Life cycle concrete demand prediction method and device and storage medium

Abstract

The application provides a full-life-cycle commercial concrete demand prediction method and device and a storage medium, and belongs to the building information technology field.The method comprises the following steps: obtaining BIM information of a building project, and converting the BIM information into a fixed-dimension vector based on a first encoder; the first encoder is trained based on a positive sample pair and a negative sample pair BiLSTM encoder; the Euclidean distance between the fixed-dimension vector of a target building project and the fixed-dimension vector of a completed building project is determined, and a number of completed building projects with the smallest distance constitute a similar project set; the context vector of the similar project set is determined; the feature sequence corresponding to the target building project, the decoder state and the context vector are input into a decoder to obtain a predicted component relative construction time, and then the total commercial concrete demand is determined.The application can realize early prediction of the commercial concrete demand, so that the production and distribution of commercial concrete raw materials can be better arranged.

Description

Technical Field

[0001] This invention relates to the field of building information technology, specifically to a method, device, and storage medium for predicting the demand for commercial concrete throughout its entire life cycle. Background Technology

[0002] With the development of the construction industry, the commercial concrete (ready-mixed concrete) supply industry has also matured. To ensure sufficient and timely supply of ready-mixed concrete according to orders, suppliers need to be able to forecast the demand for ready-mixed concrete for a construction project in advance. This allows for advance planning of raw materials, labor, transportation, and delivery. Currently, the overall demand for ready-mixed concrete is mainly confirmed by contracts signed with suppliers before the start of construction projects. However, specific demand relies on advance booking by project owners, and sometimes can only be confirmed when placing an order. This poses certain obstacles for suppliers in arranging raw materials, labor, and transportation. Advance demand forecasting can help suppliers better plan production and delivery.

[0003] Currently, there is a lack of methods to predict the demand for ready-mixed concrete in advance, making it impossible to better arrange the production and distribution of raw materials for ready-mixed concrete. Summary of the Invention

[0004] In view of this, it is necessary to provide a method, device and storage medium for predicting the demand for ready-mixed concrete throughout its entire life cycle, so as to achieve the purpose of predicting the demand for ready-mixed concrete in advance, thereby better arranging the production and distribution of ready-mixed concrete raw materials.

[0005] To address the aforementioned problems, in a first aspect, the present invention provides a method for predicting the demand for ready-mixed concrete throughout its entire lifecycle, comprising:

[0006] The BIM information of the building project is acquired, and the BIM information is converted into a fixed-dimensional vector based on the first encoder; the first encoder is trained based on a BiLSTM encoder that is used to represent positive sample pairs with high building similarity and negative sample pairs with low building similarity.

[0007] Determine the Euclidean distance between the fixed-dimensional vector of the target building project and the fixed-dimensional vector of the completed building projects, and select the completed building projects with the smallest Euclidean distance to form a set of similar projects.

[0008] Determine the context vector of the set of similar items;

[0009] The feature sequence corresponding to the component sequence of the target building project, the decoder state, and the context vector are input into the decoder to obtain the predicted relative construction time of the components; the decoder is trained based on the mean square error between the predicted relative construction time of the components and the actual relative construction time of the components as the loss function.

[0010] The total demand for ready-mixed concrete is determined based on the predicted relative construction time of components, the start time of the target building project, and the volume of components.

[0011] In one possible implementation, the BIM information is converted into a fixed-dimensional vector based on a first encoder, including:

[0012] The component sequence of the BIM information is encoded based on the first encoder to obtain the hidden state of each component;

[0013] The hidden state is subjected to global average pooling to obtain a fixed-dimensional vector of the building project.

[0014] In one possible implementation, determining the context vector of the set of similar items includes:

[0015] By integrating the component sequences corresponding to the BIM information of the similar project set, and the relative time sequences between components, a feature sequence of completed building projects is obtained.

[0016] Encode the component sequence corresponding to the BIM information of the target building project to obtain the feature sequence of the target building project;

[0017] Based on the feature sequence of the target building project, determine the current state of the decoder;

[0018] Determine the current state of the decoder and the attention score between it and the feature sequence of the completed building project;

[0019] Based on the attention score, the attention-required parts of the feature sequence of the completed building project are weighted and summed to obtain the context vector.

[0020] In one possible implementation, the component sequences corresponding to the BIM information of the similar project set, and the relative time sequences between components, are fused to obtain the feature sequence of the completed building project, including:

[0021] A new sequence is obtained by integrating the component sequences corresponding to the BIM information of the similar project set and the relative time sequences between the components;

[0022] The new sequence is input into the second encoder to obtain the feature sequence of the completed building project; the second encoder is trained based on the Transformer encoder.

[0023] In one possible implementation, the component sequence corresponding to the BIM information of the target building project is encoded to obtain the feature sequence of the target building project, including:

[0024] The component sequence corresponding to the BIM information of the target building project is encoded based on the third encoder to obtain the feature sequence of the target building project; the third encoder is trained based on the Transformer encoder.

