Wheat flour processing precision detection method and system

Abstract

The invention discloses a wheat flour processing precision detection method and system, and relates to the field of material analysis by measuring physical properties. The method comprises the following steps: carrying out near-infrared scanning on a wheat flour sample to obtain spectral data, and carrying out smooth first-order derivative + vector normalization + detrending method processing; abnormal data is eliminated through a Z-Score method, and the data accuracy is ensured; and comparing the processed spectrum data with the established processing precision prediction model to obtain the bran content of the wheat flour, and further judging the processing precision. The method has the characteristics of automation, rapidness, no damage and the like, the detection efficiency is remarkably improved, and manual intervention is reduced.

Description

Technical Field

[0001] The present invention relates to the field of analyzing materials by measuring physical properties, and particularly to a method and system for detecting the processing precision of wheat flour. Background Art

[0002] The purpose of wheat flour milling is to process wheat that has been cleaned and moisture-conditioned through mechanical means into wheat flour suitable for different needs, while separating by-products. Milling is the most complex and important section in wheat processing. In the actual production process, the detection method of wheat flour processing precision mainly depends on the amount of bran, and the ash content can also be used as an indirect evaluation method for judgment.

[0003] In Chinese Patent Publication No.: CN118730804A, a method for detecting the processing precision of wheat flour is disclosed. The present invention conducts thermogravimetric analysis on wheat flour to obtain a pyrolysis characteristic curve, and based on the pyrolysis kinetic equation and first-order derivative, establishes the correlation between the characteristic values of the pyrolysis curve (pyrolysis residual mass fraction, pyrolysis activation energy at 238 - 338 °C, and pyrolysis peak area at 290 - 320 °C) and the key indicators of wheat flour processing precision (ash content, starch content). Then, combined with principal component analysis, three principal component indicators and the weight values of their characteristic values are obtained, and a comprehensive evaluation function formula for wheat flour processing precision is constructed. However, this method has high requirements for sample quality, is not simple enough, and the detection time is too long. Summary of the Invention

[0004] The present invention aims to solve the problems that the current detection methods for wheat flour processing precision are cumbersome, time-consuming, inaccurate, and easily affected by human factors in the detection process.

[0005] A method for detecting the processing precision of wheat flour includes: Performing near-infrared scanning on a wheat flour sample to obtain near-infrared spectral data of the wheat flour sample; Performing local polynomial fitting on the near-infrared spectral data of the wheat flour sample, calculating the first derivative, normalizing the first derivative, and performing polynomial fitting on the normalized data to remove the baseline drift in the spectrum to obtain the final spectral data; Comparing the processed near-infrared spectral image of the wheat flour sample with a model to obtain the processing precision of the wheat flour, and the model is a partial least squares regression model of near-infrared spectra of wheat and the bran content in wheat flour.

[0006] Preferably, when performing near-infrared scanning on the wheat flour sample, scanning n times and using the Z-Score method to determine whether the spectral data of each scan is an outlier; after excluding the abnormal results, taking the average value of the remaining scan results as the final spectral data.

[0007] Preferably, if near-infrared scanning is performed and more than half of the scanned data is judged as abnormal data, the spectral data of the sample will be marked as "abnormal", and the operator will be prompted to perform scanning again.

[0008] Preferably, it further includes a sample weighing step. The sample is weighed before scanning. If the weight is less than 2 g, the wheat flour is sampled again.

[0009] Preferably, the method for establishing the model includes: Collect wheat flour samples with known bran contents; Perform near-infrared scanning on each sample to obtain the spectral data of each sample; Use a variety of methods to preprocess the spectral data; Use partial least squares regression to establish multiple wheat flour processing precision prediction models for the spectral data processed by different preprocessing methods; Bring the prediction set into the multiple wheat flour processing precision prediction models and calculate the model accuracy Select the model with the highest accuracy as the final wheat flour processing precision prediction model.

[0010] Preferably, the specific steps for calculating the model accuracy are as follows: Calculate the calibration correlation coefficient R C and the calibration determination coefficient R 2 C , the root mean square error of calibration RMSEC, the prediction correlation coefficient R P , the prediction determination coefficient R 2 P , the root mean square error of prediction RMSEP, and the ratio RPD of the standard deviation of the calibration set to the standard deviation of the prediction set; perform normalization processing on these parameters; take the average of the data after normalization processing as the accuracy index of the model.

