An intelligent auditing system for enterprise management system certification process

Abstract

The application discloses an intelligent auditing system for enterprise management system certification process, and relates to the technical field of intelligent auditing.The system comprises a data collection module, a data processing module and an output module.The data collection module is used for obtaining auditing material data, auditing standard data and historical auditing data.The data processing module is used for auditing the auditing material data through a correlation auditing method, and classifying unqualified elements into sampling elements and non-sampling elements.For the sampling elements, after the unqualified elements are reviewed and modified, a modified calibration factor of the modified elements is calculated, and the elements are audited according to a round auditing method.The output module is used for outputting an auditing result.The application can reduce the auditing range and the calculation amount by modifying the calibration factor of the unqualified elements and the change relationship between the unqualified elements and the related elements, and re-auditing the elements which are changed after the modification, so that the auditing efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent auditing technology, specifically to an intelligent auditing system for enterprise management system certification processes. Background Technology

[0002] In enterprise management system certification, intelligent auditing methods rely on technologies such as AI and big data to achieve efficient and accurate audits. Image recognition quickly locates product defects, improving detection speed and accuracy; NLP analyzes unstructured documents, assisting auditors in quickly obtaining key information; and the integration of IoT and cloud computing enables remote monitoring and real-time data analysis, reducing the number of on-site visits. Furthermore, intelligent auditing systems can automatically generate audit reports, providing preliminary screening results, reducing the workload of manual audits, achieving human-machine collaboration, and improving audit efficiency and quality.

[0003] During the review process, certain review methods need to be followed to improve review efficiency and accuracy. For example, a multi-dimensional feature-based intelligent review method, device, and medium (patent publication number CN119963318A) includes: constructing a data approval model and obtaining data to be approved based on the data approval model; wherein the data approval model includes a data cleaning model, a feature extraction model, a classification and recognition model, and a result approval model; cleaning and organizing the data to be approved based on the data cleaning model to generate standardized data; extracting features from the standardized data based on the feature extraction model to generate collected features; processing the collected features based on the classification and recognition model to classify the collected features and obtain classification results; and processing the classification results according to a preset weight ratio based on the result approval model to generate approval results. This method can automatically review the data to be approved, improving review efficiency, accuracy, and consistency while reducing operating costs.

[0004] In enterprise management systems, various types of data need to be audited. Auditing is usually carried out using a combination of AI and human review. In some projects, if the audit fails, it needs to be corrected and audited again to ensure that the audit passes. However, each audit requires a complete audit process. When the amount of data is large, the amount of calculation is large, which slows down the audit process and affects the production operation. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent auditing system for the enterprise management system certification process, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an intelligent auditing system for enterprise management system certification processes, comprising:

[0007] Data collection module: Acquires data on materials to be reviewed, review standards, and historical review data. Historical review data includes initial review material data and modification records.

[0008] Data processing module:

[0009] Based on the audit standard data and the data of the materials to be audited, the audit material data is audited using the correlation audit method;

[0010] Based on historical audit data, non-compliant elements are classified into sampled elements and non-sampled elements using a sampling classification method;

[0011] For the sampled elements, a calibration factor calculation model is established based on the calibration factor calculation method and the variation range between the target sampled element and the other elements.

[0012] After the unqualified elements are reviewed and corrected, they are reviewed again. When the unqualified elements are corrected and pass the review, the modified calibration factor is calculated using the calibration factor calculation model.

[0013] Based on the revised calibration factors and the round of review, the review will be conducted according to the round-based review method.

[0014] For non-sampled elements, after the data fails the review and is modified, the modified data is reviewed again using the correlation audit method.

[0015] Output module: Used to output the audit results.