[0025] In one possible implementation, the total ready-mixed concrete requirement is determined based on the predicted relative construction time of the components, the commencement time of the target building project, and the volume of the components, including:

[0026] Based on the predicted relative construction time of each component and the start time of the target building project, determine the predicted absolute construction time of each component;

[0027] Based on the predicted absolute construction time of each component, select the target component sequence that meets the preset time requirements from the component sequence corresponding to the BIM information of the target building project;

[0028] Based on the target component sequence and the volume of the target components, the total commercial concrete requirement is determined.

[0029] In one possible implementation, a preset time requirement is set to satisfy:

[0030]

[0031]

[0032] in, t c This indicates the timestamp indicating the existence of predicted demand for ready-mixed concrete. t a This indicates the predicted absolute construction time for a component. t b This indicates the predicted absolute construction time of another component. t a-1 express t a The previous moment, t b-1 express t b The previous moment.

[0033] Secondly, the present invention also provides a device for predicting the demand for commercial concrete throughout its entire life cycle, comprising:

[0034] An information encoding module is used to acquire BIM information of a building project and convert the BIM information into a fixed-dimensional vector based on a first encoder; the first encoder is trained based on a BiLSTM encoder that represents positive sample pairs that represent high building similarity and negative sample pairs that represent low building similarity.

[0035] The similar project identification module is used to determine the Euclidean distance between the fixed-dimensional vector of the target building project and the fixed-dimensional vector of the completed building projects, and select the completed building projects with the smallest Euclidean distance to form a similar project set.

[0036] A context vector component module is used to determine the context vector of the set of similar items;

[0037] The construction time prediction module is used to input the feature sequence corresponding to the component sequence of the target building project, the decoder state, and the context vector into the decoder to obtain the predicted relative construction time of the components; the decoder is trained based on the mean square error between the predicted relative construction time of the components and the actual relative construction time of the components as a loss function.

[0038] The demand determination module is used to determine the total demand for commercial concrete based on the predicted relative construction time of components, the start time of the target building project, and the volume of components.

[0039] Thirdly, the present invention also provides an electronic device, including a memory and a processor, wherein,

[0040] The memory is used to store programs;

[0041] The processor, coupled to the memory, is used to execute the program stored in the memory to implement the steps of the full life-cycle commercial concrete demand forecasting method as described in any of the preceding claims.

[0042] Fourthly, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the full life-cycle commercial concrete demand forecasting method as described in any of the preceding claims.

[0043] The beneficial effects of adopting the above implementation method are as follows: The whole life cycle commercial concrete demand prediction method, device and storage medium provided by the present invention convert the BIM information of building projects into fixed-dimensional vectors through the first encoder; the first encoder is trained based on BiLSTM encoders that are positive sample pairs used to represent high building similarity and negative sample pairs used to represent low building similarity; building projects include unfinished building projects and completed building projects, and unfinished building projects are the target building projects. The encoder is trained using positive sample pairs and negative sample pairs, that is, the encoder is trained by contrastive learning.

[0044] The encoder obtained in this way can acquire a fixed-dimensional vector from the BIM information of the building project, and determine a set of similar building projects based on the fixed-dimensional vector, i.e., a set of similar projects. This allows the determination of a context vector, which contains contextual information about the input data. Based on the feature sequence corresponding to the component sequence of the target building project, the decoder state, and the context vector, the relative construction time between two components can be predicted. Combined with the start time and construction volume of the building project, the total demand for ready-mixed concrete can be determined. The entire process does not require prior contract confirmation with the ready-mixed concrete supplier, nor does it require confirmation only when placing an order. It can predict the total demand for ready-mixed concrete in advance based on the BIM information of the building project, thereby better arranging the production and distribution of ready-mixed concrete raw materials. Attached Figure Description

[0045] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 A flowchart illustrating an embodiment of the whole lifecycle commercial concrete demand forecasting method provided by the present invention;

[0047] Figure 2 A flowchart illustrating another embodiment of the whole life cycle commercial concrete demand forecasting method provided by the present invention;

[0048] Figure 3 A schematic block diagram of an embodiment of the whole life cycle commercial concrete demand prediction device provided by the present invention;

[0049] Figure 4 A schematic diagram of an embodiment of the electronic device provided by the present invention. Detailed Implementation

[0050] 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 a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0051] In the description of the embodiments of this application, unless otherwise stated, "a plurality of" means two or more.

[0052] In this embodiment of the invention, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, apparatus, product or device that includes a series of steps or modules is not necessarily limited to those steps or modules that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to such process, method, product or device.

[0053] The naming or numbering of steps in the embodiments of the present invention does not mean that the steps in the method flow must be executed in the time / logical order indicated by the naming or numbering. The execution order of the named or numbered process steps can be changed according to the technical purpose to be achieved, as long as the same or similar technical effect can be achieved.

[0054] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0055] This invention provides a method, device, and storage medium for predicting the demand for commercial concrete throughout its entire life cycle, which will be described below.

