Engineering audit management integrated system and method thereof

Abstract

The invention relates to the technical field of audit management, and discloses an engineering audit management integration system and a method thereof, and the engineering audit management integration method comprises the following steps: collecting engineering information; generating plane data based on the engineering information; inputting the plane data into a symbiotic network model, wherein the symbiotic network model outputs a result representing the occurrence of interaction risks of all projects; inputting the plane data into a risk model, wherein the risk model outputs a result representing audit risk occurrence of each project; according to the invention, engineering information is collected regularly, and mutual influence between each project and sub-projects thereof is considered in multiple aspects, so that compared with auditing risk determination of a single project, the system can analyze the mutual relation between the projects through the symbiotic network to improve the accuracy and coverage of risk assessment.

Description

Technical Field

[0001] The present invention relates to the field of audit management technology, and more specifically, to an engineering audit management integrated system and method thereof. Background Art

[0002] Against the backdrop of rapid global economic development, the complexity and scale of power engineering projects are increasing, which has brought new challenges to project audits. Traditional audit methods usually rely on post-analysis, lack of real-time and foresight, and are difficult to effectively identify and manage risks in complex and changing project environments. Deep learning technology is used to integrate information from multiple data sources such as finance, progress, resources, and market environment, and then dynamically monitor audit risks in project execution. However, in the actual power engineering construction process, construction will be carried out in different locations according to actual construction needs. Power engineering projects in mountainous areas may be affected by the terrain. Project risk assessment may only take into account the risks of the current power engineering project, but other projects implemented in the same area may affect the terrain and affect the current power engineering project, resulting in inaccurate project risk assessment. Summary of the invention

[0003] The purpose of the present invention is to provide an engineering audit management integrated system and method in order to solve the above problems.

[0004] The present invention provides an engineering audit management integrated method, comprising the following steps:

[0005] Step 100, collecting engineering information, which is collected once at each time sampling point, including project information, sub-project information and location information;

[0006] Step 200, generating plane data based on engineering information, including facets for storing data, a serial data stream from one facet index to one data object, the plane data also including a relationship list for storing association relationships between data objects indexed by the facets, the element value in the i-th row and j-th column of the relationship list representing a quantitative value of the association relationship between the data objects indexed by the i-th and j-th facets;

[0007] The association relationships include high and low, corresponding to two quantitative values;

[0008] Data objects include items, subitems, and locations;

[0009] The serial data stream of a data object includes n data units ordered by time, and the tth data unit represents the engineering information collected at the tth moment of the data object;

[0010] Step 300, inputting the plane data into the symbiotic network model, the symbiotic network model comprising a first backbone network, the first backbone network comprising a first encoder, a second encoder and a first FC layer, wherein the serial data stream is input into the first encoder, the first encoder outputs a first identification feature to the second encoder, the second encoder also inputs a relationship list, the second encoder outputs a second identification feature to the first FC layer and a feature fusion layer, and the first FC layer outputs a result indicating the occurrence of all project interaction risks;

[0011] Step 400, input the plane data into the risk model, the risk model includes a second backbone network, the second backbone network includes a third encoder, a feature fusion layer, a fifth encoder and a second FC layer, the serial data stream is input into the third encoder, the third encoder outputs the third identification feature to the feature fusion layer, the feature fusion layer outputs the fourth combination feature to the fifth encoder, the fifth encoder also inputs the relationship list, the fifth encoder outputs the fifth identification feature to the second FC layer, the second FC layer outputs the results indicating the occurrence of audit risks for each project, the audit risks include financial risks, schedule risks, resource risks and environmental risks.

[0012] Furthermore, the time sampling point refers to the moment when data is collected at a predetermined time interval; the default value of the predetermined time interval is 24 hours.

[0013] Further, the project information includes financial data, schedule data, resource data, environmental data and topographic data of the project;

[0014] Subproject information includes the project's financial data, schedule data, resource data, environmental data, and topographic data;

[0015] The location information includes an overhead image of the location.

