Data processing device and data processing method for chromatograph, and chromatograph

Abstract

To easily obtain a peak characteristic point in a short time without being affected by noise.SOLUTION: A data processing device for chromatograph comprises: a partial regression curve calculation unit that calculates a partial regression curve for each of a plurality of data groups in which a predetermined number of measured time series data are grouped; a partial feature point acquisition unit that acquires a partial feature point for each of the calculated partial regression curves; and an arithmetic processing unit 163 that functions as a peak feature point acquisition unit that acquires peak feature points in the measured time series data based on the partial feature points.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a chromatography technique such as a liquid chromatograph, and more particularly to a data processing device for a chromatograph, a data processing method, and a chromatograph. [Background technology]

[0002] In a chromatograph, the types and amounts of components contained in an analytical sample are determined from waveform data, with the horizontal axis representing time and the vertical axis representing signal intensity. At that time, peak characteristic points, such as the start point where the signal intensity rises and the end point where it falls, are detected based on the data detected by the device, and waveform processing is performed.

[0003] As a technology for performing the above-mentioned waveform processing, for example, there is known a technology that aims to "provide a chromatographic device that can easily perform curve fitting using the nonlinear least squares method even for chromatograms containing multiple overlapping peaks" and that "in a chromatographic data processing device in which a sample to be measured is separated by a column and the separated sample is detected and data processing is performed on the chromatogram obtained by detecting the separated sample, a fitting process is performed for each peak in an arbitrary time region in which multiple peaks exist in the chromatogram, either in the early time region or the late time region, and the processed peak is subtracted from the chromatogram in the arbitrary time region, thereby resolving the multiple peaks in the chromatogram" (see, for example, Patent Document 1).

[0004] Also, a technology is known that aims to "make it easier to find appropriate feature points using a chromatographic data processing device," and that "a chromatographic data processing device that processes data based on plot data measured by a chromatograph is provided with a virtual curve calculation unit that determines a virtual curve based on the measured plot data, a provisional feature point acquisition unit that determines provisional feature points based on the determined virtual curve, and an arithmetic processing unit that functions as an actual plot data feature point extraction unit that extracts actual plot data feature points corresponding to the provisional feature points from the measured plot data" (see, for example, Patent Document 2). [Prior art documents] [Patent documents]

[0005] [Patent Document 1] JP 2006-177980 A [Patent Document 2] JP 2019-174399 A Summary of the Invention [Problem to be solved by the invention]

[0006] In the technique disclosed in Patent Document 1, the calculated virtual curve and feature points tend to be significantly affected by noise.

[0007] Furthermore, in the technology of Patent Document 2, it tends to take a long time to consider the type of appropriate regression curve when finding feature points. That is, it is necessary to consider which type of regression curve to select, as well as which time axis range of the measurement data to regress and search for feature points.

[0008] The present invention has been made in consideration of the above-mentioned problems, and has an object to make it possible to easily find peak characteristic points in a short period of time while being less susceptible to the effects of noise. [Means for solving the problem]

[0009] In order to achieve the above objectives, The present invention relates to A data processing device for a chromatograph that processes data based on plot data measured by the chromatograph, a partial regression curve calculation unit that calculates a partial regression curve, which is a partial regression curve, for each of a plurality of data groups each including a predetermined number of measured time series data groups; a partial feature point acquisition unit that acquires partial feature points that are partial feature points for each of the partial regression curves that have been acquired; a peak feature point acquisition unit that acquires peak feature points in the measured time series data based on the partial feature points; The present invention is characterized by having the following.

[0010] This allows the feature points to be found in a relatively short time by finding a partial regression curve for a predetermined number of time-series plot data. Moreover, since it is easy to perform regression using a function with a relatively low degree, it is easy to prevent the regression curve from becoming biased due to the influence of noise. Effect of the Invention

[0011] In the present invention, peak characteristic points can be easily found in a short time while being less susceptible to the effects of noise. [Brief description of the drawings]

[0012] [Figure 1] FIG. 2 is a block diagram showing a schematic configuration of a chromatograph. [Diagram 2] FIG. 2 is a block diagram showing a schematic configuration of a data processing device for a chromatograph. [Diagram 3] FIG. 1 is a diagram showing an example of a chromatogram. [Figure 4] FIG. 11 is a diagram showing a schematic example of data for explaining an example of data processing. [Diagram 5] FIG. 13 is a diagram showing an example of a curve related to data processing. [Figure 6] FIG. 13 shows an example of an envelope method in a scanning PCF. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0013] Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings.