[0056] like Figure 1 As shown, this invention provides a method for predicting the demand for commercial concrete throughout its entire life cycle, including:

[0057] S101. Obtain BIM (Building Information Modeling) information of the building project, and convert the BIM information into a fixed-dimensional vector based on the first encoder; the first encoder is trained based on a BiLSTM (Bidirectional Long Short-Term Memory) encoder, which is used to represent positive sample pairs with high building similarity and negative sample pairs with low building similarity.

[0058] It is understandable that the construction projects in this step include target construction projects (i.e., construction projects that have not yet started) and completed construction projects. BIM information can be viewed as an ordered list (sequence) of concrete components for a construction project, including component ID, type, and volume. The order of the concrete components indicates the sequence in which the various components were constructed during the construction of the entire building. The types of components significantly distinguish the different types of components in the building.

[0059] High similarity means that the vector representations of buildings are similar in the vector space; low similarity means that the vectors of buildings are quite different, that is, the vector representations of buildings are far apart in the vector space.

[0060] S102. Determine the Euclidean distance between the fixed dimension vector of the target building project and the fixed dimension vector of the completed building projects, and select several completed building projects with the smallest Euclidean distance to form a set of similar projects.

[0061] It is understandable that the smaller the Euclidean distance, the closer the two vectors are, and the more similar the corresponding building projects. In this embodiment, "several" refers to at least two.

[0062] S103. Determine the context vector of the set of similar items.

[0063] Understandably, a context vector is a specific representation that contains contextual information about the input data (such as words or phrases).

[0064] S104. Input the feature sequence corresponding to the component sequence of the target building project, the decoder state, and the context vector into the decoder to obtain the predicted relative construction time of the components; the decoder is trained based on the mean square error between the predicted relative construction time of the components and the actual relative construction time of the components as the loss function.

[0065] Understandably, the decoder is trained using the mean squared error between the predicted relative construction time of components and the actual relative construction time of components in completed building projects as a loss function. The decoder state... The corresponding characteristics of the target building project at this time step and context vector The input is fed into the decoder to predict the relative construction time of the current component. and update the decoder state. :

[0066]

[0067] Ultimately, the predicted relative construction time series for each component of the target project is obtained. .

[0068] S105. Determine the total demand for ready-mixed concrete based on the predicted relative construction time of components, the start time of the target building project, and the volume of components.

[0069] Understandably, based on the predicted relative construction time of each component and the start time of the target building project, the timestamp of each component can be obtained, and then based on the timestamp of each component and its corresponding volume, the total demand for commercial concrete can be obtained.

[0070] The method provided by this invention can be applied to a terminal or a server, and is executed by an application on the terminal or server. The terminal can be a mobile phone or a computer, and the server can be an edge server or a cloud server.

[0071] The purpose of this invention is to dynamically predict the total ready-mixed concrete demand for a target building project (a project that has not yet started construction) based on BIM information of historically completed building projects and the construction times of each corresponding component. Knowing only the BIM information of the building project, the invention predicts the construction time of each concrete component in each BIM file of the target building project, thereby generating the corresponding ready-mixed concrete demand.

[0072] In some embodiments, converting the BIM information into a fixed-dimensional vector based on a first encoder includes:

[0073] The component sequence of the BIM information is encoded based on the first encoder to obtain the hidden state of each component;

[0074] The hidden state is subjected to global average pooling to obtain a fixed-dimensional vector of the building project.

[0075] Understandably, a trained BiLSTM encoder is used to encode the ordered list of concrete components. This represents the sequence of BIM information. The input is fed into the trained BiLSTM encoder, which, due to its bidirectional design, can capture the bidirectional dependencies between concrete members at different times in the time series. After inputting the sequence, the BiLSTM will output the hidden state of each member. :

[0076]

[0077] in, and These represent the hidden states of the forward and backward LSTMs at position i, respectively. Then, the hidden states are pooled using a global average pooling layer to obtain a fixed-dimensional representation vector of the building. :

[0078]

[0079] In some embodiments, determining the context vector of the set of similar items includes:

[0080] By integrating the component sequences corresponding to the BIM information of the similar project set, and the relative time sequences between components, a feature sequence of completed building projects is obtained.

[0081] Encode the component sequence corresponding to the BIM information of the target building project to obtain the feature sequence of the target building project;

[0082] Based on the feature sequence of the target building project, determine the current state of the decoder;

[0083] Determine the current state of the decoder and the attention score between it and the feature sequence of the completed building project;

[0084] Based on the attention score, the attention-required parts of the feature sequence of the completed building project are weighted and summed to obtain the context vector.

[0085] Understandably, at each time step (from 1 to m , m (This refers to the number of components in the target project). First, an attention score is calculated for each similar project. Calculate the current state of the decoder Similar projects Feature sequences Attention score:

[0086]

[0087] in W It is a learnable weight matrix. Through this attention mechanism, based on the components that need to be built in the current state, the model can automatically focus on similar construction patterns in historically similar projects.