[0016] Preferably, the calibration factor calculation method includes:

[0017] In the historical audit data, the samples are grouped according to the type of non-compliant element, so that the samples in the same group have the same non-compliant element;

[0018] Select any group, compare the data of the elements in the target group before and after modification (excluding the unqualified elements), and mark the other elements whose data has changed as related elements of the unqualified elements;

[0019] The sample modification calibration factor is set to a constant. Based on the modification change rate of non-conforming elements and the modification change rate of related elements, a calibration factor calculation model is established, specifically as follows:

[0020] ;

[0021] in This indicates that the calibration factor has been modified, and the modified calibration factor in the sample is 1. and Represents the fitting coefficient. Indicates the total number of associated elements. This represents the modified value of the non-compliant element. This represents the value of the unqualified element before modification. This represents the modified value of the i-th associated element. This represents the value of the i-th associated element before modification;

[0022] The calibration factor calculation model is trained using multiple samples from the target group to obtain... and The numerical value is then used to input the data of the materials to be reviewed and the modification records into the model to obtain the modification calibration factor after this modification.

[0023] Preferably, the calibration factor calculation method includes:

[0024] In the historical audit data, the samples are grouped according to the type and number of non-compliant elements, so that the samples in the same group have exactly the same non-compliant elements.

[0025] Select any group, compare the data of the elements in the target group before and after modification (excluding the unqualified elements), and mark the other elements whose data has changed as related elements of the combination of multiple unqualified elements;

[0026] The sample modification calibration factor is set to a constant. Based on the modification change rate of non-conforming elements and the modification change rate of related elements, a calibration factor calculation model is established, specifically as follows:

[0027] ;

[0028] Among them The calibration factor for the sample is a constant. and Represents the fitting coefficient. Indicates the total number of associated elements. This indicates the total number of non-compliant elements. This represents the modified value of the j-th non-compliant element. This represents the value of the j-th non-compliant element before modification. This represents the modified value of the i-th associated element. This represents the value of the i-th associated element before modification;

[0029] The calibration factor calculation model is trained using multiple samples from the target group to obtain... and The numerical value is then used to input the data of the materials to be reviewed and the modification records into the model to obtain the modification calibration factor after this modification.

[0030] Preferably, the round-based review method includes:

[0031] Record the audit stage during the audit. The audit stage is 0 for the first audit. Each time the audit fails and is audited again, the audit stage is increased by 1.

[0032] Based on the audit stage and the modified calibration factors, an audit judgment formula is established, specifically as follows:

[0033] ;

[0034] in This indicates a modification of the calibration factor. This indicates the review stage. If the review judgment formula is true, the review result is qualified; if the review judgment formula is false, the review result is unqualified.

[0035] When the audit stage reaches stage 4 and the audit is still unsatisfactory, the data from the last audit is treated as data to be audited and re-audited using the correlation audit method, and the audit stage is reset.

[0036] Repeat the reset process for the review stage until the review result is satisfactory.

[0037] Preferably, the relevance verification method includes:

[0038] Each element in the data of the materials to be reviewed is reviewed individually according to the rules. If the review fails, the review result is determined to be unqualified.

[0039] When the rule is approved, the historical correlation factor and the preliminary correlation factor for each sample are calculated based on the historical review data and the correlation factor calculation method.

[0040] Then, the preliminary correlation factor of the data to be reviewed is calculated using the correlation factor calculation method. The preliminary correlation factor of the data to be reviewed is compared with the historical correlation factor. If the difference is less than or equal to the correlation threshold, the review result is deemed qualified. If the difference is greater than the correlation threshold, the review result is deemed unqualified.

[0041] Preferably, the method for calculating the correlation factor includes:

[0042] Normalize the common elements in each sample of the historical audit data so that the values ​​are between 0 and 1;

[0043] After normalization, the standard deviation and mean of multiple elements in each sample are calculated. The preliminary correlation factor for each sample is then calculated using the following formula:

[0044] ;

[0045] in Indicates preliminary correlation factors, This represents the total number of elements in the sample. This represents the normalized value of the r-th element in the sample. This represents the average value of all elements in the normalized sample. This represents the standard deviation of each element in the normalized sample.

[0046] Calculate the measure of central tendency for each initial correlation factor, as a historical correlation factor.