[0011] Preferably, the characteristic spectral absorption peak of the final wheat flour processing precision prediction model is 570 - 1100 nm.

[0012] A wheat flour processing precision detection system that executes the wheat flour processing precision detection method according to any one of claims 1 to 6, includes: A spectral acquisition module that uses a near-infrared spectrometer to acquire the spectral data of wheat flour samples and is connected to the data preprocessing module and the prediction model module through USB or Ethernet; A data preprocessing module that preprocesses the spectral data and eliminates abnormal data. This module is connected to the spectral acquisition module and the prediction model module; A prediction model module that uses the PLSR model to predict the processing precision according to the preprocessed spectral data. This module exchanges data with the data preprocessing module through the API.

[0013] Preferably, the wavelength range of the near-infrared spectrometer is 570 - 1100 nm, the spectral repeatability ≤ 0.02 nm, and the spectral accuracy ≤ 0.05 nm.

[0014] Preferably, the wheat flour processing precision detection system further includes a high-precision electronic scale, which is placed under the sample tray.

[0015] Through the above technology, the present invention realizes efficient and accurate detection of wheat flour processing precision. The near-infrared spectrometer is used to collect the spectra of samples, and combined with the automated data preprocessing and model comparison process, the entire detection process does not require manual intervention. The system can automatically identify and eliminate abnormal data to ensure the reliability of the final result. This automated operation not only reduces human errors but also improves the detection efficiency, and is especially suitable for real-time quality monitoring in large-scale production. Description of the Drawings

[0016] Figure 1 is the flow chart of the wheat flour processing precision detection method.

[0017] Figure 2 is the original near-infrared spectra of wheat flour with different processing precisions.

[0018] Figure 3 is the SG1 of wheat flour with different processing precisions st +SNV+Detrending preprocessed near-infrared spectra.

[0019] Figure 4 is the SG1 st +SNV+Detrending preprocessed comparison chart of true values and predicted values of calibration set and prediction set. Detailed Embodiments

[0020] The following are the detailed embodiments of the present invention. The technical solutions of the present invention will be described in detail in combination with the drawings.

[0021] Example 1 Higher bran content: When the bran content of wheat flour is higher, it usually means that the wheat has not undergone sufficient fine screening or milling process during processing, resulting in more bran, germ, etc. not being effectively separated, and finally resulting in rough and larger-grained powder. In this case, the processed wheat flour has a lower precision, uneven particles, and contains more crude fiber parts.

[0022] Lower bran content: When the bran content of wheat flour is lower, it usually indicates that the wheat has undergone a more refined screening, milling, and separation process during processing, removing more bran and rough parts. At this time, the obtained wheat flour has a higher precision, finer particles, and is generally considered to be a higher-quality powder.

[0023] The processing precision of wheat flour is directly affected by the bran content. A low bran content usually means a higher processing precision, and conversely, a high bran content often means a lower processing precision. Therefore, by detecting the bran content of wheat flour samples, the processing precision can be effectively deduced. This relationship provides a scientific basis for using near-infrared spectroscopy technology to detect the processing precision of wheat flour in this invention.

[0024] As Figure 1 shown: In this invention, the spectral diagram of the sample is collected by a near-infrared spectrometer, and the obtained spectral diagram contains the reflected light intensity data within a certain wavelength range. The data of this spectral diagram is used to compare with a pre-established prediction model, so as to deduce the processing precision (bran content) of wheat flour. The specific process is as follows: Before detecting the processing precision of wheat flour, it is first necessary to properly prepare the sample. The preparation of wheat flour samples is crucial and directly affects the accuracy and reliability of the detection results.

[0025] Collect multiple batches of wheat flour samples from the wheat flour production line to ensure that the sample sources are extensive and representative. Each sample should be uniform enough for accurate spectral analysis.

[0026] The weight of each sample should ensure at least 2 grams to ensure that the spectrometer can comprehensively scan the sample and obtain accurate results. The samples should be consistent among different batches, and any contamination or external interference should be avoided during the collection process.

[0027] Sample storage: The samples need to be stored in a dry and clean environment. Ensure that the temperature and humidity during storage are stable to prevent moisture changes or contamination of the samples. Each sample should be stored in a sealed bag or moisture-proof container to prevent changes in the composition of the samples due to environmental changes.

