Construction supervision instruction evaluation optimization method and system

Abstract

The application discloses a construction supervision instruction evaluation optimization method and system, and relates to the technical field of construction management, which comprises the following steps: collecting intelligent safety helmet inertial measurement unit and ultra-wideband positioning data in real time and obtaining supervision instruction texts; generating a group inertial motion feature vector containing an inferred working procedure, confidence and personnel density by using a frequency band energy decoupling method; analyzing the instruction texts to extract a target working procedure and a construction section, and calculating an instruction-site matching degree; calculating an execution damping coefficient according to the working procedure switching difficulty, the working surface inertial momentum and historical performance data; and generating an optimization scheme based on the combined results of the matching degree and the damping coefficient. The application realizes the dynamic adaptation of the supervision instruction and the real-time working condition of the site, and effectively improves the instruction execution efficiency.

Description

Technical Field

[0001] This invention relates to the field of construction management technology, specifically to a method and system for evaluating and optimizing construction supervision instructions. Background Technology

[0002] Currently, in the construction supervision process, the issuance of supervision instructions mainly relies on the experience and judgment of management personnel or comparison with static progress plans. However, construction sites are highly dynamic and complex, and existing solutions lack quantitative means to perceive the real-time status of group operations on site, making it impossible to accurately know the current actual work process type, work intensity, and personnel distribution of workers. This often leads to a mismatch between the timing of the issuance of supervision instructions and the actual construction rhythm on site. For example, issuing reinforcement rectification instructions during continuous pouring or issuing emergency stop orders during shift change off-peak periods results in situations where there are no corresponding trades available for deployment or no one responds, seriously affecting the execution efficiency of supervision instructions and the effectiveness of project management. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the prior art. This invention provides a method and system for evaluating and optimizing construction supervision instructions, which realizes real-time quantitative perception of the group operation status at the construction site and evaluation of instruction adaptability, thereby generating optimized supervision instructions.

[0004] This invention provides a method for evaluating and optimizing construction supervision instructions, the method comprising: Real-time acquisition of inertial measurement unit data and ultra-wideband positioning data from each smart safety helmet at the construction site, while simultaneously obtaining the text of the supervisor's instructions to be evaluated; Based on the data from the inertial measurement unit, the frequency band energy decoupling method is used to decompose the group motion spectrum of each smart safety helmet into low-frequency energy proportion, mid-frequency energy proportion, and high-frequency energy proportion, combined with pre-established standard procedures. A frequency band template matrix is ​​used to generate a group inertial motion feature vector for each construction grid. The group inertial motion feature vector includes at least the inferred construction procedure for each construction grid, the confidence level corresponding to the inferred construction procedure, and the personnel density. The supervision instruction text is structured and parsed to extract the target process and target construction section. The target construction section is mapped to the corresponding construction grid set. Based on the group inertial motion feature vector, the consistency degree between the actual dominant process and the target process within the construction grid set is calculated. Combined with the personnel density of the construction grid set, instructions are generated. On-site matching degree; The execution damping coefficient is calculated based on the difficulty of switching between the inferred construction process and the target process, the current working surface inertial momentum of the construction grid set, and the historical instruction performance data of the corresponding construction unit. According to the instructions The combination of the on-site matching degree and the execution damping coefficient generates an optimized scheme for the supervision instructions and outputs the optimized supervision instructions.

[0005] Furthermore, based on the inertial measurement unit data, the method of using frequency band energy decoupling to decompose the group motion spectrum of each smart safety helmet into low-frequency energy proportion, mid-frequency energy proportion, and high-frequency energy proportion includes: The composite acceleration modulus of all smart safety helmets in each construction grid is processed by short-time frame division. Fourier transform is performed on each frame to obtain the power spectral density. The power spectral density is decomposed into low-frequency energy, mid-frequency energy and high-frequency energy according to the preset frequency band range. The low-frequency band corresponds to slow personnel movement and work surface transfer, the mid-frequency band corresponds to rhythmic limb movements, and the high-frequency band corresponds to tool impact and high-frequency vibration. The proportions of the low-frequency energy, mid-frequency energy, and high-frequency energy in the total energy are calculated to obtain the proportions of low-frequency energy, mid-frequency energy, and high-frequency energy.

[0006] Furthermore, the combination of pre-established standard procedures The frequency band template matrix generates the group inertial motion feature vectors for each construction grid, including: Pre-determine the expected proportions of low-frequency bands, mid-frequency bands, and high-frequency bands for different construction procedures to form standard procedures. Frequency band template matrix; The real-time feature triplet, composed of the low-frequency energy proportion, mid-frequency energy proportion, and high-frequency energy proportion of each construction grid obtained in real time, is compared with the standard procedure. The expected proportion of each construction process in the frequency band template matrix is ​​calculated for similarity. The construction process with the highest similarity is taken as the inferred construction process of the construction grid, and the value of the similarity is output as the confidence level corresponding to the inferred construction process.

[0007] Furthermore, the evaluation and optimization method further includes: The entire construction area is divided into multiple construction grids according to the preset grid size. The grid size is matched with the construction section division and axis positioning. Each construction grid is an independent minimum analysis unit. The ultra-wideband positioning data of each smart safety helmet is mapped to the corresponding construction grid. The inertial measurement unit data of all smart safety helmets in the same construction grid are aggregated and processed to output the group inertial motion feature vector of the construction grid.