[0047] Preferably, the method for calculating the association threshold includes:

[0048] Calculate the differences between multiple preliminary correlation factors and historical correlation factors in historical audit data;

[0049] Set a range coefficient so that more than 80% of the differences are less than the correlation threshold, and then obtain the correlation threshold.

[0050] Preferably, the sampling classification method includes:

[0051] The total number of samples and the non-compliant elements that appeared in the historical audit data were statistically analyzed.

[0052] Choose any non-compliant element and calculate the proportion of the sample size containing the target non-compliant element to the total sample size.

[0053] If the number of samples containing the target non-compliant element reaches the total number of elements in the sample, or if the number of samples containing the target non-compliant element accounts for more than 6.8% of the total number of samples, then the target non-compliant element is determined to be a sampled element; otherwise, the target non-compliant element is determined to be a non-sampled element.

[0054] Compared with the prior art, the beneficial effects of the present invention are:

[0055] By modifying the calibration factors to reflect changes in non-compliant elements and related elements, after a comprehensive review, only the elements that have changed after modification are reviewed again. This narrows the scope of the review, reduces the amount of calculation, and thus improves the efficiency of the review.

[0056] Meanwhile, the modified data is reviewed through round-by-round review and modification of calibration factors. The review requirements are raised each time the review fails, which leads to a larger scope of changes and inaccurate reviews. Moreover, the review stage can be reset after multiple review stages, and the current data is used as the data to be reviewed. This prevents the direction of modification from deviating due to too many modifications. Re-review can prevent the direction of modification from deviating too much and thus failing the review repeatedly.

[0057] Furthermore, when calculating the calibration factor, non-compliant elements can be calculated as a single element or as a group of multiple elements, using different calibration factor calculation models to ensure the accuracy of the calibration factor calculation. Attached Figure Description

[0058] Figure 1 This is a schematic diagram of the intelligent review process of the present invention;

[0059] Figure 2 This is a flowchart illustrating the round-review method in this invention;

[0060] Figure 3 This is a flowchart illustrating the correlation verification method in this invention;

[0061] Figure 4 This is a flowchart illustrating the correlation factor calculation method in this invention. Detailed Implementation

[0062] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0063] In this application, for ease of understanding, the method steps used do not necessarily need to be executed in the order of steps in this embodiment during actual operation. In other embodiments, these steps may be performed simultaneously or in a different order.

[0064] Example 1:

[0065] When the initial review fails, subsequent reviews can determine whether the revised data is acceptable based on the overall changes in other elements related to the non-compliant element, eliminating the need for another comprehensive review. This can effectively improve review efficiency when there are many elements to review.

[0066] like Figures 1-4 As shown, the present invention provides a technical solution: an intelligent auditing system for enterprise management system certification process, comprising:

[0067] Data collection module: Acquires data on materials to be reviewed, review standards, and historical review data. Historical review data includes initial review material data and modification records.

[0068] Data processing module:

[0069] Based on the audit standard data and the data of the materials to be audited, the audit material data is audited using the correlation audit method;

[0070] Based on historical audit data, non-compliant elements are classified into sampled elements and non-sampled elements using a sampling classification method;

[0071] For the sampled elements, a calibration factor calculation model is established based on the calibration factor calculation method and the variation range between the target sampled element and the other elements.

[0072] After the unqualified elements are reviewed and corrected, they are reviewed again. When the unqualified elements are corrected and pass the review, the modified calibration factor is calculated using the calibration factor calculation model.

[0073] Based on the revised calibration factors and the round of review, the review will be conducted according to the round-based review method.

[0074] For non-sampled elements, after the data fails the review and is modified, the modified data is reviewed again using the correlation audit method.

[0075] Output module: Used to output the audit results.

[0076] It should be noted that the audit standard data includes the specified range of each element and the relationship between certain elements, which are restricted by different audit standards. Furthermore, the data of the materials to be audited, the audit standard data, and the historical audit data have all been preprocessed (including but not limited to deduplication and missing data). The above processes are all existing technologies and will not be elaborated here.

[0077] like Figure 3 and Figure 4 As shown, the relevance verification methods include:

[0078] Each element in the data of the materials to be reviewed is reviewed individually according to the rules. If the review fails, the review result is determined to be unqualified.