[0088] Then, a context vector is generated to represent the weighted summation of the attention scores for the parts of similar projects that require attention:

[0089]

[0090] In some embodiments, the component sequences corresponding to the BIM information of the similar project set, and the relative time sequences between components, are fused to obtain the feature sequences of completed building projects, including:

[0091] A new sequence is obtained by integrating the component sequences corresponding to the BIM information of the similar project set and the relative time sequences between the components;

[0092] The new sequence is input into the second encoder to obtain the feature sequence of the completed building project; the second encoder is trained based on the Transformer encoder.

[0093] Understandably, this is achieved by fusing component sequences. and relative time series The features at the corresponding time steps are concatenated, and the new sequence is input into the encoder. The encoder uses a Transformer encoder and outputs a feature sequence of the same length. The characteristic representation of each component It not only characterizes the features of the component, but also incorporates its contextual information throughout the building project through the Transformer encoder.

[0094] In some embodiments, the component sequence corresponding to the BIM information of the target building project is encoded to obtain the feature sequence of the target building project, including:

[0095] The component sequence corresponding to the BIM information of the target building project is encoded based on the third encoder to obtain the feature sequence of the target building project; the third encoder is trained based on the Transformer encoder.

[0096] Understandably, the codes for the target building project are largely the same as those for historically similar projects; however, they lack a relative time series. (This is also the sequence that ultimately needs to be predicted), only the component sequence exists. This sequence is input to an encoder that also uses the Transformer encoder architecture. In this process, the feature sequence of the target project is obtained. ,in m This refers to the number of components in the target project.

[0097] In some embodiments, the total ready-mixed concrete requirement is determined based on the predicted relative construction time of the components, the commencement time of the target building project, and the volume of the components, including:

[0098] Based on the predicted relative construction time of each component and the start time of the target building project, determine the predicted absolute construction time of each component;

[0099] Based on the predicted absolute construction time of each component, select the target component sequence that meets the preset time requirements from the component sequence corresponding to the BIM information of the target building project;

[0100] Based on the target component sequence and the volume of the target components, the total commercial concrete requirement is determined.

[0101] It is understandable that the total demand for commercial concrete is obtained by multiplying each value in the target component sequence by the corresponding component volume and summing the products corresponding to each component in the target component sequence.

[0102] In some embodiments, a preset time requirement is provided to satisfy:

[0103]

[0104]

[0105] in, t c This indicates the timestamp indicating the existence of predicted demand for ready-mixed concrete. t a This indicates the predicted absolute construction time for a component. t b This indicates the predicted absolute construction time of another component. t a-1 express t a The previous moment, t b-1 express t b The previous moment.

[0106] It is understandable that to obtain a specific day (assuming the timestamp for each day is...), To predict the demand for ready-mixed concrete, it is only necessary to find the subsequence index in the timestamp sequence. and The above formula must be satisfied.

[0107] In some embodiments, the decoder is trained using the mean squared error between the predicted relative construction time of a component and the actual relative construction time of a component as a loss function.

[0108] Understandably, by minimizing the predicted time series... Compared with the real time series The model is trained using the error between the two values, with mean squared error (MSE) as the loss function, i.e.:

[0109]

[0110] in It is the target project number The actual relative construction time of each component.

[0111] In some embodiments, the method provided by the present invention may refer to Figure 2 As shown, inputs and outputs are represented by right-angled rectangles, and modules are represented by rounded rectangles.

[0112] A BIM information encoding module based on contrastive learning: This module encodes a building's BIM information into a vector. Similar buildings have similar vector representations in the vector space, while buildings with significant differences have more distant vector representations. To train the encoder, positive and negative sample pairs annotated with expert knowledge are used for contrastive learning training.

[0113] A BIM-based component construction time prediction module enhanced with prior knowledge: Based on a BIM information encoding module and a database of historical building projects, the module uses an encoder to encode the sequence of concrete components in the target building project's BIM, obtaining a feature sequence. An attention mechanism is then used to fuse feature information from similar building projects in the database, and a decoder is used to progressively predict the relative construction time of each component.

[0114] The ready-mixed concrete demand generation module based on component construction time calculates the absolute timestamp of each concrete component based on the previously obtained relative construction time, and calculates the ready-mixed concrete demand for that day according to the queried date.

[0115] To enhance the forecasting of ready-mixed concrete demand using BIM information, BIM information is viewed as an ordered list of concrete components for a building project, denoted as... Each component The order of concrete components, determined by their ID, type, and volume, indicates the sequence in which they were constructed during the overall building construction. The type of component significantly distinguishes the various components within the building, including but not limited to those shown in Table 1.

[0116] Table 1: Component Types and Corresponding Names

[0117]

[0118] The objective of this invention is to predict the estimated start time of construction for each component in a BIM information model (an ordered list of concrete components) of a construction project that has not yet commenced (i.e., the target construction project). To achieve this, data from historically completed construction projects is used for model training. This historical data includes not only the corresponding BIM information model but also the start time of construction for each component. After training with this historical data, it will be possible to predict the start time of construction for each component in a new construction project, and consequently, to predict the planned demand for concrete in each period.