[0014] (Overall explanation) (Configuration of liquid chromatograph 100) 1 shows a schematic configuration of a liquid chromatograph 100. This liquid chromatograph 100 includes a mobile phase container 110 in which a liquid that is a mobile phase is stored, a pump 120 for delivering the mobile phase, an autosampler 130 for injecting a sample, a column 140 that is kept at a constant temperature by a column oven 141 and separates components in the sample, a detector 150 for detecting the separated components, a data processing device 160 for processing the detection results, and a display unit 170 for displaying the processing results. Each unit that constitutes the liquid chromatograph 100 can be configured in the same way as an ordinary device, except for the processing contents of the data processing device 160, and therefore detailed explanations will be omitted.

[0015] (Detailed configuration of data processing device 160) As shown in FIG. 2, the data processing device 160 includes a control processing unit 161, a data holding unit 162, and an arithmetic processing unit 163.

[0016] The control processing unit 161 controls the operation of the entire liquid chromatograph 100, and is provided with a control unit 161a, a measurement condition setting unit 161b that sets measurement conditions in response to operations on an operation panel (not shown), and a recording unit 161c that records measurement results, etc.

[0017] The data storage unit 162 stores data processed based on the measurement results.

[0018] The calculation processing unit 163 performs processing based on the measurement results, and functions as a partial regression curve calculation unit, a partial feature point acquisition unit, and a feature point acquisition unit. Specifically, for example, it includes a signal processing unit 163a that performs D / A conversion of an analog signal output from the detector 150, a calculation unit 163b that performs extraction and analysis of feature points, and a determination unit 163c that determines the analysis results.

[0019] (Data processing operation) In the liquid chromatograph 100, waveform data such as that shown in Fig. 3 can be obtained by a measurement operation. This waveform data has a horizontal axis representing time and a vertical axis representing signal intensity.

[0020] Since the relationship between time and components is known in advance, the components contained in the analytical sample are identified based on the retention time on the horizontal axis of the peak apex of the waveform data (qualitative processing).

[0021] Furthermore, the amount of the component contained is quantified based on the peak area of ​​the waveform data (quantitative calculation process).

[0022] This type of processing is carried out by extracting peak characteristic points such as peak apexes, start points, end points, valley points, and shoulder points, as shown in a typical example in Figure 3, and then setting a baseline line segment based on these.

[0023] The extraction of the peak characteristic points is performed by the processing of the arithmetic processing unit 163 as described below.

[0024] (Extraction of peak feature points) For example, 1000 points of plot data P1 to P1000 are acquired for each predetermined sampling period, which are discrete data actually detected by detector 150, and among the plot data P1 to P1000, for example, consecutive plot data P_Start to P_End in a predetermined time interval from time t_Start to time t_End as shown in Fig. 4 is set as a processing target. Here, time t_Start and time t_End may be determined in advance, or may be specified by the user.

[0025] Seven consecutive plot data points P_n1 to P_n7 are selected while being shifted one by one from the start point to the end point of the time interval, and a quadratic curve C is obtained as a partial regression curve approximating the seven plot data points P_n1 to P_n7, as shown in Fig. 5. The quadratic curve C is obtained, for example, by the least squares method. Note that a curve of a cubic or higher degree function, a Gaussian curve, or various approximation methods other than the least squares method may be used.

[0026] For example, if the start point of the time interval is P1 and the end point is P15, Quadratic curve C_P1-P7 based on plot data P1 to P7, Quadratic curve C_P2-P8 based on plot data P2 to P8, … Quadratic curve C_P8-P14 based on plot data P9-P15, Quadratic curves C_P9-P15 are obtained from the plot data P9 to P15, respectively.