[0079] When the rule is approved, the historical correlation factor and the preliminary correlation factor for each sample are calculated based on the historical review data and the correlation factor calculation method.

[0080] Then, the preliminary correlation factor of the data to be reviewed is calculated using the correlation factor calculation method. The preliminary correlation factor of the data to be reviewed is compared with the historical correlation factor. If the difference is less than or equal to the correlation threshold, the review result is deemed qualified. If the difference is greater than the correlation threshold, the review result is deemed unqualified.

[0081] The correlation factor is calculated as follows:

[0082] Normalize the common elements in each sample of the historical audit data so that the values ​​are between 0 and 1;

[0083] After normalization, the standard deviation and mean of multiple elements in each sample are calculated. The preliminary correlation factor for each sample is then calculated using the following formula:

[0084] ;

[0085] in Indicates preliminary correlation factors, This represents the total number of elements in the sample. This represents the normalized value of the r-th element in the sample. This represents the average value of all elements in the normalized sample. This represents the standard deviation of each element in the normalized sample.

[0086] Calculate the measure of central tendency for each initial correlation factor, as a historical correlation factor.

[0087] The methods for calculating the correlation threshold include:

[0088] Calculate the differences between multiple preliminary correlation factors and historical correlation factors in historical audit data;

[0089] Select the maximum difference from the differences, set a range coefficient for the maximum difference, and obtain the correlation threshold so that more than 80% of the differences are less than the correlation threshold.

[0090] It should be noted that, for ease of understanding, the following simulated data is used:

[0091] Suppose Company A is auditing the factory's exhaust emissions. The parameters to be audited are: emission temperature (°C), carbon dioxide content (mg / m³), carbon monoxide content (mg / m³), and humidity (g / m³). The audit standard data are (units are the same as above and omitted here): emission temperature 80-120°C, carbon dioxide content 30-40 mg / m³, carbon monoxide content 0-3 mg / m³, and humidity 8-23%.

[0092] Assume that the historical review data (after modification and approval) contains:

[0093] Sample 1: Emission temperature was 90°C, carbon dioxide content was 35%, carbon monoxide content was 2%, and humidity was 13%.

[0094] Sample 2: Emission temperature was 95°C, carbon dioxide content was 37%, carbon monoxide content was 1%, and humidity was 10%.

[0095] Sample 3: Emission temperature was 113°C, carbon dioxide content was 39 g / L, carbon monoxide content was 0 g / L, and humidity was 8 g / L.

[0096] ...

[0097] Data in materials pending review: Emission temperature is 93°C, carbon dioxide content is 40%, carbon monoxide content is 3%, and humidity is 12%.

[0098] First, the rules for each element are reviewed to ensure that the value of each element is within the range specified by the review standard data.

[0099] The above samples are normalized according to element type, using the following formula:

[0100] ;

[0101] in This represents the normalized value of the element in the p-th sample. This represents the value of the element in the p-th sample before normalization. This represents the largest value among all sample elements. This represents the smallest value among all sample elements.

[0102] Assume that the average values ​​obtained from the above multiple samples are emission temperature 98, carbon dioxide content 35, carbon monoxide content 1, and humidity 10.

[0103] The standard deviations calculated for the above samples were 5.82 for emission temperature, 1.36 for carbon dioxide content, 0.82 for carbon monoxide content, and 2.06 for humidity.

[0104] After normalization according to the formula, it becomes:

[0105] Sample 1: (0.26, 0.38, 0.67, 0.36);

[0106] Sample 2: (0.40, 0.63, 0.33, 0.14);

[0107] Sample 3: (0.91, 0.95, 0, 0);

[0108] ...

[0109] Meanwhile, the data of the materials to be reviewed is normalized according to the maximum and minimum values ​​of the historical review data, so the data of the materials to be reviewed after processing is (0.34, 1, 1, 0.29).

[0110] The mean of the multiple elements in Sample 1 is 0.42, and the standard deviation is 0.15.