[0008] Furthermore, the consistency degree between the actual dominant process and the target process within the construction grid set is calculated based on the group inertial motion feature vector, and instructions are generated by combining this with the personnel density of the construction grid set. On-site matching includes: All construction grids corresponding to the target construction section in space are grouped into a construction grid set. The inferred construction procedure, the confidence level corresponding to the inferred construction procedure, and the personnel density of each construction grid in the construction grid set are extracted from the group inertial motion feature vector. The inferred construction procedures are weighted and statistically analyzed based on the personnel density of each construction grid to obtain the actual dominant procedures and overall confidence level of the construction grid set. The process consistency score is obtained by comparing the process logic relationship between the target process and the actual dominant process. The personnel saturation is calculated based on the ratio between the actual personnel density of the construction grid set and the baseline personnel density corresponding to the actual dominant process. Instructions are generated by comprehensively calculating the process consistency score, the overall confidence level, and the personnel on-site saturation level. On-site matching degree.

[0009] Furthermore, the step of comparing the target process with the actual dominant process based on their logical relationship to obtain a process consistency score includes: When the target process is the same as the actual dominant process, the process consistency score is assigned the first score. When the target process and the actual dominant process are either the preceding or succeeding processes in the preset standard process logic chain, the process consistency score is assigned a second score, which is less than the first score. When the target process and the actual dominant process are not the same process and are not the immediate predecessor or successor of each other, the process consistency score is assigned a third score, which is less than the second score.

[0010] Furthermore, calculating the personnel saturation based on the ratio between the actual personnel density of the construction grid set and the baseline personnel density corresponding to the actual dominant process includes: Retrieve the baseline personnel density corresponding to the actual dominant process, where the baseline personnel density is the minimum personnel density threshold required to perform the normal operation of the process. Calculate the ratio of the actual personnel density within the construction grid set to the baseline personnel density; When the ratio is greater than or equal to 1, the personnel saturation is assigned a value of 1; when the ratio is less than 1, the personnel saturation is assigned the value of the ratio.

[0011] Furthermore, the calculation of the execution damping coefficient based on the inferred difficulty of switching between the construction process and the target process, the current working surface inertial momentum of the construction grid set, and the historical instruction fulfillment data of the corresponding construction unit includes: Calculate the process switching damping factor based on the theoretical minimum transition time required to switch from the actual dominant process to the target process; The inertial momentum factor of the work surface is calculated based on the mid-frequency average energy, actual personnel density, and consistency of personnel movement direction within the construction grid set. The ratio of the actual closing time to the standard closing time of the construction unit for the same type of supervision instructions in the past preset time period is retrieved to calculate the historical performance damping factor. The process switching damping factor, the working surface inertial momentum factor, and the historical performance damping factor are weighted and synthesized to obtain the execution damping coefficient, wherein the weight of each factor is dynamically updated according to the historical instruction execution result data.

[0012] Furthermore, the calculation of the inertial momentum factor of the working face based on the mid-frequency average energy, actual personnel density, and the degree of consistency of personnel movement direction within the construction grid set includes: The horizontal movement angles of all smart safety helmets within the construction grid set are statistically analyzed, and the standard deviation of the movement angles is calculated. The smaller the standard deviation, the more consistent the movement direction of the personnel. The inertial momentum value is obtained by combining the mid-frequency average energy, the actual personnel density, and the standard deviation of the direction of movement angle. When the inertial momentum value is higher than the preset inertial threshold, the inertial momentum factor of the working surface tends to be high, indicating that the current working surface is in a high-intensity orderly operation state, and interrupting the current operation will generate a large execution resistance. When the inertial momentum value is lower than or equal to the preset inertial threshold, the inertial momentum factor of the working surface tends to be low, indicating that the current working surface is in a low-intensity or disordered state, and the execution resistance to interrupting the current operation is small.

[0013] This invention also provides a construction supervision instruction evaluation and optimization system, comprising: Data acquisition module: used to collect inertial measurement unit data and ultra-wideband positioning data of each smart safety helmet at the construction site in real time, and to obtain the text of the supervision instructions to be evaluated at present; Group state feature extraction module: Based on the data from the inertial measurement unit, this module uses a frequency band energy decoupling method to decompose the group motion spectrum of each smart safety helmet into low-frequency energy proportion, mid-frequency energy proportion, and high-frequency energy proportion, combined with pre-established standard procedures. A frequency band template matrix is ​​used to generate a group inertial motion feature vector for each construction grid. The group inertial motion feature vector includes at least the inferred construction procedure for each construction grid, the confidence level corresponding to the inferred construction procedure, and the personnel density. The instruction parsing and matching degree evaluation module is used to perform structured parsing of the supervision instruction text, extract the target process and target construction section, map the target construction section to the corresponding construction grid set, calculate the consistency degree between the actual dominant process and the target process within the construction grid set based on the group inertial motion feature vector, and generate instructions by combining the personnel density of the construction grid set. On-site matching degree; Damping prediction module: used to calculate the execution damping coefficient based on the difficulty of switching between the inferred construction process and the target process, the current working surface inertial momentum of the construction grid set, and the historical instruction performance data of the corresponding construction unit; Optimize decision output module: used to perform the following based on the instructions. The combination of the on-site matching degree and the execution damping coefficient generates an optimized scheme for the supervision instructions and outputs the optimized supervision instructions.