[0016] Furthermore, data objects with a high association relationship are defined as: an item and the locations it includes, an item and its adjacent items, an item and its subordinate sub-items, a sub-item and the locations it includes, and a location and the locations to which it is connected.

[0017] Furthermore, the engineering information is input into the model after feature engineering. The engineering information includes image modalities. For data of various modalities, corresponding feature engineering methods are used to encode and generate features, and then they are spliced. The information of the image modality is used as a feature engineering method through the CNN algorithm.

[0018] Furthermore, the output of the first FC layer is an interaction risk vector, which represents the probability of occurrence of interaction risks of all projects. A probability greater than 0.5 indicates that there is a risk, otherwise it indicates that there is no risk.

[0019] Furthermore, the output of the second FC layer is an audit risk vector. The wth component of the audit risk vector represents the audit risk of the wth project. The audit risk includes the probability of occurrence of financial risk, schedule risk, resource risk and environmental risk. A probability greater than 0.5 indicates risk.

[0020] Furthermore, the loss function calculation method of the symbiotic network model is:

[0021] ;

[0022] in represents the symbiotic network loss value, represents the total number of symbiotic samples, It is The true symbiosis classification label of the symbiosis sample has a value of 0 or 1, where 1 indicates that the symbiosis sample has an interaction risk, and 0 indicates that the symbiosis sample does not have an interaction risk. The first The probability of the interaction risk of the symbiotic samples, Represents the logarithmic function with the natural constant e as the base.

[0023] Furthermore, the loss function calculation method of the risk model is:

[0024] ;

[0025] in, represents the risk loss value, is the number of risk samples, is the number of risk categories. There are four risk categories, including financial risk, schedule risk, resource risk and environmental risk. The risk sample represents the actual risk label of each risk category for all projects in a time period. The value is 0 or 1. 1 means that the project has the risk of this risk category, and 0 means that the project has not the risk of this risk category. Indicates The risk sample The actual risk label of the risk class, represents the risk model prediction The risk sample The probability of occurrence of such risk, Represents the logarithmic function with the natural constant e as the base.

[0026] The present invention provides an engineering audit management integration system, which is used to store computer-readable instructions. When the computer-readable instructions are read, the aforementioned engineering audit management integration method can be executed.

[0027] The beneficial effect of the present invention is that the present invention collects engineering information at regular intervals and considers the mutual influence between each project and its sub-projects in many aspects. Compared with the audit risk determination of a single project, the system can analyze the relationship between projects through a symbiotic network to improve the accuracy and coverage of risk assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 It is a flow chart of an engineering audit management integrated method of the present invention;

[0029] Figure 2 This is Example 1 of the desensitized data sample of the present invention;

[0030] Figure 3 This is Example 2 of the desensitized data sample of the present invention;

[0031] Figure 4 It is a sub-item sample of the desensitized data sample of the present invention;

[0032] Figure 5 It is a location sample of the desensitized data sample of the present invention;

[0033] Figure 6 It is a result diagram of training the symbiotic network model and risk model of the present invention. DETAILED DESCRIPTION

[0034] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that the discussion of these embodiments is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the contents of this specification. Each example may omit, replace or add various processes or components as needed. In addition, the features described relative to some examples may also be combined in other examples.

[0035] At least one embodiment of the present invention discloses an integrated method for project audit management, such as Figure 1 As shown, the following steps are included:

[0036] Step 100, collecting engineering information, which is collected once at each time sampling point, including project information, sub-project information and location information;

[0037] Project information includes financial data (budget, expenditure, cash flow), progress data (project timetable, key progress), resource data (planned personnel, equipment, materials, actual use of materials), environmental data (raw material prices, weather conditions) and terrain data (terrain changes, soil characteristics, geological data, expected impact on the surrounding environment);

[0038] Subproject information includes financial data, progress data, resource data, environmental data and terrain data of the subproject;

[0039] Location information includes an overhead image of the location, altitude, and geological type;

[0040] In one embodiment of the present invention, the entire monitoring area is divided into uniform grids, each grid representing a location;

[0041] The time sampling point refers to the moment when data is collected at a predetermined time interval;

[0042] In one embodiment of the present invention, the default value of the predetermined time interval is 24 hours.