[0028] Stir each wheat flour sample evenly to ensure uniform distribution of the components inside the wheat flour. The samples should be loaded in a dust-free environment to avoid interference from impurities or the external environment to the data.

[0029] The wheat flour samples should be evenly loaded into the sample cell of the spectrometer. Ensure that the scanning area of the spectrometer is completely covered during the loading of each sample to ensure that the collected data represents the characteristics of the entire sample. The surface of the sample should be flat to avoid unevenness or caking on the sample surface, which may affect the stability and accuracy of the spectral data.

[0030] Spectral diagram collection and preprocessing The reflection spectral data of wheat flour samples are collected by a near-infrared spectrometer. Generally, the wavelength range for collection is between 570 - 1100 nm, and this range includes the absorption peaks of the main components (such as starch, protein, fat, fiber, etc.) in the wheat flour samples. The reflection intensity at each wavelength reflects the content of different chemical components in the sample. The wheat spectral diagram is as shown in Figure 2 as follows.

[0031] To eliminate possible outliers, the present invention uses the Z-Score method in multiple scans of each sample to determine whether the spectral data of each scan is an outlier. The specific steps are as follows: Z-Score is a commonly used statistical method for standardizing data. By calculating the Z-Score of each scanned data point, it is determined whether it is an outlier.

[0032] z = (x - μ) / σ Here, x is the original score, z is the transformed z-score, μ is the mean score of the overall sample space, and σ is the standard deviation of the overall sample space.

[0033] Generally, data with a Z-Score exceeding ±3 is considered an outlier (which means the data point is more than 3 times the standard deviation away from the mean). If the Z-Score of a certain scan is greater than 3 or less than -3, the data is considered abnormal data.

[0034] If there is abnormal data, the data will be excluded, and the average of the remaining three scans is taken as the final spectral data.

[0035] For each wheat flour sample, the system automatically detects the Z-Score of the four scans. If the Z-Score of a certain scan exceeds the threshold (±3), the spectral data of this scan will be excluded. The remaining three data will be used to calculate the final average spectral value.

[0036] This method effectively eliminates abnormal data caused by instrument instability, sample accidental interference, or external factors, ensuring that the final data is more stable and accurate.

[0037] If all four scanned data are judged as abnormal data, the spectral data of this sample will be marked as "abnormal", and the operator is prompted to rescan.

[0038] If only some of the scanned data are excluded, the system will automatically use the mean of the remaining data for predicting the processing accuracy.

[0039] To further improve the quality of spectral data, the final data after abnormal data rejection will enter the subsequent preprocessing and prediction steps. The system will monitor each scan result in real time during the collection of each wheat flour sample to ensure that each sample can provide the most accurate spectral data.

[0040] Before comparison, it is necessary to preprocess the spectral data. The present invention uses the method of smoothing first derivative + vector normalization + detrending (SG1 st +SNV+Detrending) to preprocess the spectral data. The combination steps of this method can be carried out according to the following process: 1.SG1 st (Savitzky-Golay first derivative): First, use the Savitzky-Golay filter to smooth the original data, which helps to reduce noise. Then, calculate the first derivative of the data, aiming to capture the rate of change of the data (i.e., the slope), so as to highlight the dynamic changes of the signal. This step is usually used to smooth fluctuations and reveal the trend changes of the data; 2.SNV (standard normal variate): After SG1 st processing, perform standard normal transformation. Specifically, standardize the data by subtracting the mean of each data point and dividing by the standard deviation. The purpose of this step is to eliminate the influence caused by dimensionality, amplitude difference or systematic deviation between different data sets, so that the data is comparable; 3.Detrending: After applying SG1 st and SNV processing, the data may still contain some long-term trends (such as low-frequency drift). To better reveal the periodic changes or short-term fluctuations of the signal, the detrending step will remove these long-term trends, usually using methods such as difference or regression analysis to achieve.

[0041] The purpose of the preprocessing step is to make the collected spectral data more comparable and consistent, facilitating accurate comparison with the model. The preprocessed image is as Figure 3 shown.

