calcite, and dolomite in cement clinker samples measured by the Rietveld method, respectively.

[0032] The chemical analysis values ​​of free calcium oxide (f-CaO) and loss on ignition (LOI) (LOI is derived from the decomposition calculations of magnesium carbonate and calcium carbonate) are important parameters representing the quality of clinker firing, reflecting the degree of mineral transformation and carbonate decomposition. These parameters are usually obtained through chemical analysis methods, which are more direct and reliable, and can compensate for the shortcomings of Rietveld fitting results in the detection of specific minerals.

[0033] The petrographic analysis values ​​are True_C3S, True_C2S, True_C3A, True_C4AF, True_pericarpegite, True_calcite, and True_dolomite, which are the actual contents of tricalcium silicate, dicalcium silicate, tricalcium aluminate, tetracalcium aluminoferrite, periclase, calcite, and dolomite after petrographic analysis correction, respectively.

[0034] S2: The input vector is constructed using preliminary XRD analysis values, free calcium oxide chemical analysis values, and loss on ignition. The output vector is constructed using petrographic analysis values ​​as actual values. The constructed artificial neural network model is then trained.

[0035] The input vector consists of 10 features: (XRD_C3S, XRD_C2S, XRD_C3A, XRD_C4AF, XRD_f-CaO, LOI, f-CaO, XRD_pericargillaceous, XRD_calcite, XRD_dolomite). The input data can be Z-score standardized as required.

[0036] The output vector includes 7 features: (True_C3S, True_C2S, True_C3A, True_C4AF, True_pericarpegite, True_calcite, True_dolomite). Note: f-CaO is a result measured by a traditional chemical method and is only used as an input feature in modeling. It is not used as a prediction target by the model and is used in the correction process but is not in the output.

[0037] Artificial neural networks (ANNs), as a nonlinear modeling technique, are capable of learning nonlinear relationships between complex variables. In this invention, ANNs do not replace Rietveld analysis, but rather serve as a secondary correction tool for its results: by using Rietveld results along with key process parameters such as f-CaO and LOI as inputs, ANNs can further reduce systematic errors and improve the stability and accuracy of quantitative results.

[0038] The artificial neural network model is a three-layer feedforward artificial neural network structure. The input layer includes 10 neurons, the hidden layer includes no less than 14 neurons, the activation function is the Sigmoid function, and the output layer includes 7 neurons, corresponding to the corrected mineral content.

[0039] When training the model, follow these rules:

[0040] Gradient descent and mean squared error (MSE) are used as the loss function.

[0041] The Adam optimizer is applied with an initial learning rate of 0.01.

[0042] The sample set is divided into training and test sets at 80% and 20% respectively.

[0043] An early stopping strategy and an L2 regularization term are introduced to avoid overfitting.

[0044] Five-fold cross-validation was used to evaluate the model's generalization ability.

[0045] S3: Input the preliminary XRD analysis value, free calcium oxide chemical analysis value and loss on ignition of the cement clinker sample to be tested into the trained artificial neural network model for correction, and output the corrected mineral component content.

[0046] For a new cement clinker sample to be tested, obtain its preliminary Rietveld XRD analysis results, f-CaO and LOI, input them into the trained neural network model, and output the corrected contents of C3S, C2S, C3A, C4AF, periclase, calcite and dolomite.

[0047] This invention aims to provide a quantitative correction method for XRD analysis of cement clinker based on the detection results of free calcium oxide and loss on ignition, and an artificial neural network model. This method, building upon Rietveld quantitative analysis, combines f-CaO and LOI parameters from chemical analysis with a secondary correction using an artificial neural network to achieve quantitative results that more closely approximate the true chemical composition. This invention can significantly improve the accuracy of quantitative analysis of minerals in cement clinker, providing an effective tool for production control and quality evaluation.

[0048] The invention is illustrated in detail below with several specific experimental examples.