[0043] Step 200, generating plane data based on engineering information, including facets for storing data, a serial data stream from one facet index to one data object, the plane data also including a relationship list for storing association relationships between data objects indexed by the facets, the element value in the i-th row and j-th column of the relationship list representing a quantitative value of the association relationship between the data objects indexed by the i-th and j-th facets;

[0044] The association relationship includes high and low, corresponding to two quantitative values ​​respectively.

[0045] Data objects include items, subitems, and locations;

[0046] Data objects with high association relationships are defined as: items and the locations they include, items and their adjacent items, items and their attached sub-items, sub-items and the locations they include, and locations and the locations they are connected to;

[0047] The serial data stream of a data object includes n data units sorted by time, and the tth data unit represents the engineering information collected at the tth moment of the data object.

[0048] In one embodiment of the present invention, the engineering information is input into the model after feature engineering;

[0049] Engineering information includes image modalities. For data of various modalities, corresponding feature engineering methods are used to encode and generate features, and then splice them;

[0050] The information of image modality is used as a feature engineering method through CNN algorithm.

[0051] Step 300, inputting the plane data into the symbiotic network model, the symbiotic network model comprising a first backbone network, the first backbone network comprising a first encoder, a second encoder and a first FC (fully connected layer), wherein the serial data stream is input into the first encoder, the first encoder outputs a first identification feature to the second encoder, the second encoder also inputs a relationship list, the second encoder outputs a second identification feature to the first FC layer and a feature fusion layer, and the first FC layer outputs a result indicating the occurrence of all project interaction risks;

[0052] In one embodiment of the present invention, the formula of the first encoder is as follows:

[0053]

[0054] in Represents the plane data The tth first identification feature of the serial data stream corresponding to the facet, Indicates The t-1th first identification feature of the serial data stream corresponding to the facet, Indicates The tth data unit of the serial data stream of bins, and are the first and second serial recognition weight matrices, is the serial identification bias parameter, yes Activation function.

[0055] In one embodiment of the present invention, the formula of the second encoder is as follows:

[0056]

[0057]

[0058] in Represents the first layer The second aggregate representation of facets is, and Represents the first layer and the first layer The second facet feature of the facets, Representation and Facets A collection of facets with high association relationships; Represents the universal quantifier symbol; represents the first layer of the l-1th layer The second facet feature of the facets, , E represents the total number of layers, when l=1 , ,in and Respectively represent and The facet index of the object The first identifying feature, Indicates the total number of data items in the serial data stream; when l=E Equal to the surface element The second identification feature, represents the aggregation function of layer l, such as average or maximum pooling function, represents the first plane recognition weight matrix of the lth layer, σ represents the Sigmoid function, Represents the concatenation function.

[0059] In one embodiment of the present invention, the formula of the first FC layer is as follows:

[0060]

[0061] in Represents the interaction risk vector, which represents the probability of occurrence of interaction risks of all projects. A probability greater than 0.5 indicates that there is a risk, otherwise it indicates that there is no risk. Represents the plane data The second identification feature of the face element, E represents the total number of layers, Represents the collection of cells of all plane data, and FC means fully connected.

[0062] In one embodiment of the present invention, the loss function calculation method of the symbiotic network model is:

[0063]

[0064] in represents the symbiotic network loss value, represents the total number of symbiotic samples, It is The true symbiosis classification label of the symbiosis sample has a value of 0 or 1, where 1 indicates that the symbiosis sample has an interaction risk, and 0 indicates that the symbiosis sample does not have an interaction risk. The first The probability of the interaction risk of the symbiotic samples, Represents the logarithmic function with the natural constant e as the base.