[0014] The construction supervision instruction evaluation and optimization method and system provided in this application processes the inertial measurement unit data of a group of people wearing smart safety helmets using a frequency band energy decoupling method. This decouples the complex group motion signal into low-frequency, mid-frequency, and high-frequency energy proportions with clear engineering physical meaning, avoiding the privacy risks and high computational consumption associated with relying on visual recognition of individual behavior. This makes it possible to achieve robust perception of the macro-construction rhythm at the edge with extremely low computational cost. Based on this, by constructing a group inertial motion feature vector that includes inferred procedures, confidence levels, and personnel density, and combining it with a standard procedure-frequency band template matrix for similarity matching, the method allows for the existence of procedure transition states, more accurately reflecting the actual working conditions at the construction site, and overcoming the lag and distortion problems caused by traditional solutions that rely solely on plan reporting. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the 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.

[0016] Figure 1 This is a flowchart of the construction supervision instruction evaluation and optimization method in an embodiment of the present invention; Figure 2This is a flowchart of the matching degree analysis between construction supervision instructions and the site in an embodiment of the present invention; Figure 3 This is a schematic diagram of the construction supervision instruction evaluation and optimization system in an embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Example 1 like Figure 1 As shown in the figure, this embodiment provides a method for evaluating and optimizing construction supervision instructions. This method solves the problem of misalignment between the timing of instruction issuance and the actual construction rhythm by sensing the group operation status at the construction site and dynamically evaluating its adaptability to the supervision instructions. The evaluation and optimization method mainly includes the following steps.

[0019] Step S100: Real-time acquisition of inertial measurement unit data and ultra-wideband positioning data of each smart safety helmet at the construction site, and acquisition of the current supervision instruction text to be evaluated.

[0020] Specifically, inertial measurement unit (IMU) data typically includes triaxial acceleration and triaxial angular velocity information, reflecting the wearer's head posture and limb movement intensity; ultra-wideband positioning data provides high-precision three-dimensional spatial coordinates to determine the specific location of personnel on the construction site. In this embodiment, the data acquisition frequency can be set to 50Hz to meet the requirements for capturing construction movement characteristics. Simultaneously acquiring supervisory instruction text refers to synchronously receiving instruction content from the supervisory management system or mobile terminal within the time window of acquiring on-site sensing data, ensuring temporal alignment between the sensing state and the instruction object. This synchronous acquisition mechanism of multi-source heterogeneous data lays the data foundation for subsequent correlation analysis between motion signals from the physical world and semantic information from the management world.

[0021] Step S200: Based on inertial measurement unit data, the frequency band energy decoupling method is used to decompose the group motion spectrum of each smart safety helmet into low-frequency energy proportion, mid-frequency energy proportion, and high-frequency energy proportion, combined with pre-established standard procedures. The frequency band template matrix generates the group inertial motion feature vector of each construction grid. The group inertial motion feature vector includes at least the inferred construction procedure, the confidence level of the inferred construction procedure, and the personnel density of each construction grid.

[0022] It is important to note that the group inertial motion feature vector in this embodiment differs from traditional individual behavior recognition results. Traditional methods often attempt to identify the specific action category of each worker, which not only involves privacy risks but also consumes enormous computing power and struggles to guarantee accuracy in complex construction scenarios. The group inertial motion feature vector used in this application is an intermediate statistical representation. It does not focus on who an individual is or what specific action they are performing. Instead, it aggregates the group motion signals within the entire construction grid and maps them to state parameters reflecting the macroscopic construction rhythm. For example, inferring the construction sequence indicates the most likely type of work currently underway in the area, confidence level reflects the reliability of the inference, and personnel density characterizes the scale of resource investment on the work surface. This approach avoids privacy issues and enables robust perception of on-site conditions with extremely low latency on edge computing devices, providing necessary contextual anchors for instruction evaluation.

[0023] Step S300: Perform structured parsing on the supervision instruction text, extract the target process and target construction section, map the target construction section to the corresponding construction grid set, calculate the consistency between the actual dominant process and the target process within the construction grid set based on the group inertial motion feature vector, and generate instructions by combining the personnel density of the construction grid set. On-site matching accuracy. Structured parsing refers to using natural language processing technology to extract key elements from unstructured instruction text, such as the name of the required rectification process, the specified floor, or axis location.

[0024] Mapping the target construction section to a construction grid set converts the spatial description in management semantics into physical spatial units that the system can compute. (Instruction) On-site matching degree is not simply a string comparison result, but a comprehensive evaluation indicator. It considers both the logical relevance between the current actual work content and the instruction requirements, as well as whether there are sufficient human resources on-site to respond to the instruction. For example, even if the current process is the same as the target process, if the on-site personnel density is extremely low, the matching degree will decrease accordingly, reflecting the reality that although the instruction is correct, there is a lack of conditions for execution.