[0027] That is, when there are m points of plot data within a time interval, and n consecutive points of plot data (where m>n) are selected, m-n+1 quadratic curves C are obtained.

[0028] For the group of quadratic curves C thus obtained over the entire time interval, it is confirmed whether differential feature points (maximum points, minimum points), which are partial feature points, are included within the range of the selected data point group in each quadratic curve. If differential feature points are included, the times corresponding to each differential feature point are set as candidates for the times of peak feature points (peak apex, peak start point, peak end point). Based on the data points at the times corresponding to these differential feature points, the peak feature points (peak apex, peak start point, peak end point) of the approximate waveform of the time series data (plot data) of the detection signal are set, and these peak feature points are output.

[0029] Specifically, for a group of quadratic curves C obtained over the entire time interval, peak characteristic points are obtained by the following procedure.

[0030] (1) That is, for example, the points on the quadratic curve C at the times of the plot data P2 in Fig. 5(a) and the plot data P13 in Fig. 5(i) are determined to be local minimum points, or local maximum points. Also, the points on the quadratic curve C at the times of the plot data P8 in Figs. 5(d) to (f) are determined to be local maximum points, or local maximum points.

[0031] (2) However, if the time of such a maximum point or minimum point is not within the time range for obtaining each partial regression curve, it may be ignored. Also, if it is not within the time range for obtaining each partial regression curve, but is within a time range of ±2 sampling periods from the center of the time range, it may be determined as a partial characteristic point.

[0032] (3) In addition, in the examples of Figures 5(d) to (f), all of the maximum points occur at the time of plot data P8. However, if there are multiple quadratic curves with such maximum points within a specified time range and the times at which the maximum points occur are different from one another, the maximum points may be grouped together as follows.

[0033] For example, if the difference in time between the maximum points is within two sampling periods (sense range), it is determined that one peak exists, and the time at which the maximum value or raw data of the maximum points is maximum may be determined as the peak characteristic point. The sense range as described above is set, for example, according to the expected peak width or sampling period, and is not limited to being set fixedly, but may be set by the user.

[0034] Here, liquid chromatography is a measurement in which the measured values ​​change over time, and each measured value is a series of values ​​per unit time. In contrast, the data collected per unit time changes due to the separation performance of the column and the effects of noise. Therefore, a certain range of expected data can be considered as a single inseparable event mass (represented by a single piece of data).

[0035] Furthermore, even if there are local maximum points at times that are separated by, for example, two or more sampling periods, if the raw data of all the plots at the times in between are within a predetermined noise amplitude, the plot data can be determined to be from one peak affected by noise, and the point with the larger raw data can be determined to be the peak characteristic point.

[0036] Here, the above-mentioned predetermined noise amplitude is, for example, as follows. That is, a value of a certain width is added to the measurement data due to various factors accompanying the measurement. This value is defined only by the width, and there are non-periodic and periodic components, but since the acceptance correlation between the period and the measurement peak is not constant, it is considered to be a noise component together. In other words, the noise width in the measurement value is considered to be an uncertain value. In other words, even if measurement value A>measurement value B, if measurement value A-measurement value B<noise width, the exact difference between A and B is uncertain. The above-mentioned noise amplitude can be set, for example, according to the amplitude of the measurement value at a time sufficiently distant from the peak point.

[0037] As described above, by finding a partial regression curve for a predetermined number of time-series plot data, it is possible to find feature points in a relatively short time. Moreover, since it is easy to perform regression using a function with a relatively low degree, it is easy to prevent the regression curve from becoming biased due to the influence of noise. In particular, since it is not necessary to select the type of regression curve or the time axis range for finding the regression curve, it is easy to perform objective processing quickly.

[0038] The sets of plot data for obtaining each partial regression curve are not limited to those shifted by one sampling period, but may be those shifted by two or more sampling periods, as long as the peak characteristic points can be obtained with the required accuracy. Moreover, the plot data of each set is not necessarily limited to continuous plot data, and discrete data or bunched data may be used.