[0119] The problem is formally defined as follows: Given an ordered list (in construction order) of concrete components representing a building information model. Predict timestamp sequences of the same length Each timestamp Represents components Construction time.

[0120] A BIM information encoding module based on contrastive learning is proposed to obtain the BIM of a new building project, encode its information, and perform vector comparison with historical completed projects, referencing those similar projects.

[0121] Given multiple building projects, each building project is represented by a BIM model as an ordered list of components sorted by construction time. Each component Represented as eigenvectors ,in It is a one-hot encoded vector of component type. It is the volume value of the component.

[0122] To obtain positive and negative sample pairs for contrastive learning, expert knowledge is used to label the data, dividing these buildings into several sets. Buildings within each set are very similar in terms of building type (e.g., hospitals) and size (e.g., the same number of floors), while buildings in different sets differ significantly. The goal is to extract a representation vector for each building through contrastive learning using BIM information, making buildings within the same set closer together in the embedding space, and buildings in different sets further apart.

[0123] A trained BiLSTM (Bidirectional Long Short-Term Memory) encoder is used to encode the ordered list of concrete components. This encodes the sequence representing BIM information. The input is fed into the trained BiLSTM encoder, which, due to its bidirectional design, can capture the bidirectional dependencies between concrete components at different times in the time series. After inputting the sequence, the BiLSTM encoder will output the hidden state of each component. :

[0124]

[0125] in, and These represent the hidden states of the forward and backward LSTMs at position i, respectively. Then, the hidden states are pooled using a global average pooling layer to obtain a fixed-dimensional representation vector of the building. :

[0126]

[0127] After obtaining the BIM encoding, we need to obtain positive and negative sample pairs for training. Positive sample pairs are two buildings selected from the same set of experts, indicating that the buildings are very similar; negative sample pairs are two buildings selected from different sets of experts. A contrastive loss function is used to make the representation vectors of positive sample pairs closer together and the representation vectors of negative sample pairs further apart. Let the number of positive and negative sample pairs both be N, then the loss function is as follows:

[0128] The contrastive learning loss function based on sample pairs can be defined as:

[0129]

[0130] in, It is the number of sample pairs. Indicates the first i The loss for a positive sample pair Indicates the first i The loss for each negative sample pair.

[0131] For positive sample pairs:

[0132]

[0133] in, and These are the representation vectors of the two samples in a positive sample pair. The goal of the loss function for positive sample pairs is to make the representation vectors of the positive sample pairs (two similar buildings) as close as possible, so the square of the Euclidean distance is used as the loss.

[0134] For negative sample pairs:

[0135]

[0136] in, and These are the representation vectors of the two samples in the negative sample pair. Margin is a predefined value representing the minimum allowed distance between negative sample pairs. The goal of the negative sample loss function is to make the representation vectors of negative sample pairs (two significantly different buildings) as far apart as possible, at least beyond the Margin value.

[0137] In this way, the contrastive learning loss function based on sample pairs can simultaneously perceive similar buildings and buildings with large differences, optimize the BiLSTM encoder, and learn an effective BIM information representation.

[0138] For the BIM of a new building, its component sequence is input into a trained model to obtain the representation vector of the entire building. This BIM representation vector will be used for subsequent component construction time prediction.

[0139] BIM-based component construction time prediction module enhanced with prior knowledge:

[0140] Based on the above description, an encoder for BIM information has been obtained, capable of obtaining a vector representation of a building from its BIM information (a sequence of concrete components). Similar buildings have similar vector representations. The following describes how to fuse information from similar buildings to enhance the prediction of construction time for individual concrete components of a target building.

[0141] Top-k similar building discovery:

[0142] Because a historical information database exists, containing BIM information for multiple completed projects, this information is encoded and converted into a vector representation. For the target building project, its BIM is encoded as a vector. Then, calculate Find the k projects with the smallest Euclidean distances to the BIM vectors of all completed projects in the database, and denote them as the set of similar projects. .

[0143] Historical similar project codes:

[0144] For each similar project Since it is a historically completed project, it not only contains BIM information (component sequence) (in It is a project i The number of components also corresponds to a construction time sequence. The original time sequence is a timestamp sequence, with each element representing the absolute construction time of the corresponding component. However, absolute timestamps are meaningless; a more meaningful representation is the relative construction time between components. For example, the foundation pad was built first, and the foundation slab was built two days later. Therefore, this invention transforms the original timestamp sequence into a relative time sequence. The relative time corresponding to each component This indicates how many days after the previous component was constructed. .

[0145] By fusing component sequences and relative time series The features at the corresponding time steps are concatenated, and the new sequence is input into the encoder. The encoder uses a Transformer encoder and outputs a feature sequence of the same length. The characteristic representation of each component It not only characterizes the features of the component, but also incorporates its contextual information throughout the building project through the Transformer encoder.