[0042] Spectral map feature extraction and model comparison: After preprocessing, the reflected light intensity data of each wavelength in the spectral map will be used to compare with the established processing accuracy prediction model. The reflected data of each wavelength in the spectral map can be regarded as a feature, and these features will be matched with the features already learned in the model. The specific process includes: A PLSR (Partial Least Squares Regression) model is established through a large number of wheat flour samples with known bran contents. This model is trained based on the relationship between the spectrogram and the known wheat flour processing precision (such as bran content). During the training process, the model will learn which spectral features (reflection intensities at specific wavelengths) are correlated with the processing precision (such as bran content).

[0043] During the model training process, the system extracts characteristic parameters from the spectrogram through the Partial Least Squares (PLSR) technique. These parameters can effectively reflect the material composition of wheat flour. The model will automatically determine which wavelength data is most relevant to the processing precision and assign appropriate weights to these relevant wavelengths.

[0044] Different chemical components will have different absorption peaks at different wavelengths. For example, components such as moisture, protein, and fat may have strong absorption in different wavelength regions. These absorption peaks can be identified in the spectrogram and compared with the known data in the model.

[0045] By extracting the features of the spectrogram (such as the wavelength position of the reflected light intensity, absorption peaks, etc.) and comparing them with the features learned in the training model, the processing precision indicators such as the bran content of the sample can be deduced.

[0046] The preprocessed spectral data of the sample is input into the PLSR model.

[0047] The model determines the processing precision of the sample by calculating the similarity between the sample spectral data and the spectral features in the training data.

[0048] Based on the comparison result, the model will predict the processing precision (such as bran content) of the wheat flour sample, and this predicted value is the processing precision indicator of the sample.

[0049] For the output of the processing precision prediction result, after the spectrogram of the sample is matched with the spectrogram in the calibration set through comparison with the model, the model will give a predicted value of the processing precision of the wheat flour, usually expressed as a percentage of the bran content. This predicted value provides a basis for the quality control of each sample.

[0050] In the present invention, the comparison between the sample spectrogram and the model is achieved by extracting the characteristic parameters of the sample spectrogram and comparing them with the spectral features already learned in the model. Through the Partial Least Squares Regression (PLSR) model, combined with the correlation between the spectral data of the sample and the known processing precision data, the processing precision (bran content) of the wheat flour can be deduced. This process relies on the precise processing of the spectrogram and the effective establishment of the model, thus achieving efficient and accurate processing precision prediction.

[0051] Example 2: To achieve efficient detection of the processing precision of wheat flour, the present invention establishes a wheat flour processing precision prediction model through the following implementation steps. This model can predict the bran content of wheat flour samples based on near-infrared spectroscopy data, thereby judging its processing precision.

[0052] In this embodiment, more than three thousand wheat flour samples are first collected and prepared: The present invention collects and prepares three thousand six hundred wheat flour samples with known and different processing precisions from well-known domestic flour mills such as Luhua, Wudeli, COFCO, and Chen Keming. Each sample weighs more than 5 kg. Sampling is carried out according to the quartering method, divided into 2 parts, respectively packed into self-sealing bags, and stored in a dry environment for later use.

[0053] The selected samples include wheat flour with low bran content and high bran content, ensuring that the model can cover wheat flour samples with different processing precisions.

[0054] The bran content of each sample is determined by traditional chemical analysis methods (such as the ash method) and recorded as the true value for use as the target value for subsequent model training.

[0055] Each wheat flour sample is scanned by a near-infrared spectrometer to obtain the corresponding near-infrared spectroscopy data. The wavelength range for spectral acquisition is set to 570 - 1100 nm to cover the main spectral absorption regions in wheat flour samples.

[0056] Each sample is scanned four times independently to ensure data stability.

[0057] The scanning resolution is 0.5 nm. To improve the quality and reliability of the data, the spectrometer is calibrated each time during scanning to ensure stable equipment performance during the measurement process.

[0058] Construction of the model training dataset The preprocessed spectroscopy data is combined with the known bran content (determined by traditional chemical analysis methods) to construct a training dataset. This dataset contains the spectral characteristics of each sample and its corresponding processing precision (bran content) for establishing a wheat flour processing precision prediction model.

[0059] To ensure the generalization ability and prediction accuracy of the model, the K-S (Kennard-Selection) method is used to divide all sample data into a calibration set and a prediction set. Generally, the calibration set accounts for 70% - 80% of the total dataset, and the remaining part is used to verify the prediction accuracy of the model.