[0065] Step 400, input the plane data into the risk model, the risk model includes a second backbone network, the second backbone network includes a third encoder, a feature fusion layer, a fifth encoder and a second FC layer, input the serial data stream into the third encoder, the third encoder outputs a third identification feature to the feature fusion layer, the feature fusion layer outputs a fourth combination feature to the fifth encoder, the fifth encoder also inputs a relationship list, the fifth encoder outputs a fifth identification feature to the second FC layer, the second FC layer outputs a result indicating the occurrence of audit risks for each project, the audit risks include financial risks, schedule risks, resource risks and environmental risks;

[0066] In one embodiment of the present invention, the formula of the third encoder is as follows:

[0067]

[0068]

[0069]

[0070] in Represents the plane data The first bin of the serial data stream corresponds to A third identifying feature, Indicates The first bin of the serial data stream corresponds to A third identifying feature, and Respectively represent The first and second intermediate features of the facets, Indicates The first of the serial data streams of bins data units, and are the third and fourth serial recognition weight matrices, is the second serial identification bias parameter, is the hyperbolic tangent function, represents the dot product, Represents the Sigmoid function.

[0071] In one embodiment of the present invention, the formula of the feature fusion layer is as follows:

[0072]

[0073] in, Represents the plane data The fourth combination feature corresponding to the face element is, Represents the plane data The second identification feature of the face element, E represents the total number of layers, Represents the plane data The first bin of the serial data stream corresponds to A third identifying feature, represents the total number of data items in the serial data stream, Represents the feature fusion function (splicing function).

[0074] In one embodiment of the present invention, the formula of the fifth encoder is as follows:

[0075]

[0076] in Represents the first layer The fifth facet feature of the facets is Representation and Facets A collection of facets with high associations. Indicates The second plane identification weight matrix of the layer, It is Layer Opposite element The attention weight is calculated by the following formula:

[0077]

[0078] in is a learnable attention vector, represents the vector concatenation operation, represents the third plane recognition weight matrix of layer l, represents transpose;

[0079] and Respectively represent the first The fifth facet feature of the facets is , F represents the total number of layers, when l=1 , , hour is equal to the fifth identification feature of the face element v; and Respectively represent and The fourth combined feature of the object indexed by the facets.

[0080] In one embodiment of the present invention, the formula of the second FC layer is as follows:

[0081]

[0082] in Represents the audit risk vector. The i-th component of the audit risk vector represents the audit risk of the i-th project. Audit risk includes the probability of occurrence of financial risk, schedule risk, resource risk and environmental risk. A probability greater than 0.5 indicates risk. Represents the plane data The fifth identification feature of the facets, Represents the set of facets of all plane data, and FC means fully connected.

[0083] In one embodiment of the present invention, the loss function calculation method of the risk model is:

[0084]

[0085] in, represents the risk loss value, is the number of risk samples, is the number of risk categories. There are four risk categories, including financial risk, schedule risk, resource risk and environmental risk. The risk sample represents the actual risk label of each risk category for all projects in a time period. The value is 0 or 1. 1 means that the project has the risk of this risk category, and 0 means that the project has not the risk of this risk category. Indicates The risk sample The actual risk label of the risk class, represents the risk model prediction The risk sample The probability of occurrence of such risk, Represents the logarithmic function with the natural constant e as the base.

[0086] The following is an example of a specific application scenario of the above method:

[0087] This study selected a large-scale power engineering construction project in China as the research object; the project is located in the mountainous area of ​​southwest China, with complex topography and changeable climatic conditions. The project is large-scale and covers multiple sub-projects, such as transmission line erection and substation construction. Due to its special geographical location, the project faces many uncertainties during the construction process, such as geological disasters and weather changes, which pose challenges to the smooth progress of the project.