[0146] Target building project code:

[0147] The target building project's coding is largely the same as that of historically similar projects, except that it lacks a relative time series. (This is also the sequence that ultimately needs to be predicted), only the component sequence exists. This sequence is input to an encoder that also uses the Transformer encoder architecture. In this process, the feature sequence of the target project is obtained. ,inm This refers to the number of components in the target project.

[0148] Construction time prediction by incorporating attention from similar projects:

[0149] The relative construction time (relative to the previous component) of each component needs to be generated progressively. This is achieved using an RNN (Recurrent Neural Network) decoder implemented with an MLP (Multi-Layer Perceptron). At each time step, the input is the state and other features, and the output is the relative construction time of that component. Initial state... Set as a global feature for encoding the target item, which is the target item feature sequence. Average pooling results:

[0150]

[0151] At each time step (from 1 to m , m (This refers to the number of components in the target project). First, an attention score is calculated for each similar project. Calculate the current state of the decoder Similar projects Feature sequences Attention score:

[0152]

[0153] in W It is a learnable weight matrix. Through this attention mechanism, based on the components that need to be built in the current state, the model can automatically focus on similar construction patterns in historically similar projects.

[0154] Then, a context vector is generated to represent the weighted summation of the attention scores for the parts of similar projects that require attention:

[0155]

[0156] The decoder predicts the construction time of the component corresponding to this time step. The decoder state is then set. The corresponding characteristics of the target project at this time step and context vector The input is fed into the decoder to predict the relative construction time of the current component. and update the decoder state. :

[0157]

[0158] Ultimately, the predicted relative construction time series for each component of the target project is obtained. .

[0159] By minimizing the predicted time series Compared with the real time series The model is trained using the error between the two values, with mean squared error (MSE) as the loss function, i.e.:

[0160]

[0161] in It is the target project number The actual relative construction time of each component.

[0162] Module for generating ready-mixed concrete demand based on component construction time:

[0163] In the previous section, we used BIM prior knowledge to predict the construction time of each component of the target building project and obtained the sequence. abbreviated as Next, we need to analyze the relative time series. Convert to absolute timestamp sequence .

[0164] Let the start time of the target project be... So, the first i The estimated construction time (timestamp) for each component is:

[0165]

[0166] Having learned the first i The estimated construction time for each component, and for each component, a volume... To retrieve the timestamp of a specific day (assuming the days are timestamps) To predict the demand for ready-mixed concrete, it is only necessary to find the subsequence index in the timestamp sequence. and ,satisfy:

[0167]

[0168]

[0169] This allows us to calculate the total demand for commercial concrete for that day:

[0170]

[0171] The present invention has the following advantages:

[0172] This invention enables the prediction of ready-mixed concrete demand at different stages of a project, even before construction begins, using the project's BIM information (information on concrete components). Most existing solutions are based on time series forecasting or market research, and no research has yet explored BIM-based ready-mixed concrete demand forecasting. Furthermore, this invention expands the application of BIM information.

[0173] Key points of this invention:

[0174] BIM Information Encoding Module Based on Comparative Learning: This module can encode the BIM information of buildings. Similar building vectors are close in distance, while dissimilar building vectors are far apart.

[0175] BIM Prior Knowledge Enhanced Component Construction Time Prediction Module: This module can integrate construction progress information from similar building projects to assist the target project in predicting the construction time of each concrete component.

[0176] like Figure 3 As shown, the present invention also provides a whole life cycle commercial concrete demand prediction device 300, comprising:

[0177] The information encoding module 301 is used to acquire BIM information of the building project and convert the BIM information into a fixed-dimensional vector based on the first encoder; the first encoder is trained based on a BiLSTM encoder that is used to represent positive sample pairs with high building similarity and negative sample pairs with low building similarity.

[0178] The similar project determination module 302 is used to determine the Euclidean distance between the fixed dimension vector of the target building project and the fixed dimension vector of the completed building project, and select the few completed building projects with the smallest Euclidean distance to form a similar project set.

[0179] Context vector component module 303 is used to determine the context vector of the similar item set;

[0180] The construction time prediction module 304 is used to input the feature sequence corresponding to the component sequence of the target building project, the decoder state, and the context vector into the decoder to obtain the predicted relative construction time of the components; the decoder is trained based on the mean square error between the predicted relative construction time of the components and the actual relative construction time of the components as a loss function.

[0181] The demand determination module 305 is used to determine the total demand for commercial concrete based on the predicted relative construction time of components, the start time of the target building project, and the volume of components.

[0182] The full life cycle commercial concrete demand forecasting device provided in the above embodiments can realize the technical solutions described in the above full life cycle commercial concrete demand forecasting method embodiments. The specific implementation principles of each module or unit can be found in the corresponding content in the above full life cycle commercial concrete demand forecasting method embodiments, which will not be repeated here.

[0183] like Figure 4 As shown, the present invention also provides an electronic device 400. The electronic device 400 includes a processor 401, a memory 402, and a display 403. Figure 4 Only some components of the electronic device 400 are shown, but it should be understood that it is not required to implement all the components shown, and more or fewer components may be implemented instead.