[0111] The mean of the multiple elements in Sample 2 is 0.38, and the standard deviation is 0.18.

[0112] The mean of the multiple elements in sample 3 is 0.47, and the standard deviation is 0.47.

[0113] ...

[0114] The average value of the materials to be reviewed is 0.66, and the standard deviation is 0.34.

[0115] According to the formula Calculations show that:

[0116] The initial association factor for sample 1 is -0.07;

[0117] The preliminary association factor for sample 2 is -1.2;

[0118] The preliminary association factor for sample 3 is 1.2;

[0119] ...

[0120] The preliminary correlation factor for the data in the materials to be reviewed is 6.33.

[0121] Further calculation of the average value (any measure of central tendency can be used; in this embodiment, the average value is removed) yields a historical correlation factor of -0.06 (hypothetical value). Therefore, the difference between the preliminary correlation factor and the historical correlation factor of the data to be reviewed is 6.27.

[0122] Calculate the association threshold:

[0123] The calculated differences between the initial correlation factors and the historical correlation factors for multiple samples were 0.01, 1.14, 1.26, etc., and 80% of the differences were less than 1.9. Therefore, the correlation threshold was set to 1.9 (it can also be slightly larger, as long as 80% of the differences are less than this value, no specific restriction is imposed).

[0124] The difference between the preliminary correlation factor and the historical correlation factor of the data to be reviewed (6.27) is compared with the correlation threshold (1.9). If the difference is greater than the correlation threshold, it is judged as unqualified.

[0125] Therefore, although the rule review of individual elements is qualified, there may be potential factors that cause the overall data of the materials to be reviewed to be abnormal (in this example, the main one is the simultaneous large increase of carbon dioxide and carbon monoxide). The correlation review method can determine whether there are potential anomalies based on the correlation between the data. When potential anomalies are found, technical personnel can be notified to conduct manual review and confirm or modify the review results.

[0126] Sampling classification methods include:

[0127] The total number of samples and the non-compliant elements that appeared in the historical audit data were statistically analyzed.

[0128] It should be noted that non-conforming elements can be identified through rule-based review, or they can be identified manually by technicians upon receiving a notification of non-conformity, by inspecting and addressing various possible causes of non-conformity.

[0129] Choose any non-compliant element and calculate the proportion of the sample size containing the target non-compliant element to the total sample size.

[0130] If the number of samples containing the target non-compliant element reaches the total number of elements in the sample, or if the number of samples containing the target non-compliant element accounts for more than 6.8% of the total number of samples, then the target non-compliant element is determined to be a sampled element; otherwise, the target non-compliant element is determined to be a non-sampled element.

[0131] It should be noted that, for ease of understanding, the following simulated data is used:

[0132] Using the above simulation data, assuming that the total number of samples in the historical audit data is 100, the number of samples with emission temperature as the non-compliant element is 5. Although it is less than 6.8%, the number exceeds the total number of elements (emission temperature, carbon dioxide content, carbon monoxide content, and humidity, a total of 4). Therefore, emission temperature can be marked as the sampling element.

[0133] Suppose that in addition to these four elements, there are three other elements such as sulfur dioxide content. In this case, the number of samples with emission temperature as the non-compliant element is 5, which is less than 6.8%, and the number does not exceed the total number of elements. Therefore, emission temperature can be marked as a non-sampling element (in this embodiment, it is a sampling element).

[0134] The calibration factor calculation methods include:

[0135] In the historical audit data, the samples are grouped according to the type of non-compliant element, so that the samples in the same group have the same non-compliant element;

[0136] Select any group, compare the data of the elements in the target group before and after modification (excluding the unqualified elements), and mark the other elements whose data has changed as related elements of the unqualified elements;

[0137] The sample modification calibration factor is set to a constant. Based on the modification change rate of non-conforming elements and the modification change rate of related elements, a calibration factor calculation model is established, specifically as follows:

[0138] ;

[0139] in This indicates that the calibration factor has been modified, and the modified calibration factor in the sample is 1. and Represents the fitting coefficient. Indicates the total number of associated elements. This represents the modified value of the non-compliant element. This represents the value of the unqualified element before modification. This represents the modified value of the i-th associated element. This represents the value of the i-th associated element before modification;

[0140] The calibration factor calculation model is trained using multiple samples from the target group to obtain... and The numerical value is then used to input the data of the materials to be reviewed and the modification records into the model to obtain the modification calibration factor after this modification.