[0025] Step S400: Calculate the execution damping coefficient based on the inferred difficulty of switching between the construction process and the target process, the current working surface inertial momentum of the construction grid set, and the historical instruction performance data of the corresponding construction unit.

[0026] In this embodiment, the execution damping coefficient is a quantitative parameter characterizing the resistance or delay tendency of command execution. Its physical metaphor is similar to damping in a mechanical system; a higher value means that the inertia that needs to be overcome to implement the command is greater. The difficulty of process switching reflects the theoretical cost of transitioning from the current state to the state required by the command. The inertial momentum of the work surface is the core innovative parameter introduced in this application. It describes the intensity and orderliness of the current work surface operation from a purely kinematic perspective. High-intensity and consistent group movement implies huge interruption costs. Historical command fulfillment data introduces empirical corrections in the time dimension. The combination of these three factors enables the execution damping coefficient to transcend subjective experience judgment and objectively quantify the on-site immediate response capability to commands.

[0027] Step S500, according to the instruction The combined results of on-site matching degree and execution damping coefficient generate an optimized scheme for the supervision instructions and output the optimized supervision instructions.

[0028] Specifically, the generation of optimization solutions is based on joint judgment in a two-dimensional decision space. For example, when the matching degree is high and the damping coefficient is low, it indicates that the site conditions are fully met, and the system can suggest immediate issuance and follow-up according to the conventional process; when the matching degree is low but the damping coefficient is also low, it may mean that the site is idle or in a transitional state but the work type does not match, and the system can suggest postponing issuance until the appropriate process begins; when the matching degree is high but the damping coefficient is also high, it indicates that although relevant work is being carried out on site, it is in a high-intensity continuous construction phase, and forcibly interrupting it would result in significant efficiency losses, and the system can suggest strengthening supervision measures or adjusting the wording of instructions to reduce interference; when both are in unfavorable ranges, the system can suggest upgrading to a higher level of management intervention. Through this differentiated optimization output, this embodiment realizes the transformation of supervision instructions from mechanical issuance to adaptive intelligent control, significantly improving the level of precision in project management.

[0029] Specifically, based on inertial measurement unit data, the frequency band energy decoupling method is used to decompose the group motion spectrum of each smart safety helmet into low-frequency energy proportion, mid-frequency energy proportion, and high-frequency energy proportion. This process is not a simple mathematical filtering, but rather a deep binding of signal processing parameters with the physical motion characteristics of the construction site.

[0030] First, the composite acceleration modulus values ​​of all smart safety helmets within each construction grid are processed using short-time frame segmentation. In this embodiment, the window length is preferably set to 2 seconds, and the step size is set to 0.5 seconds. The 2-second window length is chosen because the cycle of most construction limb movements (such as a complete masonry swing or rebar binding cycle) is usually between 1 and 2 seconds, which ensures sufficient frequency resolution to distinguish the movement characteristics of different processes; while the 0.5-second step size ensures data continuity while also taking into account the real-time processing load of the edge computing device. Subsequently, a fast Fourier transform is performed on each frame of data to obtain the power spectral density, and the power spectral density is decomposed into low-frequency energy, mid-frequency energy, and high-frequency energy according to a preset frequency band range.

[0031] To assign clear engineering and physical meaning to the spectral characteristics, this application defines and divides the three frequency bands as follows: The low-frequency band corresponds to slow personnel movement and work surface transfers, such as worker walking and inspection, and material handling displacement; its frequency range is typically set to 0.3Hz to 1.5Hz, reflecting the activity level of site relocation or material preparation. The mid-frequency band corresponds to rhythmic limb movements, such as repeated bending and hand operations during rebar tying, arm swinging during masonry plastering, and the shaking of a concrete vibrator; its frequency range is typically set to 1.5Hz to 5.0Hz, and is the core frequency band characterizing the intensity of the main construction work. The high-frequency band corresponds to tool impacts and high-frequency vibrations, such as transient high-energy signals generated by electric picks, cutting machine operations, and impact drilling; its frequency range is typically set to 5.0Hz to 15.0Hz, reflecting the execution status of high-disturbance or high-risk processes. The proportions of low-frequency, mid-frequency, and high-frequency energy in the total energy are calculated separately to obtain the proportions of low-frequency, mid-frequency, and high-frequency energy. This frequency band decoupling method based on physical semantics enables subsequent algorithms to directly understand what type of activity is taking place on-site, rather than simply processing abstract digital signals.