[0060] To eliminate the interference caused by physical differences between samples, equipment noise, or environmental changes, the collected spectroscopy data must be preprocessed. The preprocessing steps include smoothing first derivative, vector normalization (SNV), and detrending to improve the quality of the spectroscopy data.

[0061] Using the Savitzky-Golay first derivative (SG1 st ), standard normal variate (SNV), detrending, standard normal variate + Savitzky-Golay first derivative (SNV+SG1 st ), detrending + Savitzky-Golay first derivative (Detrending+SG1 st ), standard normal variate + detrending (SNV+Detrending), Savitzky-Golay first derivative + standard normal variate + detrending (SG1 st +SNV+Detrending) to explore and optimize the near-infrared spectroscopy pretreatment method applicable to the detection of wheat flour processing accuracy. After pretreatment, the spectral data is more comparable, ensuring more accurate prediction results can be obtained in subsequent model training.

[0062] The present invention uses the partial least squares regression (PLSR) method to establish and train the model. PLSR is a powerful statistical method that can handle high-dimensional data and find the relationship between independent variables (spectral data) and dependent variables (processing accuracy). Its advantages include: Handling high-dimensional data: PLSR can effectively handle the situation where the number of variables is greater than the number of samples (i.e., high-dimensional data). Dimensionality reduction ability: PLSR reduces the dimensionality of the data by extracting latent principal components, thereby improving the computational efficiency and solving the problem of multicollinearity. It helps to extract the most meaningful information in the data while ignoring the noise. Eliminating multicollinearity: In traditional regression models, if there is a strong correlation (multicollinearity) between independent variables, it may lead to unstable regression coefficients. PLSR avoids the problem of multicollinearity by constructing new components (principal components), enhancing the stability and reliability of the model.

[0064] Using the calibration set wheat flour sample data, a partial least squares (PLSR) model is established, and the prediction set wheat flour sample data is substituted to obtain the calibration correlation coefficient R C , calibration determination coefficient R 2 C , calibration root mean square error RMSEC, prediction correlation coefficient R P , prediction determination coefficient R 2 P , prediction root mean square error RMSEP, and the ratio RPD of the prediction set standard deviation to the prediction set standard deviation, calibration correlation coefficient R C , calibration determination coefficient R 2 C , prediction correlation coefficient R P and prediction determination coefficient R 2 PThe larger it is, the better the prediction effect, the smaller the calibration root mean square error RMSEC and the prediction root mean square error RMSEP, the better the model effect. When the ratio of the calibration set standard deviation to the prediction set standard deviation RPD≥3, it indicates that the model is reliable. Data normalization processing: For R C 、R 2 C 、RMSEC、R P 、R 2 P 、RMSEP, range normalization is performed respectively, and the formula is: Among them, the extreme values of each index are calculated independently to ensure that all parameters are normalized to the [0,1] interval. Finally, a comprehensive evaluation index is constructed: The seven normalized parameters are arithmetically averaged, and the higher the final score, the better the overall performance of the model. This method combines multi-dimensional indicators, considering both the model fitting ability (calibration indicators) and the generalization performance (prediction indicators), and introduces RPD to strengthen the sensitivity to data distribution differences. The normalization processing effectively eliminates the scale differences of indicators with different dimensions, forming a unified and comparable comprehensive evaluation system. In this way, the optimal near-infrared prediction model is determined.

[0065] After determination, for the present invention, the model obtained by using SG1 st +SNV+Detrending preprocessing has the best effect.

[0066] The model effect is as Figure 4 shown.

[0067] After preliminary training, the prediction set is used to test the model. The prediction accuracy of the model is evaluated by comparing the error between the predicted value and the actual value. If the performance of the model does not meet the requirements (for example, the prediction error is large), it can be optimized in the following ways: Adjust the number or range of calibration set samples; Change the parameter settings in the PLSR model; Select more principal components for model training; Through optimization, a wheat flour processing precision prediction model with high accuracy and strong generalization ability is finally obtained.

[0068] After the model is trained and passes the prediction set test, the model can be applied to predict the processing precision of unknown samples. By inputting the near-infrared spectrum data of wheat flour samples, the model can quickly and accurately predict the bran content of the sample, providing a reliable basis for the quality control of wheat flour.

[0069] For a new wheat flour sample, the system preprocesses its spectral data and then inputs it into the PLSR model for prediction. The model calculates the bran content of the sample and provides relevant information on the processing precision.