[0141] It should be noted that, for ease of understanding, the following simulated data is used:

[0142] Using the simulation data above, emission temperature was labeled as a sampling element;

[0143] Assuming the historical review data contains (the value before modification):

[0144] Sample 1: Emission temperature was 123°C, carbon dioxide content was 39 g / L, carbon monoxide content was 0 g / L, and humidity was 9 g / L.

[0145] Sample 2: Emission temperature was 125°C, carbon dioxide content was 40%, carbon monoxide content was 0%, and humidity was 8%.

[0146] Sample 3: Emission temperature was 133°C, carbon dioxide content was 40%, carbon monoxide content was 0%, and humidity was 8%.

[0147] ...

[0148] The calibration factor calculation model is trained using data from multiple samples (using nonlinear least squares, or other existing methods; no specific restrictions apply). and The values ​​are approximately 0.89 and -0.5, respectively.

[0149] Assuming that the correlation review fails, the technicians adjust the parameters of the exhaust equipment, and the subsequent data are: exhaust temperature 96, carbon dioxide content 38, carbon monoxide content 1, and humidity 11.

[0150] By substituting the original and modified values ​​of the data in the materials to be reviewed into the formula, the modification calibration factor can be calculated. ≈0.60.

[0151] like Figure 2 As shown, the round-based review methods include:

[0152] Record the audit stage during the audit. The audit stage is 0 for the first audit. Each time the audit fails and is audited again, the audit stage is increased by 1.

[0153] Based on the audit stage and the modified calibration factors, an audit judgment formula is established, specifically as follows:

[0154] ;

[0155] in This indicates a modification of the calibration factor. Indicates the amplitude coefficient. This indicates the review stage. If the review judgment formula is true, the review result is qualified; if the review judgment formula is false, the review result is unqualified.

[0156] When the audit stage reaches the set value and the audit is still unsatisfactory, the data from the last audit is treated as data to be audited and re-audited using the correlation audit method, and the audit stage is reset.

[0157] Repeat the reset process for the review stage until the review result is satisfactory.

[0158] It should be noted that, for ease of understanding, the following simulated data is used:

[0159] Using the simulation data above, when the first audit was conducted using the correlation audit method, the audit stage was 0, but the audit failed. Modifications were then made, and the modification calibration factor was calculated after the modifications. =1, and during the second review, the review stage =1.

[0160] at this time =0.4, =0.13, therefore the audit judgment formula is invalid, and it is still unqualified after modification. At this time, the equipment in the factory can continue to be debugged, and the above audit process can be repeated again. If the subsequent audit passes, it means that it is qualified. If it fails multiple times in the subsequent audit, when the audit stage accumulates to 4, it will be re-audited through the correlation audit method and the audit stage will be reset to prevent too many modifications from causing the modification direction to deviate. Re-auditing can prevent the modification direction deviation from being too large, which will cause the audit to fail all the time.

[0161] Example 2:

[0162] In Example 1, when distinguishing between sampled elements and non-sampled elements, only one non-conforming element is always present in the sample. The individual non-conforming element is distinguished. However, in actual production, there may be more than one non-conforming element in the same sample. In this case, other methods such as sampling classification and round-based auditing can treat non-conforming elements as a group of elements for calculation. However, for the calibration factor calculation method, it is necessary to determine the calibration factor calculation model separately. Therefore, this embodiment provides another calibration factor calculation method.

[0163] The calibration factor calculation methods include:

[0164] In the historical audit data, the samples are grouped according to the type and number of non-compliant elements, so that the samples in the same group have exactly the same non-compliant elements.