[0032] Based on this, combined with pre-established standard procedures The frequency band template matrix generates the group inertial motion feature vector of each construction grid. Specifically, standard procedures... The frequency band template matrix can be established in two ways: one is to select typical processes at the beginning of the project and conduct on-site measurement and calibration for several days, collecting the frequency distribution of each process under stable operating conditions as a benchmark; the other is to directly import the project's BIM construction organization simulation data and generate a theoretical template based on the trade ratio and process standards. Regardless of the method used, the matrix includes the expected proportions of low-frequency, mid-frequency, and high-frequency bands set for different construction processes. For example, for the rebar tying process, its typical frequency band energy distribution might be 0.25 for low-frequency, 0.65 for mid-frequency, and 0.10 for high-frequency; while for the electromechanical pre-embedded cutting process, it might be 0.15 for low-frequency, 0.30 for mid-frequency, and 0.55 for high-frequency. During real-time operation, the system will compare the real-time feature triplets composed of the low-frequency, mid-frequency, and high-frequency energy proportions of each construction grid calculated in real time with the standard process... The expected proportion of each construction process in the frequency band template matrix is ​​used to calculate similarity. The cosine similarity algorithm can be used for similarity calculation. This algorithm effectively measures the consistency of two vectors in direction, thus ignoring the interference of absolute energy magnitude caused by differences in wearing position. The construction process with the highest similarity is selected as the inferred construction process for that construction grid, and the similarity value is used as the confidence output corresponding to the inferred construction process. This confidence parameter is crucial; it quantifies the degree of certainty regarding the current sensing state, provides a reliability weight for subsequent multi-grid weighted statistics, and avoids the risk of misclassifying noise as a valid process.

[0033] Furthermore, to achieve refined spatial management, the evaluation and optimization method also includes dividing the entire construction area into multiple construction grids according to a preset grid size. The grid size matches the construction segment division and axis positioning, with each construction grid serving as an independent minimum analysis unit. In this embodiment, the grid size is preferably set to 5m × 5m, a scale that not only matches the granularity of conventional building construction segment division but also effectively covers the activity range of an independent work team. The ultra-wideband positioning data of each smart safety helmet is mapped to the corresponding construction grid, and the inertial measurement unit data of all smart safety helmets within the same construction grid are aggregated to output the group inertial motion feature vector of that construction grid.

[0034] It is important to note that this embodiment employs a group spectrum aggregation approach rather than individual behavior recognition, which offers several technical advantages. First, from a privacy compliance perspective, the group spectrum only reflects macroscopic motion statistics, completely stripping away biometric information that could identify individuals, thus meeting increasingly stringent privacy protection requirements at construction sites. Second, from a computational efficiency perspective, aggregating data within the grid before performing FFT transformation and template matching requires far less computation than classifying each individual's behavior individually using AI, enabling the algorithm to run in real-time on low-cost edge gateways. Finally, from a business adaptation perspective, supervisory instructions are typically issued to work areas or teams rather than individuals. The group inertial motion feature vector directly characterizes the macroscopic construction rhythm and resource allocation status of the work area, naturally aligning with the decision-making granularity of supervisory management and offering greater guidance value than scattered individual behavior tags. Through the aforementioned frequency band energy decoupling and grid aggregation mechanism, the system can transform massive amounts of raw sensor data into structured, semantic, and spatially resolved group state descriptions, laying a solid data foundation for subsequent instruction matching and damping prediction.

[0035] Based on the characteristic vector of group inertial motion, the consistency between the actual dominant process and the target process within the construction grid set is calculated, and instructions are generated by combining the personnel density of the construction grid set. On-site matching degree. Specifically, this process is not a simple text comparison, but a multi-dimensional evaluation system that integrates spatial statistics, process sequence logic and human resource allocation, aiming to truly reflect the executability and adaptability of supervision instructions at the current time and space node.

[0036] For details, please refer to Figure 2 Step S300 includes: S301: Merge all construction grids corresponding to the target construction section in space into a construction grid set, and extract the inferred construction procedures, confidence levels, and personnel density of each construction grid in the construction grid set from the feature vector of the group inertial motion.

[0037] S302: Based on the personnel density of each construction grid, the inferred construction procedures are weighted and statistically analyzed to obtain the actual dominant procedures and overall confidence level of the construction grid set.

[0038] On construction sites, the distribution of personnel within different grids is often extremely uneven. Using simple averaging or voting mechanisms can easily lead to misinterpretations by a few grids where personnel are resting, moving between locations, or performing sporadic work. By using personnel density weighting, the system can automatically focus on areas with the most concentrated resource investment and the most intensive work activities, thereby accurately identifying the actual dominant process representing the true progress of the current construction section. For example, a construction section may contain 10 grids, with 8 grids in a low-density, waiting state and the inferred process being "cleaning," while 2 grids are in a high-density, work-in-progress state and the inferred process being "rebar tying." After density weighting, the actual dominant process for this construction section will be robustly determined to be "rebar tying," effectively filtering out background noise.

[0039] S303: Compare the process logic relationship between the target process and the actual dominant process to obtain a process consistency score.

[0040] To adapt to the dynamic and complex nature of process connections at construction sites, this embodiment does not employ a black-and-white binary judgment, but instead designs a tiered scoring mechanism. Specifically, when the target process is the same as the actual leading process, the process consistency score is assigned the first score; when the target process and the actual leading process are either immediate predecessors or successors in the preset standard process logic chain, the process consistency score is assigned the second score, which is less than the first score; when the target process and the actual leading process are not the same process and are not immediate predecessors or successors, the process consistency score is assigned the third score, which is less than the second score. The engineering significance of this tiered design lies in acknowledging and accommodating reasonable process transition states. In actual construction, there are often necessary overlaps, interleavings, or preparation times between processes. For example, issuing a "concrete pouring" instruction when "formwork installation" is about to be completed, although the names are different, belongs to a closely connected normal process. At this time, assigning a second score (such as 0.5) reflects the difference in processes and affirms the rationality of their timing, avoiding the system misjudging a normal transition as a serious misalignment and generating an invalid warning. Only when the two are completely unrelated (such as issuing a "waterproofing construction" instruction during the "masonry" stage) is the lowest third score (such as 0.1) assigned to clearly identify the real timing conflict.