[0184] In some embodiments, memory 402 may be an internal storage unit of electronic device 400, such as a hard disk or memory of electronic device 400. In other embodiments, memory 402 may also be an external storage device of electronic device 400, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc. equipped on electronic device 400.

[0185] Furthermore, the memory 402 may include both internal storage units of the electronic device 400 and external storage devices. The memory 402 is used to store application software and various types of data installed on the electronic device 400.

[0186] In some embodiments, processor 401 may be a central processing unit (CPU), microprocessor, or other data processing chip, used to run program code stored in memory 402 or process data, such as the whole life cycle commercial concrete demand forecasting method of the present invention.

[0187] In some embodiments, display 403 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. Display 403 is used to display information from electronic device 400 and to display a visual user interface. Components 401-403 of electronic device 400 communicate with each other via a system bus.

[0188] In some embodiments of the present invention, when the processor 401 executes the full lifecycle commercial concrete demand forecasting program in the memory 402, the following steps can be implemented:

[0189] The BIM information of the building project is acquired, and the BIM information is converted into a fixed-dimensional vector based on the first encoder; the first encoder is trained based on a BiLSTM encoder that is used to represent positive sample pairs with high building similarity and negative sample pairs with low building similarity.

[0190] Determine the Euclidean distance between the fixed-dimensional vector of the target building project and the fixed-dimensional vector of the completed building projects, and select the completed building projects with the smallest Euclidean distance to form a set of similar projects.

[0191] Determine the context vector of the set of similar items;

[0192] The feature sequence corresponding to the component sequence of the target building project, the decoder state, and the context vector are input into the decoder to obtain the predicted relative construction time of the components; the decoder is trained based on the mean square error between the predicted relative construction time of the components and the actual relative construction time of the components as the loss function.

[0193] The total demand for ready-mixed concrete is determined based on the predicted relative construction time of components, the start time of the target building project, and the volume of components.

[0194] It should be understood that when the processor 401 executes the full lifecycle commercial concrete demand forecasting program in the memory 402, in addition to the functions mentioned above, it can also perform other functions, as detailed in the description of the corresponding method embodiments above.

[0195] Furthermore, the embodiments of the present invention do not specifically limit the type of electronic device 400 mentioned. Electronic device 400 can be a mobile phone, tablet computer, personal digital assistant (PDA), wearable device, laptop computer, or other portable electronic device. Exemplary embodiments of portable electronic devices include, but are not limited to, portable electronic devices running iOS, Android, Microsoft, or other operating systems. The aforementioned portable electronic device can also be other portable electronic devices, such as a laptop computer with a touch-sensitive surface (e.g., a touch panel). It should also be understood that in some other embodiments of the present invention, electronic device 400 may not be a portable electronic device, but rather a desktop computer with a touch-sensitive surface (e.g., a touch panel).

[0196] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the full lifecycle commercial concrete demand forecasting method provided by the methods described above, the method comprising:

[0197] The BIM information of the building project is acquired, and the BIM information is converted into a fixed-dimensional vector based on the first encoder; the first encoder is trained based on a BiLSTM encoder that is used to represent positive sample pairs with high building similarity and negative sample pairs with low building similarity.

[0198] Determine the Euclidean distance between the fixed-dimensional vector of the target building project and the fixed-dimensional vector of the completed building projects, and select the completed building projects with the smallest Euclidean distance to form a set of similar projects.

[0199] Determine the context vector of the set of similar items;

[0200] The feature sequence corresponding to the component sequence of the target building project, the decoder state, and the context vector are input into the decoder to obtain the predicted relative construction time of the components; the decoder is trained based on the mean square error between the predicted relative construction time of the components and the actual relative construction time of the components as the loss function.

[0201] The total demand for ready-mixed concrete is determined based on the predicted relative construction time of components, the start time of the target building project, and the volume of components.

[0202] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0203] The above provides a detailed description of the method, apparatus, and storage medium for predicting the demand of ready-mixed concrete throughout its entire life cycle. Specific examples have been used to illustrate the principles and implementation methods of this invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this invention. At the same time, those skilled in the art will recognize that there will be changes in the specific implementation methods and application scope based on the ideas of this invention. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A full life cycle commercial concrete demand prediction method, characterized in that, The method comprises the following steps: obtaining BIM information of a building project and converting the BIM information into a fixed-dimensional vector based on a first encoder; the first encoder is trained based on a BiLSTM encoder for representing positive sample pairs with high building similarity and negative sample pairs with low building similarity; determining the Euclidean distance between the fixed-dimensional vector of the target building project and the fixed-dimensional vectors of the completed building projects, and selecting a number of completed building projects with the smallest Euclidean distance to form a similar project set; determining a context vector of the similar project set; inputting the feature sequence corresponding to the component sequence of the target building project, the decoder state and the context vector into a decoder to obtain a predicted component relative construction time, and the decoder is trained based on the mean square error between the predicted component relative construction time and the actual component relative construction time as a loss function; determining the total commercial concrete demand based on the predicted component relative construction time, the construction time of the target building project and the component volume; determining a context vector of the similar project set, comprising: fusing the component sequence corresponding to the BIM information of the similar project set and the relative time sequence between components to obtain a feature sequence of the completed building project; encoding the component sequence corresponding to the BIM information of the target building project to obtain a feature sequence of the target building project; determining the current state of the decoder based on the feature sequence of the target building project; determining the attention score between the current state of the decoder and the feature sequence of the completed building project; based on the attention score, performing weighted summation on the part of the feature sequence of the completed building project that needs to be focused to obtain a context vector; fusing the component sequence corresponding to the BIM information of the similar project set and the relative time sequence between components to obtain a feature sequence of the completed building project, comprising: fusing the component sequence corresponding to the BIM information of the similar project set and the relative time sequence between components to obtain a new sequence; inputting the new sequence into a second encoder to obtain a feature sequence of the completed building project; the second encoder is trained based on a Transformer encoder.