[0039] (Scanning CCF: Complementary Curve Fitting and Scanning PCF: Scanning Partial Curve-Fitting) The Savitzky-Golay method is a type of moving average and a smoothing method, but the purpose of suppressing the influence of noise is the same as in this embodiment. The moving average has a window that cuts out a certain time period, and a representative value is assigned to the representative point at the center time of the window. The window may be, for example, 7 points or 9 points, and the detection signal representative value at the representative time is replaced with the moving average value one after another by sweeping backward in time. Alternatively, it may be swept forward. In either case, the function form is y=f(t), several signal data points within the time window are summarized, and one value of the objective variable y is assigned to the explanatory variable time t. By switching the weighting coefficient of the Savitzky-Golay method, it is also possible to calculate a differential coefficient such as y=f'(t). The smoothing method is a method that simply cancels out noise, but the purpose of this embodiment is to find a true curve (assumed to be true).

[0040] This embodiment also has a time window, and the window is swept, which is common to the Savitzky-Golay method, etc. However, this embodiment does not assign a representative value to the center time, but rather regresses a virtual curve in each window, which is a feature of the scanning CCF or scanning PCF, as described later.

[0041] In the case of a quadratic curve, a vertex or valley point may be obtained. For example, when analyzing the vicinity of a vertex, the t and y coordinates of the vertex are determined for each window. In other words, these are the vertex time t and signal intensity y as differential characteristic points of each regression curve. Even if the window is shifted slightly, if the vertex coordinates (t, y) are almost the same, it can be determined that this is the true vertex. The same is true for valley points. The method of finding this differential characteristic point falls under the category of scanning CCF.

[0042] This method can be extended to not only linear and quadratic polynomials but also higher-order polynomials. This virtual curve is not limited to probability density functions such as normal distribution, or analytical mathematical functions such as exponential functions and logarithmic functions. It can also be extended to any function. This method is called scanning CCF (Complementarily Curve Fitting) because the curve fitting process, which is performed supplementarily for each time window, is swept one after another. The name comes from the fact that the curve fitting process is used supplementarily to find differential feature points.

[0043] On the other hand, the method to find the true curve is called Scanning Partial Curve-Fitting (PCF) because it uses the same time window for the entire true curve and scans the curve fitting process partially one after another. Scanning PCF is not used as a supplement to curve fitting, but is mainly used to find the true curve. Scanning CCF is used to search for differential feature points, and scanning PCF is used to search for the true curve.

[0044] Another similar signal analysis method is spline interpolation, such as cubic spline. Spline interpolation also uses polynomials, but it is a method for interpolating between real data points, and does not aim to find the true curve buried in noise. Also, the moving average mentioned above does not have an interpolation function. Scanning PCF removes the effects of noise and also has an interpolation function.

[0045] (Examples of scanning CCF and scanning PCF) While the scanning CCF finds differential characteristic points such as vertices and valleys, the scanning PCF examines the difference between virtual curves in the overlapping section between two windows, i.e., the residual. For example, by evaluating the sum of squared residuals of two virtual curves, it is possible to compare whether the virtual curves are identical or not. By considering the virtual curves that match as the true curve, the true curve for the entire time window section can be found. In fact, if the y coordinates along the vertical axis for each time window match within a certain tolerance, the virtual curves that match are extracted and the true curve is generated by connecting them next to each other. The scanning PCF is characterized by finding the true curve, and is different from smoothing processing that cancels out noise components.

[0046] In the scanning PCF, it is desirable to use a quadratic or cubic polynomial for the virtual curve. However, since a special case of a curve is a linear straight line, it is not impossible to use a regression line for the scanning PCF. As mentioned above, the main purpose of this embodiment is to examine the degree of difference between adjacent regression curves. In the case of a linear straight line, the degree to which the first-order differential coefficient, i.e., the slope, matches or does not match for each time window is examined. If they match almost, it can be determined that the slope is almost constant in the matching section. When determining the first-order differential coefficient, i.e., the rate of change of the regression line, it is sufficient to regress to a linear polynomial, i.e., a straight line. This embodiment is a method for finding a true curve including a true straight line from detection signal data that is scattered, i.e., where noise is superimposed. At the same time, if the regression function is a differentiable function, the differential coefficient of the function can also be calculated.