[0165] Select any group, compare the data of the elements in the target group before and after modification (excluding the unqualified elements), and mark the other elements whose data has changed as related elements of the combination of multiple unqualified elements;

[0166] The calibration factor of the sample is set as a constant. Based on the modification rate of non-conforming elements and the modification rate of related elements, a calibration factor calculation model is established, specifically as follows:

[0167] ;

[0168] Among them The calibration factor representing the sample modification is a constant. and Represents the fitting coefficient. Indicates the total number of associated elements. This indicates the total number of non-compliant elements. This represents the modified value of the j-th non-compliant element. This represents the value of the j-th non-compliant element before modification. This represents the modified value of the i-th associated element. This represents the value of the i-th associated element before modification;

[0169] The calibration factor calculation model is trained using multiple samples from the target group to obtain... and The numerical value is then used to input the data of the materials to be reviewed and the modification records into the model to obtain the modification calibration factor after this modification.

[0170] It should be noted that, for ease of understanding, the following simulated data is used:

[0171] Use the modified data from Example 1;

[0172] Assuming the historical review data contains (the value before modification):

[0173] Sample 1: Emission temperature was 123°C, carbon dioxide content was 42%, carbon monoxide content was 2%, and humidity was 9%.

[0174] Sample 2: Emission temperature was 125°C, carbon dioxide content was 43%, carbon monoxide content was 3%, and humidity was 8%.

[0175] Sample 3: Emission temperature was 133°C, carbon dioxide content was 45%, carbon monoxide content was 2%, and humidity was 8%.

[0176] ...

[0177] Assume the data of the material to be reviewed is as follows: emission temperature 122°C, carbon dioxide content 41%, carbon monoxide content 0%, and humidity 9%.

[0178] In the case of failure to pass the correlation review, the technicians adjusted the parameters of the exhaust equipment, and the subsequent data were: exhaust temperature 96°C, carbon dioxide content 38°C, carbon monoxide content 1°C, and humidity 11%.

[0179] The calibration factor calculation model was trained using data from multiple samples to obtain... and The value of is calculated using existing techniques (such as nonlinear least squares or Bayesian methods), and will not be elaborated upon here.

[0180] By substituting the original and modified values ​​of the data in the materials to be reviewed into the formula, the modification calibration factor can be calculated. .

[0181] This method can be used when there are multiple non-compliant elements, treating the non-compliant elements as a group of elements to accurately calculate the modification calibration factor.

[0182] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended embodiments and their equivalents.