[0088] The data comes from various information collected during the project implementation, mainly including but not limited to

[0089] Project information: including financial data (e.g., budget, expenditure, cash flow), schedule data (e.g., project timeline, key milestones), resource data (e.g., staffing, equipment usage, material consumption), environmental data (e.g., raw material market prices, weather conditions), and terrain data (e.g., terrain changes, soil properties, geological data);

[0090] Sub-project information: similar to project information, but focusing on the specific details of each sub-project;

[0091] Location information: including the project's geographical location coordinates and overhead images, which are used to assist in analyzing the impact of the project's geographical location on construction;

[0092] Data preprocessing

[0093] To ensure the effectiveness and accuracy of the model, the data preprocessing stage includes the following steps:

[0094] Data cleaning: remove outliers and missing values ​​to ensure data quality;

[0095] Feature engineering: Convert data of different modalities (such as text, numbers, and images) accordingly. For example, use a convolutional neural network (CNN) algorithm to process image data and convert it into a feature vector that can be used as model input.

[0096] Standardization: Standardize or normalize numerical data to facilitate model training;

[0097] Constructing plane data: constructing plane data based on project information and location information, including facets for storing data, lists of associations between data objects indexed by facets, etc.

[0098] Data Annotation

[0099] Interaction risk annotation: Based on historical data and expert knowledge, the data is annotated to mark items with interaction risks.

[0100] Audit risk marking: Mark projects where interactive risks occur, including financial risks, schedule risks, resource risks and environmental risks.

[0101] A set of anonymized data samples and their sub-items and location samples are as follows: Figure 2 , Figure 3 , Figure 4 , Figure 5 As shown;

[0102] Model Training

[0103] Symbiotic network model training: The constructed plane data is input into the first backbone network of the symbiotic network model, features are extracted through the first encoder and the second encoder, the interaction risk is calculated using the first fully connected layer, and finally the model parameters are adjusted by minimizing the symbiotic network loss function until the model converges.

[0104] Risk model training: Similarly, the plane data is input into the second backbone network of the risk model. The features are gradually extracted and fused through the third encoder, feature fusion layer, fifth encoder and second fully connected layer. The audit risk of each project is calculated, and the model performance is optimized by minimizing the loss function of the risk model.

[0105] like Figure 6 The figure shows the result of training the symbiotic network model and the risk model through a set of data samples. Loss 1 is the loss value of the symbiotic network model training, and loss 2 is the loss value of the risk model. The symbiotic network model is a binary loss and can converge to a lower loss value faster than the risk model, but the risk model can still converge to a loss value of 0.18 and can be applied to risk auditing.

[0106] Model checking

[0107] After the model training is completed, it needs to be tested to verify its effectiveness and reliability; specific measures include:

[0108] Internal validation: Use the cross-validation method to use part of the training set as a validation set to evaluate the performance of the model on unseen data;

[0109] External validation: Select data from different time periods or different geographical regions as the training set as the test set to further verify the generalization ability of the model;

[0110] Performance evaluation: By comparing the model prediction results with the actual risk events, calculating indicators such as accuracy, recall rate, and F1 score, the performance of the model is comprehensively evaluated;

[0111] Case analysis: Select several typical engineering project cases, conduct in-depth analysis on the consistency between the model prediction results and the actual situation, and explore the applicability of the model under specific conditions.

[0112] In at least one embodiment of the present invention, there is provided an engineering audit management integration system for storing computer-readable instructions, which can execute the aforementioned engineering audit management integration method when the computer-readable instructions are read.

[0113] The above describes an embodiment of the present invention, but this embodiment is not limited to the above-mentioned specific implementation mode. The above-mentioned specific implementation mode is merely illustrative and not restrictive. Under the guidance of this embodiment, ordinary technicians in this field can also make more forms of equivalent embodiments, all of which are within the protection of this embodiment.