[0047] The shoulder point that characterizes the shoulder peak, i.e., the inflection point, can also be found by extending this logic. The inflection point is at the time when the second-order partial differential coefficient becomes zero. If a quadratic function is regressed for each time window, the first-order differential coefficient can of course be obtained, and if the rate of change per unit time is observed, the inflection point can also be found. Another feature of the scanning PCF is that the first-order differential coefficient is not a representative value for each time, as in smoothing processing, but is found as a first-order differential curve.

[0048] The smoothing process mainly shifts the time window from one to the next, but in this embodiment, it is not necessary to shift it by one frame. Sweeping is possible with every other frame or every two frames. The number of time window points can be selected from 7, 9, 11, etc. Furthermore, the expected peak width can be input and the sampling period and the number of time window points can be automatically calculated. In the case of every other peak or every two peaks, bunching processing that adds up data points is also possible. In fact, for both scanning PCF and scanning CCF, it is desirable to set the number of data points per peak to 20 to 30 points, or at most 50 points, and it is important for this embodiment that the sampling period and bunching points are automatically calculated with a representative peak width as an input value. In other words, the typical sampling period of HPLC (high performance liquid chromatography) is several hundred milliseconds, and that of UHPLC (ultra high performance liquid chromatography) is several tens of milliseconds.

[0049] In the case of scanning CCF, there are several methods for comparing time windows. For example, if the valley point time t of each time window is within plus or minus the sampling period, it can be judged as stable. Similarly, the vertical axis y value of the valley point can also be judged to be within a certain tolerance range. It is also possible to set a rectangular tolerance range for the coordinates (t, y).

[0050] In the case of scanning PCF, there is also a tolerance range for finding the true curve. An envelope of a certain width, such as plus or minus a few mV, is set along the vertical axis on one of the two virtual curves to be compared, and the other virtual curve is considered to match if it is within that envelope. If it falls outside the envelope, it does not match. In some cases, the envelope is more convenient than the comparison of the residual sum of squares mentioned above. This will be explained in more detail later using Figure 6.

[0051] A more advanced technique is to consider the contribution ratio. A series of time windows are numbered T1, T2, T3, etc., and the result of the regression calculation is evaluated as the contribution ratio R1. 2 , R2 2 , R3 2The more successful the regression process is, the better the explanation is provided by the regression curve, and the closer the contribution rate is to 1. The more likely the regression curve is to be weighted toward the regression curve with a good regression, and this is used as an index to find the true curve. The method of using the contribution rate as an index can also be used to find differential feature points.

[0052] Finally, by using scanning PCF, the first derivative of the true curve can be obtained, and the start or end point of the peak can be found. This can also be interpreted as finding the differential characteristic points. In addition, the true curve obtained by scanning PCF is the true peak waveform itself, so accurate retention times and peak areas can be calculated from it.

[0053] (Simple scanning PCF) For example, in the case of a time window that includes five data points, the weighted regression method is to assign a weight of 1 to the central data point, a weight of 2 / 3 to the data point next to it, a weight of 1 / 3 to the data point next to that, and so on, and perform the least squares calculation. Then, a set of coefficients of a regression curve such as a quadratic polynomial is assigned to the time of the central point. By sweeping this central time, each set of coefficients is assigned to each central time in turn. When interpolating, these adjacent regression curves are mixed at the internal division points. In other words, at the relevant time, the ratio of the regression curve at the relevant time is set to 1.0, and the ratio of the regression curve at the adjacent time is set to 0.0, and then they are added. At the internal division points between the relevant time and the adjacent time, the respective regression curves are mixed according to the ratio of the internal division points according to the time difference. At the midpoint, they are mixed at a ratio of 0.5. At the adjacent time, the ratio of the regression function of the adjacent time is 1.0. In this way, the curve obtained is one true curve.