Claims

1. An intelligent auditing system for enterprise management system certification processes, comprising: Data collection module: Acquires data on materials to be reviewed, review standards, and historical review data. Historical review data includes initial review material data and modification records. Its features are: Data processing module: Based on the audit standard data and the data of the materials to be audited, the audit material data is audited using the correlation audit method; Based on historical audit data, non-compliant elements are classified into sampled elements and non-sampled elements using a sampling classification method; For the sampled elements, a calibration factor calculation model is established based on the calibration factor calculation method and the variation range between the target sampled element and the other elements. After the unqualified elements are reviewed and corrected, they are reviewed again. When the unqualified elements are corrected and pass the review, the modified calibration factor is calculated using the calibration factor calculation model. Based on the revised calibration factors and the round of review, the review will be conducted according to the round-based review method. For non-sampled elements, after the data fails the review and is modified, the modified data is reviewed again using the correlation audit method. Output module: Used to output the audit results; The calibration factor calculation method includes: In the historical audit data, the samples are grouped according to the type of non-compliant element, so that the samples in the same group have the same non-compliant element; Select any group, compare the data of the elements in the target group before and after modification (excluding the unqualified elements), and mark the other elements whose data has changed as related elements of the unqualified elements; The sample modification calibration factor is set to a constant. Based on the modification change rate of non-conforming elements and the modification change rate of related elements, a calibration factor calculation model is established, specifically as follows: ； in This indicates that the calibration factor has been modified, and the modified calibration factor in the sample is 1. and Represents the fitting coefficient. Indicates the total number of associated elements. This represents the modified value of the non-compliant element. This represents the value of the unqualified element before modification. This represents the modified value of the i-th associated element. This represents the value of the i-th associated element before modification; The calibration factor calculation model is trained using multiple samples from the target group to obtain... and The numerical value is then used to input the data of the materials to be reviewed and the modification records into the model to obtain the modification calibration factor after this modification. The round-based review method includes: Record the audit stage during the audit. The audit stage is 0 for the first audit. Each time the audit fails and is audited again, the audit stage is increased by 1. Based on the audit stage and the modified calibration factors, an audit judgment formula is established, specifically as follows: ； in This indicates a modification of the calibration factor. This indicates the review stage. If the review judgment formula is true, the review result is qualified; if the review judgment formula is false, the review result is unqualified. When the audit stage reaches stage 4 and the audit is still unsatisfactory, the data from the last audit is treated as data to be audited and re-audited using the correlation audit method, and the audit stage is reset. Repeat the reset process for the review stage until the review result is satisfactory; The relevance verification method includes: Each element in the data of the materials to be reviewed is reviewed individually according to the rules. If the review fails, the review result is determined to be unqualified. When the rule is approved, the historical correlation factor and the preliminary correlation factor for each sample are calculated based on the historical review data and the correlation factor calculation method. Then, the preliminary correlation factor of the data to be reviewed is calculated by the correlation factor calculation method. The preliminary correlation factor of the data to be reviewed is compared with the historical correlation factor. If the difference is less than or equal to the correlation threshold, the review result is judged to be qualified. If the difference is greater than the correlation threshold, the review result is judged to be unqualified. The method for calculating the correlation factor includes: Normalize the common elements in each sample of the historical audit data so that the values ​​are between 0 and 1; After normalization, the standard deviation and mean of multiple elements in each sample are calculated. The preliminary correlation factor for each sample is then calculated using the following formula: ； in Indicates preliminary correlation factors, This represents the total number of elements in the sample. This represents the normalized value of the r-th element in the sample. This represents the average value of all elements in the normalized sample. This represents the standard deviation of each element in the normalized sample. Calculate the measure of central tendency for each initial correlation factor, as a historical correlation factor; The sampling classification method includes: The total number of samples and the non-compliant elements that appeared in the historical audit data were statistically analyzed. Choose any non-compliant element and calculate the proportion of the sample size containing the target non-compliant element to the total sample size. If the number of samples containing the target non-compliant element reaches the total number of elements in the sample, or if the number of samples containing the target non-compliant element accounts for more than 6.8% of the total number of samples, then the target non-compliant element is determined to be a sampled element; otherwise, the target non-compliant element is determined to be a non-sampled element.

2. The intelligent auditing system for enterprise management system certification process according to claim 1, characterized in that: The calibration factor calculation method can also be: In the historical audit data, the samples are grouped according to the type and number of non-compliant elements, so that the samples in the same group have exactly the same non-compliant elements. Select any group, compare the data before and after modification for all elements in the target group except for the unqualified elements, and mark the other elements whose data has changed as related elements of multiple unqualified elements; The sample modification calibration factor is set to a constant. Based on the modification change rate of non-conforming elements and the modification change rate of related elements, a calibration factor calculation model is established, specifically as follows: ； in The calibration factor representing the sample modification is a constant. and Represents the fitting coefficient. Indicates the total number of associated elements. This indicates the total number of non-compliant elements. This represents the modified value of the j-th non-compliant element. This represents the value of the j-th non-compliant element before modification. This represents the modified value of the i-th associated element. This represents the value of the i-th associated element before modification; The calibration factor calculation model is trained using multiple samples from the target group to obtain... and The numerical value is then used to input the data of the materials to be reviewed and the modification records into the model to obtain the modification calibration factor after this modification.

3. The intelligent auditing system for enterprise management system certification process according to claim 1, characterized in that: The method for calculating the association threshold includes: Calculate the differences between multiple preliminary correlation factors and historical correlation factors in historical audit data; Set a range coefficient so that more than 80% of the differences are less than the correlation threshold, and then obtain the correlation threshold.

Images