Claims

1. An integrated method for engineering audit management, characterized in that: include: Step 100, collecting engineering information, which is collected once at each time sampling point, including project information, sub-project information and location information; Step 200, generating plane data based on engineering information, including facets for storing data, a serial data stream from one facet index to one data object, the plane data also including a relationship list for storing association relationships between data objects indexed by the facets, the element value in the i-th row and j-th column of the relationship list representing a quantitative value of the association relationship between the data objects indexed by the i-th and j-th facets; The association relationships include high and low, corresponding to two quantitative values; Data objects include items, subitems, and locations; The serial data stream of a data object includes n data units ordered by time, and the tth data unit represents the engineering information collected at the tth moment of the data object; Step 300, inputting the plane data into the symbiotic network model, the symbiotic network model comprising a first backbone network, the first backbone network comprising a first encoder, a second encoder and a first FC layer, wherein the serial data stream is input into the first encoder, the first encoder outputs a first identification feature to the second encoder, the second encoder also inputs a relationship list, the second encoder outputs a second identification feature to the first FC layer and a feature fusion layer, and the first FC layer outputs a result indicating the occurrence of all project interaction risks; Step 400, input the plane data into the risk model, the risk model includes a second backbone network, the second backbone network includes a third encoder, a feature fusion layer, a fifth encoder and a second FC layer, the serial data stream is input into the third encoder, the third encoder outputs the third identification feature to the feature fusion layer, the feature fusion layer outputs the fourth combination feature to the fifth encoder, the fifth encoder also inputs the relationship list, the fifth encoder outputs the fifth identification feature to the second FC layer, the second FC layer outputs the results indicating the occurrence of audit risks for each project, the audit risks include financial risks, schedule risks, resource risks and environmental risks.

2. According to claim 1, a method for integrated engineering audit management is characterized in that: The time sampling point refers to the moment when data is collected at a predetermined time interval; the default value of the predetermined time interval is 24 hours.

3. The engineering audit management integrated method according to claim 1 is characterized in that: Project information includes the project's financial data, schedule data, resource data, environmental data, and topographic data; Subproject information includes the project's financial data, schedule data, resource data, environmental data, and topographic data; The location information includes an overhead image of the location.

4. The engineering audit management integrated method according to claim 1 is characterized in that: Data objects with high association relationships are defined as: an item and the locations it includes, an item and its adjacent items, an item and its attached sub-items, a sub-item and the locations it includes, and a location and the locations it is connected to.

5. The engineering audit management integrated method according to claim 1 is characterized in that: The engineering information is input into the model after feature engineering. The engineering information includes image modality. For data of various modalities, the corresponding feature engineering method is used to encode and generate features, and then spliced. The information of image modality is used as a feature engineering method through the CNN algorithm.

6. The engineering audit management integrated method according to claim 1 is characterized in that: The output of the first FC layer is an interaction risk vector, which represents the probability of occurrence of interaction risks of all projects. A probability greater than 0.5 indicates that there is a risk, otherwise it indicates that there is no risk.

7. The engineering audit management integrated method according to claim 1 is characterized in that: The output of the second FC layer is an audit risk vector. The wth component of the audit risk vector represents the audit risk of the wth project. Audit risk includes the probability of occurrence of financial risk, schedule risk, resource risk and environmental risk. A probability greater than 0.5 indicates risk.

8. The engineering audit management integrated method according to claim 1 is characterized in that: The loss function calculation method of the symbiotic network model is: ； in represents the symbiotic network loss value, represents the total number of symbiotic samples, It is The true symbiosis classification label of the symbiosis sample has a value of 0 or 1, where 1 indicates that the symbiosis sample has an interaction risk, and 0 indicates that the symbiosis sample does not have an interaction risk. is the first predicted by the symbiotic network model. The probability of the interaction risk of the symbiotic samples, Represents the logarithmic function with the natural constant e as the base.

9. The engineering audit management integrated method according to claim 1, characterized in that: The loss function calculation method of the risk model is: ； in, represents the risk loss value, is the number of risk samples, is the number of risk categories. There are four risk categories, including financial risk, schedule risk, resource risk and environmental risk. The risk sample represents the actual risk label of each risk category for all projects in a time period. The value is 0 or 1. 1 means that the project has the risk of this risk category, and 0 means that the project has not the risk of this risk category. Indicates The risk sample The actual risk label of the risk class, represents the risk model prediction The risk sample The probability of occurrence of such risk, Represents the logarithmic function with the natural constant e as the base.

10. An integrated system for engineering audit management, characterized in that: It is used to store computer-readable instructions, which, when read, can execute an engineering audit management integration method as described in any one of claims 1-9.

Images