[0041] S304: Calculate the personnel saturation based on the ratio between the actual personnel density of the construction grid set and the baseline personnel density corresponding to the actual dominant process.

[0042] The baseline personnel density refers to the minimum personnel density threshold required for the normal operation of this process. Its data source can be national or industry-issued labor quota standards, or the historical average personnel density of similar processes operating efficiently, as determined in the early stages of the project. In specific calculations, when the ratio of the actual personnel density to the baseline personnel density is greater than or equal to 1, the personnel saturation is assigned a value of 1; when the ratio is less than 1, the personnel saturation is assigned the ratio. The physical meaning of this truncation (i.e., taking min(1, ratio)) is that, for instruction execution, meeting the baseline personnel configuration is considered sufficient. Excessive personnel accumulation does not linearly improve instruction response efficiency; instead, it may indicate idle work or management chaos on-site. Therefore, locking the saturation upper limit to 1 can prevent artificially high matching degrees caused by abnormal personnel aggregation, ensuring that the evaluation results objectively reflect the true execution capability.

[0043] S305: Generate instructions by comprehensively calculating process consistency score, overall confidence level, and personnel attendance saturation. On-site matching degree.

[0044] In a specific embodiment, the matching degree M can be synthesized by the following formula: ; in, Indicates the process consistency score. Indicates the overall confidence level. Indicates the saturation level of personnel in place. and These are the weighting coefficients for the process logic dimension and the personnel resource dimension, respectively, and they satisfy... In this embodiment, The preferred setting is 0.7. The preferred setting is 0.3, which reflects the dominant role of process sequence correctness in the matching degree assessment, while also taking into account the supporting role of personnel allocation.

[0045] Furthermore, the weighting coefficients are not fixed and can be dynamically adjusted in other implementations based on project type, construction stage, or historical feedback data. For example, if a reinforcement rectification order (target process) is issued while a floor is undergoing continuous concrete pouring (the actual dominant process), since the two are neither identical nor sequential, Only 0.1; even with a high density of people on site, Even if the score is close to 1, the overall matching degree M will still be below 0.45, and the system will therefore determine it as a "mismatch" state. Conversely, if a safety reminder is issued during the lunch break, although the personnel density is low, it will still be effective. The safety reminders are relatively minor, but since they do not strongly conflict with any process (or can be considered as a general accompanying process), they are not necessary. It can maintain a high level, and the final matching degree can still fall into the "normal" or "matching" range, thereby supporting differentiated decision output. Through the above-mentioned multi-dimensional synthesis mechanism, this embodiment realizes the control of instructions. The refined quantification of on-site matching provides reliable data support for subsequent joint decision-making.

[0046] Based on the inferred difficulty of switching between construction procedures and target procedures, the current inertial momentum of the work surface in the construction grid set, and the historical instruction fulfillment data of the corresponding construction unit, the execution damping coefficient is calculated. Specifically, the execution damping coefficient is not a single-dimensional evaluation indicator, but a comprehensive parameter obtained by weighted synthesis of three orthogonal factors. It aims to quantify the resistance to instruction implementation from three dimensions: theoretical cost, physical state, and historical experience. Among them, the procedure switching damping factor reflects the theoretical time cost of job allocation and process conversion; the work surface inertial momentum factor characterizes the physical trend strength of the current work surface maintaining its existing operating state; and the historical fulfillment damping factor introduces a time-dimensional credit correction. This multi-factor synthesis mechanism avoids the one-sidedness caused by relying solely on subjective experience or a single indicator, making the calculation of the damping coefficient both objective in a mechanical sense and taking into account the dynamic complexity of engineering management.

[0047] Specifically, the construction of the working face inertial momentum factor is entirely based on pure kinematic parameters, without relying on any semantic labels or manually entered data. The working face inertial momentum factor is calculated based on the mid-frequency average energy, actual personnel density, and the consistency of personnel movement directions within the construction grid set. The underlying logic of this design is that the operational inertia at a construction site is essentially a physical quantity determined by the intensity of action, the scale of participation, and the degree of orderliness. Only when all three factors are simultaneously at a high level will forcibly interrupting the current operation result in significant conversion costs and efficiency losses.