[0049] Example 1:

[0050] 1. Input data is shown in the table below:

[0051]

[0052] 2. The output results are shown in the table below:

[0053]

[0054] Example 2:

[0055] 1. Input data is shown in the table below:

[0056]

[0057] 2. The output results are shown in the table below:

[0058]

[0059] Example 3:

[0060] 1. Input data is shown in the table below:

[0061]

[0062] 2. The output results are shown in the table below:

[0063]

[0064] Example 4:

[0065] 1. Input data is shown in the table below:

[0066]

[0067] 2. The output results are shown in the table below:

[0068]

[0069] As can be seen from the above examples, all four experimental examples show that the ANN correction results are highly consistent with the true values ​​of petrographic analysis, with an average error of less than ±0.3%, proving that the method of the present invention has stability and reliability in different batches and under actual working conditions.

[0070] Example 2:

[0071] This invention provides a quantitative correction device for XRD of cement clinker, such as... Figure 2 As shown, the device includes:

[0072] Sample data acquisition module 1 is used to acquire preliminary Rietveld XRD analysis values, free calcium oxide chemical analysis values, loss on ignition and petrographic analysis values ​​of multiple batches of cement clinker samples.

[0073] Model training module 2 is used to construct an input vector using preliminary XRD analysis values, free calcium oxide chemical analysis values, and loss on ignition, and to construct an output vector using petrographic analysis values ​​as actual values, thereby training the constructed artificial neural network model.

[0074] The calibration module 3 is used to input the preliminary XRD analysis value, free calcium oxide chemical analysis value and loss on ignition of the cement clinker sample to be tested into the trained artificial neural network model for calibration, and output the corrected mineral component content.

[0075] The input vector is (XRD_C3S, XRD_C2S, XRD_C3A, XRD_C4AF, XRD_f-CaO, LOI, f-CaO, XRD_pergasse, XRD_calcite, XRD_dolomite).

[0076] Among them, XRD_C3S, XRD_C2S, XRD_C3A, XRD_C4AF, XRD_f-CaO, XRD_pericarpegite, XRD_calcite, and XRD_dolomite are the contents of tricalcium silicate, dicalcium silicate, tricalcium aluminate, tetracalcium aluminoferrite, free calcium oxide, periclase, calcite, and dolomite in cement clinker samples measured by the Rietveld method, respectively. LOI and f-CaO are the loss on ignition and chemical analysis values ​​of free calcium oxide, respectively.

[0077] The output vector is (True_C3S, True_C2S, True_C3A, True_C4AF, True_Palmite, True_Calcite, True_Dolomite).

[0078] Among them, True_C3S, True_C2S, True_C3A, True_C4AF, True_pericarpegite, True_calcite, and True_dolomite represent the actual contents of tricalcium silicate, dicalcium silicate, tricalcium aluminate, tetracalcium aluminoferrite, periclase, calcite, and dolomite after petrographic analysis correction, respectively.

[0079] The artificial neural network model is a three-layer feedforward artificial neural network structure, with an input layer containing 10 neurons, a hidden layer containing no fewer than 14 neurons, an activation function of the sigmoid function, and an output layer containing 7 neurons.

[0080] The apparatus provided in this embodiment of the invention has the same implementation principle and technical effects as the aforementioned method embodiment. For the sake of brevity, any parts not mentioned in the apparatus embodiment can be referred to the corresponding content in the aforementioned method embodiment 1. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the apparatus and units described above can all be referred to the corresponding processes in the above method embodiments, and will not be repeated here.

[0081] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit them. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the scope of the technology disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the scope of protection of the present invention.