[0054] (Envelope Method) There are cases where one of the two virtual curves matches the other virtual curve within a certain tolerance range, and cases where they do not match. Figure 6 illustrates the envelope method. First, the virtual curve 0 is regressed using five points of actual measurement data before and after the representative time t0. The time is t ―2 From ―1There are five points: t0, t1, and t2. These subscripts are the time notations for the virtual curve 0.

[0055] Next, shift one data point backward and regress the virtual curve 1. Evaluate how well the virtual curve 1 matches the virtual curve 0. To do this, set an envelope with the input width above and below the virtual curve 0. The virtual curve 1 is t ―1 There are four overlapping points from t0 to t2. In the example of Figure 6, among those four time points, virtual curve 1 at t0 and t1 exists within the envelope. Therefore, virtual curve 0 and virtual curve 1 are considered to coincide at only those two points. In this case, the true curve is calculated as shown below (Equation 1) and is applied only to the coincident times. y0 represents virtual curve 0, y1 represents virtual curve 1, etc., and y represents the true curve (assumed to be true). The coefficient of the true curve y is the average value of the coefficients of the virtual curves.

[0056]

number

[0057] In the case of Figure 6, the true curve y is n=1 (n is the number of virtual curves used to derive true curve y - 1), which means virtual curve 0 and virtual curve 1, and the time is applied only to t0 and t1, the time intervals when virtual curve 0 and virtual curve 1 coincide. Now that the true curve y has been obtained, both its first-order differential curve y' and its second-order differential curve y" can be found from the differential coefficients of the polynomial.

[0058] Here we will discuss the relationship between the upper and lower tolerances of the envelope and the likelihood of the true curve, in other words, reliability. If the tolerance is narrow, reliability improves, but the possibility that adjacent virtual curves will not fall within the envelope increases. In extreme cases, only one true curve will be assigned to the center time. Conversely, if the tolerance is too wide, the number of virtual curves to calculate the average value will increase, and although an average value can be obtained, the variation will be large and reliability will decrease. One guideline for an appropriate tolerance is a few percent of the peak height. Of course, it is also possible to automatically input the envelope tolerance based on this few percent.

[0059] (Method of time-varying regression interval) In the above embodiment, the time interval of the five data points is shifted backward one by one. If the virtual curve obtained from the previous five points fits the actual data well, there are cases where the next sixth data point can be explained with a certain tolerance. It can be imagined using the above-mentioned envelope method, or it can be understood by the residual between the virtual curve and the actual data. For example, the average value of the sum of squared residuals in the case of five points is compared with the average value of the sum of squared residuals in the case of six points, and if the former is smaller, the regression process is stopped at the five point. Conversely, if the average value of the sum of squared residuals obtained from the latter six points is not large, it is determined that the regression curve can be extended, and the regression process is re-executed with a time interval of six points. As a result, a new virtual curve in the six-point interval is obtained. By repeating this process, the time interval to be regressed can be made variable. This method of time-variable regression interval is different from the above-mentioned envelope method in that it does not compare two virtual curves, but rather compares a virtual curve with a sequence of actual data points. It can also be extended to compare two hypothetical curves by, for example, taking the feasible curve obtained in the seven-point interval as the base and shifting it to the next five-point time interval ending at the eighth data point.The two hypothetical curves are connected only in the interval where they coincide to generate the true curve.

[0060] (Gaussian Connection Using Machine Learning) The ideal form of a chromatographic peak is considered to be a normal distribution, i.e., a Gaussian. Furthermore, a probability density function that represents a tailing peak, i.e., an exponentially modified Gaussian EMG, is often expressed by a single function. In light of the idea of ​​the scanning PCF of this embodiment, it is not necessary to express it by a single function, but a method can be devised in which a partial adaptation to a Gaussian is made and then finally connected to a true curve. Simply put, the peak can be divided vertically into three parts, namely, the apex, the mid-mountain near the inflection point, and the skirt part. In each divided part, the apex of the Gaussian, the left and right mid-mountains, and the left and right skirts are approximated by separate Gaussians. This is not a vertical division into three parts, but the number of divisions can be increased to four or five. A vertical division into three parts results in five sections in the time direction, a vertical division into four parts results in seven sections in the time direction, and a vertical division into five parts results in nine time sections.