[0048] In practical implementation, firstly, the horizontal movement direction angles of all smart safety helmets within the construction grid are statistically analyzed, and the standard deviation of these movement direction angles is calculated. A smaller standard deviation indicates a more consistent movement direction among the workers. For example, in concrete pouring or assembly line masonry scenarios, workers often move in an orderly manner along specific paths or towards the same work surface; in this case, the standard deviation of the movement direction angles is extremely small, reflecting a high degree of organizational coordination. However, in rest, waiting, or disorderly cleaning scenarios, the movement directions of workers are randomly distributed, and the standard deviation increases significantly. Subsequently, the mid-frequency average energy, actual personnel density, and standard deviation of the movement direction angles are combined to calculate the inertial momentum value. In a preferred embodiment, the inertial momentum value... It can be calculated using the following formula:

[0049] in, This represents the average energy in the mid-frequency band within the construction grid set. This parameter was obtained in Example 2 through frequency band energy decoupling and directly reflects the intensity of limb rhythmic operations. This represents the actual personnel density and characterizes the scale of resource input at the work site; This represents the standard deviation of the normalized movement direction angle, ranging from 0 to 1, and is used to characterize the degree of orderliness of the motion. The physical meaning of this formula is that only when the workload is high (…). Large number of participants Large) and the direction of movement is highly consistent ( The inertial momentum value only reaches its peak when the inertial momentum is small. The absence of any dimension will lead to the decay of inertial momentum, which accurately describes the non-linear characteristic of "interruption cost". For example, even if there are many people on site and the movements are violent, if the direction is chaotic (such as sudden chaos or disorderly handling), the actual interruptibility is relatively high, because there is no flow rhythm that requires overall coordination to stop.

[0050] To map continuous inertial momentum values ​​into damping factors that facilitate decision-making, this embodiment employs a Sigmoid-type normalization function. When the inertial momentum value is higher than a preset inertial threshold, the inertial momentum factor of the working surface tends to be high, indicating that the current working surface is in a high-intensity, orderly operating state, and interrupting the current operation will generate significant execution resistance. When the inertial momentum value is lower than or equal to the preset inertial threshold, the inertial momentum factor of the working surface tends to be low, indicating that the current working surface is in a low-intensity or disordered state, and interrupting the current operation will generate less execution resistance.

[0051] Specifically, the normalization formula can be expressed as: ; in To preset the inertia threshold, This represents the slope coefficient. Threshold. The calibration of the inertial momentum is crucial for the accuracy of the algorithm. It is not a fixed constant but should be obtained through field measurements during the project initiation phase. For example, select several samples each of typical high-inertia processes (such as continuous pouring) and low-inertia processes (such as sporadic repairs) from the project, calculate the statistical distribution of their inertial momentum values, and take the median of the means of the two types of samples as the initial... .

[0052] In addition to the inertial momentum factor of the work surface, the calculation of the execution damping coefficient also needs to consider two other factors. First, the process switching damping factor is calculated based on the theoretical minimum transfer time required to switch from the actual dominant process to the target process. The theoretical minimum transfer time can be extracted from the construction organization design or BIM schedule, reflecting the rigid time window required for job changes, equipment relocation, and process preparation. For example, switching from "rebar tying" to "concrete pouring" involves multiple steps such as pump truck placement and formwork acceptance, resulting in a long transfer time and a higher corresponding process switching damping factor; while switching from "masonry" to "plastering" can usually be done continuously by the same team, with a short transfer time and a lower damping factor. Second, the historical performance damping factor is calculated by retrieving the ratio of the actual closing time to the standard closing time of similar supervision instructions from the construction unit within a preset time period. This factor incorporates a credit evaluation mechanism; if a team delays rectification for a long period, its historical performance damping factor will increase, prompting the system to allow more buffer time or take stronger supervisory measures when issuing instructions.

[0053] Finally, the execution damping coefficient is obtained by weighting and synthesizing the process switching damping factor, the working surface inertial momentum factor, and the historical performance damping factor. The weighting of each factor is dynamically updated based on historical instruction execution results data. In a specific embodiment, the execution damping coefficient... The calculation formula is: ; in, , , These are the dynamic weights of the three factors, and they satisfy... Dynamic weight updates are key to achieving adaptive optimization in this scheme. After each instruction closure, the system records the deviation between the actual execution time and the predicted damping coefficient, and uses Bayesian updates or gradient descent algorithms to correct the weights in reverse. For example, in a project where process switching is difficult but on-site response is consistently rapid, it indicates strong work allocation capabilities, and the system will automatically reduce the weights. And improve or The weight of the event; conversely, if the situation is extremely sensitive to interruptions of a high-inertia state, then The coefficient will be gradually increased. This closed-loop feedback mechanism ensures that the execution damping coefficient continuously approaches the actual working conditions as the project progresses, rather than remaining at the level of a static empirical model. Through the above-mentioned multi-factor synthesis and dynamic calibration mechanism, this embodiment achieves accurate quantification of instruction execution resistance, providing a solid physical basis for subsequent joint decision-making.

[0054] Specifically, according to the instructions The system combines the results of on-site matching degree and execution damping coefficient to generate an optimized solution for the supervision instructions and output the optimized instructions. This optimization solution is not based on a simple triggering of a single threshold, but rather constructs a two-dimensional decision space composed of matching degree and damping coefficient. The system automatically maps differentiated handling strategies based on the high and low combinations of these two indicators, thereby transforming the issuance of supervision instructions from a mechanical process to context-aware intelligent control.