Claims

1. A method of cement clinker XRD quantitative calibration, characterized in that, The method includes: S1: Obtain preliminary Rietveld XRD analysis values, free calcium oxide chemical analysis values, loss on ignition and petrographic analysis values ​​for multiple batches of cement clinker samples; S2: The input vector is constructed using preliminary XRD analysis values, free calcium oxide chemical analysis values, and loss on ignition, and the output vector is constructed using petrographic analysis values ​​as actual values. The constructed artificial neural network model is then trained. S3: Input the preliminary XRD analysis value, free calcium oxide chemical analysis value and loss on ignition of the cement clinker sample to be tested into the trained artificial neural network model for correction, and output the corrected mineral component content; The input vector is (XRD_C3S, XRD_C2S, XRD_C3A, XRD_C4AF, XRD_f-CaO, LOI, f-CaO, XRD_pergasse, XRD_calcite, XRD_dolomite). Among them, XRD_C3S, XRD_C2S, XRD_C3A, XRD_C4AF, XRD_f-CaO, XRD_pericarpegite, XRD_calcite and XRD_dolomite are the contents of tricalcium silicate, dicalcium silicate, tricalcium aluminate, tetracalcium aluminoferrite, free calcium oxide, periclase, calcite and dolomite in cement clinker samples measured by the Rietveld method, respectively; LOI and f-CaO are the loss on ignition and chemical analysis values ​​of free calcium oxide, respectively. The output vector is (True_C3S, True_C2S, True_C3A, True_C4AF, True_pericarpium, True_calcite, True_dolomite); Among them, True_C3S, True_C2S, True_C3A, True_C4AF, True_pericarpegite, True_calcite, and True_dolomite represent the actual contents of tricalcium silicate, dicalcium silicate, tricalcium aluminate, tetracalcium aluminoferrite, periclase, calcite, and dolomite after petrographic analysis correction, respectively.

2. The method of cement clinker XRD quantitative correction according to claim 1, characterized in that, The artificial neural network model is a three-layer feedforward artificial neural network structure, with an input layer containing 10 neurons, a hidden layer containing no fewer than 14 neurons, an activation function of the sigmoid function, and an output layer containing 7 neurons.

3. A quantitative calibration device for XRD of cement clinker, characterized in that, The device includes: The sample data acquisition module is used to acquire preliminary Rietveld XRD analysis values, free calcium oxide chemical analysis values, loss on ignition and petrographic analysis values ​​of multiple batches of cement clinker samples; The model training module is used to construct an input vector with preliminary XRD analysis values, free calcium oxide chemical analysis values, and loss on ignition, and to construct an output vector with petrographic analysis values ​​as actual values, in order to train the constructed artificial neural network model. The calibration module is used to input the preliminary XRD analysis value, free calcium oxide chemical analysis value and loss on ignition of the cement clinker sample to be tested into the trained artificial neural network model for calibration, and output the corrected mineral component content. The input vector is (XRD_C3S, XRD_C2S, XRD_C3A, XRD_C4AF, XRD_f-CaO, LOI, f-CaO, XRD_pergasse, XRD_calcite, XRD_dolomite). Among them, XRD_C3S, XRD_C2S, XRD_C3A, XRD_C4AF, XRD_f-CaO, XRD_pericarpegite, XRD_calcite and XRD_dolomite are the contents of tricalcium silicate, dicalcium silicate, tricalcium aluminate, tetracalcium aluminoferrite, free calcium oxide, periclase, calcite and dolomite in cement clinker samples measured by the Rietveld method, respectively; LOI and f-CaO are the loss on ignition and chemical analysis values ​​of free calcium oxide, respectively. The output vector is (True_C3S, True_C2S, True_C3A, True_C4AF, True_pericarpium, True_calcite, True_dolomite); Among them, True_C3S, True_C2S, True_C3A, True_C4AF, True_pericarpegite, True_calcite, and True_dolomite represent the actual contents of tricalcium silicate, dicalcium silicate, tricalcium aluminate, tetracalcium aluminoferrite, periclase, calcite, and dolomite after petrographic analysis correction, respectively.

4. The cement clinker XRD quantitative correction device according to claim 3, characterized in that, The artificial neural network model is a three-layer feedforward artificial neural network structure, with an input layer containing 10 neurons, a hidden layer containing no fewer than 14 neurons, an activation function of the sigmoid function, and an output layer containing 7 neurons.

Images