[0061] It is desirable to apply machine learning to the problem of how to determine the appropriate number of divisions and what kind of function, such as Gaussian, EMG, or in some cases polynomial, should be used for regression. Many examples of chromatographic peaks are input into supervised machine learning in advance and applied to individual analyses. Adding overlapping peaks and shoulder peaks will make the analysis even more complicated, but the adaptability of this embodiment will increase accordingly. The examples input in advance may be actual measured data or artificially generated chromatographic peaks.

[0062] This method of connecting Gaussians using machine learning, like the time-variable regression interval method described above, is a method of comparing a virtual curve with an actual data time series. [Explanation of symbols]

[0063] 100 Liquid Chromatograph 110 Mobile phase container 120 Pump 130 Autosampler 140 Column 141 Column Oven 150 Detectors 160 Data Processing Device 161 Control processing section 161a Control section 161b Measurement condition setting section 161c Recording Department 162 Data storage unit 163 Processing Unit 163a Signal processing section 163b Arithmetic unit 163c Judgment part 170 Display section

Claims

1. A data processing device for a chromatograph that processes data based on plot data measured by the chromatograph, a partial regression curve calculation unit that calculates a partial regression curve, which is a partial regression curve, for each of a plurality of data groups each including a predetermined number of measured time series data groups; a partial feature point acquisition unit that acquires partial feature points that are partial feature points for each of the partial regression curves that have been acquired; a peak feature point acquisition unit that acquires peak feature points in the measured time series data based on the partial feature points; A chromatographic data processing device comprising:

2. 2. The data processing device of claim 1, 4. A data processing device for a chromatography, wherein the regression curve is either a curve of a function of second degree or higher or a Gaussian curve.

3. 3. The data processing device of claim 2, further comprising:

2. A data processing device for a chromatography, wherein the plurality of data groups are data groups shifted in time from one another.

4. 4. The data processing device of claim 3, further comprising:

2. A data processing apparatus for a chromatography, wherein the local characteristic points are at least one of a maximum point and a minimum point.

5. 3. The data processing device of claim 2, further comprising:

2. A data processing device for a chromatography, wherein the peak characteristic point is any one of a peak apex, a peak start point, and a peak end point.

6. A chromatographic unit that separates and measures components contained in a sample; A data processing device for a chromatograph according to claim 1; A chromatograph comprising:

7. A data processing method for a chromatograph, which performs data processing based on plot data measured by a chromatograph, comprising the steps of: a partial regression curve calculation step of calculating a partial regression curve, which is a partial regression curve, for each of a plurality of data groups each including a predetermined number of measured time series data groups; a partial feature point obtaining step of obtaining partial feature points which are partial feature points for each of the obtained partial regression curves; a peak feature point acquisition step of determining a peak feature point in the measured time series data based on the partial feature points; A method for processing chromatographic data, comprising:

8. A data processing device for a chromatograph that processes data based on plot data measured by the chromatograph, a partial regression curve calculation unit that calculates a partial regression curve, which is a partial regression curve, for each of a plurality of data groups each including a predetermined number of measured time series data groups; a partial curve comparison unit for comparing the partial regression curves thus obtained with each other; an integrated regression curve acquisition unit that obtains a regression curve in a wider range than each of the partial regression curves in the measured time series data based on a result of the comparison; A chromatographic data processing device comprising:

9. A data processing device for a chromatograph that processes data based on plot data measured by the chromatograph, a partial regression curve calculation unit that calculates a partial regression curve, which is a partial regression curve, for each of a plurality of data groups each including a predetermined number of measured time series data groups; a partial curve comparison unit for comparing each of the partial regression curves thus obtained with the time series data; an integrated regression curve acquisition unit that obtains a regression curve in a wider range than each of the partial regression curves in the measured time series data based on a result of the comparison; A chromatographic data processing device comprising:

Images