2. The full life cycle commercial concrete demand prediction method of claim 1, wherein, converting the BIM information into a fixed-dimensional vector based on a first encoder, comprising: encoding the component sequence of the BIM information based on the first encoder to obtain the hidden state of each component; performing global average pooling on the hidden state to obtain the fixed-dimensional vector of the building project.

3. The full life cycle commercial concrete demand prediction method of claim 1, wherein, encoding the component sequence corresponding to the BIM information of the target building project to obtain a feature sequence of the target building project, comprising: encoding the component sequence corresponding to the BIM information of the target building project based on a third encoder to obtain a feature sequence of the target building project; the third encoder is trained based on a Transformer encoder.

4. The full life cycle commercial concrete demand prediction method of claim 1, wherein, determining the total commercial concrete demand based on the predicted component relative construction time, the construction time of the target building project and the component volume, comprising: determine a predicted absolute construction time of each component based on the predicted relative construction time of each component and a start time of the target construction project; select a target component sequence that meets a preset time requirement from the component sequence corresponding to the BIM information of the target construction project based on the predicted absolute construction time of each component; determine a total commercial concrete demand based on the target component sequence and a volume of the target component.

5. The full life cycle commercial concrete demand prediction method of claim 4, wherein, The preset time requirement meets: wherein, t c a timestamp indicating a forecasted demand for ready-mix concrete, t a a forecasted absolute construction time for one component, t b a forecasted absolute construction time for another component, t a-1 a timestamp indicating a forecasted demand for ready-mix concrete, t a a last time of t b-1 a timestamp indicating a forecasted demand for ready-mix concrete, t b a last time of 6. A device for predicting the demand for ready-mix concrete throughout its life cycle, characterized by The method comprises the following steps: an information encoding module, configured to obtain BIM information of a construction project and convert the BIM information into a fixed-dimensional vector based on a first encoder; the first encoder is trained based on a BiLSTM encoder for positive sample pairs representing high building similarity and negative sample pairs representing low building similarity; a similar project determination module, configured to determine the Euclidean distance between the fixed-dimensional vector of the target construction project and the fixed-dimensional vector of a completed construction project, and select a number of completed construction projects with the smallest Euclidean distance to form a similar project set; a context vector component module, configured to determine a context vector of the similar project set; a construction time prediction module, configured to input a feature sequence corresponding to a component sequence of the target construction project, a decoder state and the context vector into a decoder to obtain a predicted relative construction time of the component; the decoder is trained based on the mean square error between the predicted relative construction time of the component and the actual relative construction time of the component as a loss function; a demand determination module, configured to determine a total commercial concrete demand based on the predicted relative construction time of the component, the start time of the target construction project and the volume of the component; determining the context vector of the similar project set comprises: fusing the component sequence corresponding to the BIM information of the similar project set and the relative time sequence between components to obtain a feature sequence of the completed construction project; encoding the component sequence corresponding to the BIM information of the target construction project to obtain a feature sequence of the target construction project; determining a current state of the decoder based on the feature sequence of the target construction project; determining an attention score between the current state of the decoder and the feature sequence of the completed construction project; based on the attention score, performing weighted summation on the part of the feature sequence of the completed construction project that needs to be focused to obtain a context vector; fusing the component sequence corresponding to the BIM information of the similar project set and the relative time sequence between components to obtain a feature sequence of the completed construction project comprises: fusing the component sequence corresponding to the BIM information of the similar project set and the relative time sequence between components to obtain a new sequence; inputting the new sequence into a second encoder to obtain a feature sequence of the completed construction project; the second encoder is trained based on a Transformer encoder.

7. An electronic device, comprising: comprise a memory and a processor, wherein the memory is configured to store a program; the processor is coupled to the memory and is configured to execute the program stored in the memory to implement the steps of the full-life-cycle commercial concrete demand prediction method according to any one of claims 1 to 5.

8. A non-transitory computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program, when executed by a processor, implements the steps of the full life cycle commercial concrete demand prediction method as claimed in any one of claims 1 to 5.

Images