[0070] This embodiment demonstrates how to efficiently predict the processing precision of wheat flour by collecting spectral data of wheat flour samples, preprocessing them, establishing a PLSR prediction model, and optimizing the model. Through these steps, the present invention can achieve efficient and accurate detection of the processing precision of wheat flour, solve problems such as high cost, long cycle, and human error in traditional detection methods, and has broad application prospects.

[0071] The above embodiments are only for the detailed description of the present invention and are not used to limit the present invention. It should be noted that appropriate modifications and decorations made by those skilled in the art within the scope of the technology and concept of the present invention should also fall within the protection scope of the claims of the present invention.

Claims

1. A method for detecting wheat flour processing accuracy, characterized in that: include: Perform near infrared scanning on the wheat flour sample to obtain near infrared spectrum data of the wheat flour sample; After performing local polynomial fitting on the near-infrared spectrum data of the wheat flour sample, the first-order derivative is calculated, the first-order derivative is normalized, and the normalized data is polynomially fitted to remove the baseline drift in the spectrum to obtain final spectrum data; The near infrared spectrum image of the processed wheat flour sample is compared with the model to obtain the wheat flour processing accuracy. The model is a partial least squares regression model of the wheat near infrared spectrum and the bran content in the wheat flour.

2. The wheat flour processing accuracy detection method according to claim 1, characterized in that: The wheat flour sample is subjected to near-infrared scanning, scanned n times, and a Z-Score method is used to determine whether the spectral data of each scan is an abnormal value; after the abnormal results are removed, the average value of the remaining scan results is used as the final spectral data.

3. The wheat flour processing accuracy detection method according to claim 1 or 2, characterized in that: In the near-infrared scanning, if more than half of the scanned data are judged to be abnormal data, the spectral data of the sample will be marked as "abnormal" and the operator will be prompted to rescan.

4. A wheat flour processing accuracy detection method according to claim 1, characterized in that: The method also includes a sample weighing step, in which the sample is weighed before scanning, and if the weight is less than 2 g, the wheat flour is resampled.

5. The wheat flour processing accuracy detection method according to claim 1, characterized in that: The method for establishing the model includes: collecting wheat flour samples with known bran content; Perform near infrared scanning on each sample to obtain spectral data of each sample; Various methods are used to pre-process the spectral data; The partial least squares regression method was used to establish multiple wheat flour processing accuracy prediction models for the spectral data after different pretreatment methods. The prediction set is brought into the multiple wheat flour processing accuracy prediction models to calculate the model accuracy The model with the highest accuracy was selected as the final wheat flour processing accuracy prediction model.

6. The wheat flour processing accuracy detection method according to claim 5, wherein the specific steps of calculating the accuracy of the model are as follows: calculating the correction correlation coefficient R C , Corrected determination coefficient R 2 C , corrected root mean square error RMSEC, prediction correlation coefficient R P , prediction coefficient R 2 P , prediction root mean square error RMSEP, and the ratio of the standard deviation of the correction set to the standard deviation of the prediction set RPD; normalize these parameters; take the average of the normalized data as the accuracy indicator of the model.

7. The wheat flour processing accuracy detection method according to claim 5, characterized in that: The characteristic spectrum absorption peak of the final wheat flour processing accuracy prediction model is 570-1100nm.

8. A wheat flour processing accuracy detection system, which executes the wheat flour processing accuracy detection method according to any one of claims 1 to 7, characterized in that: include: The spectrum acquisition module uses a near-infrared spectrometer to collect spectrum data of wheat flour samples, and connects the data preprocessing module and the prediction model module via USB or Ethernet; A data preprocessing module preprocesses the spectral data and removes abnormal data. This module is connected with the spectral acquisition module and the prediction model module; The prediction model module uses the PLSR model to predict the processing accuracy based on the preprocessed spectral data. This module exchanges data with the data preprocessing module through the API.

9. The wheat flour processing accuracy detection system according to claim 8, characterized in that: The wavelength range of the near-infrared spectrometer is 570-1100nm, the spectrum repeatability is ≤0.02nm, and the spectrum accuracy is ≤0.05nm.

10. The wheat flour processing accuracy detection system according to claim 8, characterized in that: Also included is a high-precision electronic scale that is placed beneath the sample tray.

Images