[0055] To more clearly illustrate the joint decision-making logic of this application, two typical construction site scenarios are used as examples for detailed explanation. In the first scenario, a floor is in the continuous concrete pouring stage. At this time, the characteristic vector of the group's inertial motion shows a high proportion of mid-frequency energy, high personnel density, and highly consistent movement direction, indicating that the work surface is in a high-intensity, orderly operation state. If the supervisor intends to issue an instruction regarding the rectification of rebar binding quality, the system will first calculate the instruction. On-site matching degree. Because the target process "reinforcement rectification" and the actual leading process "concrete pouring" are neither the same nor consecutive processes, the process consistency score is extremely low, causing the overall matching degree to fall into the "misalignment" range (e.g., less than 0.45). Simultaneously, the system calculates the execution damping coefficient. Due to the current working surface's inertial momentum factor approaching a high value, and the theoretical transition time from pouring to reinforcement rectification being extremely long, the process switching damping factor is also high, resulting in a final execution damping coefficient significantly higher than the preset high damping threshold (e.g., greater than 0.65). Faced with this double disadvantage of "low matching degree + high damping," the system determines that now is not the appropriate time to issue the instruction. Forcing its issuance would not only fail to obtain an immediate response but may also disrupt the continuity of critical processes. Therefore, the optimization decision output module will automatically generate suggestions such as "upgrade the supervision notice and copy it to the project department" or "postpone it after pouring is completed," instead of directly pushing it to the on-site team leader. This processing mechanism effectively avoids the disruption of the construction site's rhythm by invalid instructions, demonstrating the technical solution's precise response to the actual pain points of the project.

[0056] In contrast, in the second scenario, during lunch breaks or work transitions, the on-site group spectrum shows a predominance of low-frequency energy, low personnel density, and random movement, indicating a low-intensity or disordered work environment. If a routine safety reminder or civilized construction rectification order is issued at this time, the system will calculate the corresponding instruction. The on-site matching degree may be at a "normal" level due to low personnel saturation. However, since the inertial momentum factor and process switching damping factor of the work surface are both close to low values, the execution damping coefficient is far below the low damping threshold (e.g., less than 0.35). For this combination of "acceptable matching degree + low damping", the system judges that the on-site conditions are ready for immediate response and the interruption cost is minimal. The optimization decision output module will suggest "issue immediately and follow up according to the normal process".

[0057] Specifically, once a supervisor's instruction is executed and closed on-site, the system automatically collects the actual execution time data and compares it with the predicted execution damping coefficient. If the actual closure time of an instruction is significantly longer than expected, it indicates that the system has underestimated the current execution resistance. The closed-loop feedback correction module will use Bayesian updates or gradient descent algorithms to increase the weight of the factor that caused the underestimation. For example, in a project where process switching is difficult but on-site response is always rapid, it indicates that the project has strong work allocation capabilities. The system will automatically reduce the weight w1 of the process switching damping factor and increase the weight w2 of the work surface inertial momentum factor or the weight w3 of the historical performance damping factor. Conversely, if the site is extremely sensitive to interruptions in a high-inertia state, w2 will be gradually increased. This adaptive calibration based on measured data is usually performed offline every night, avoiding the occupation of edge gateway computing power by real-time calculations. Through this closed-loop feedback mechanism, the calculation model of the damping coefficient can continuously approach the real working conditions as the project progresses, overcoming the problem of static experience models not being suitable for different projects when transferred, and significantly improving the long-term accuracy and practicality of the evaluation and optimization of supervision instructions.

[0058] Example 2: Please refer to Figure 3 This embodiment provides a construction supervision instruction evaluation and optimization system, which includes a data acquisition module 10, a group state feature extraction module 20, an instruction parsing and matching degree evaluation module 30, a damping prediction module 40, and an optimization decision output module 50.

[0059] The data acquisition module 10 includes multiple smart safety helmets and ultra-wideband positioning base stations deployed at the construction site. The smart safety helmets have built-in inertial measurement units (IMUs) for real-time acquisition of IMU data from each helmet, while the ultra-wideband positioning base stations acquire ultra-wideband positioning data from each helmet in real-time. Specifically, the smart safety helmets, acting as the end nodes of the perception layer, typically acquire triaxial acceleration and triaxial angular velocity data at a frequency of 50Hz using their built-in IMUs, and transmit the raw data stream to the edge gateway via a wireless communication protocol. The ultra-wideband positioning base stations provide centimeter-level accuracy in three-dimensional spatial coordinates through time difference of arrival (TDOA) or bidirectional ranging algorithms. This dual-modal data acquisition mechanism ensures that the system can simultaneously acquire the microscopic motion characteristics and macroscopic spatial distribution of personnel, laying the physical foundation for subsequent group state analysis.

[0060] The group state feature extraction module 20 receives inertial measurement unit data and ultra-wideband positioning data output by the data acquisition module 10. It uses a frequency band energy decoupling method to decompose the group motion spectrum into low-frequency, mid-frequency, and high-frequency energy proportions, combined with pre-established standard procedures. The frequency band template matrix generates the group inertial motion feature vectors for each construction grid. In specific deployments, this module is preferably run on the gateway device at the edge processing layer, rather than on a cloud server. The core consideration for this architecture design is that the raw IMU data throughput is enormous and latency-sensitive; processing it in the cloud would face severe bandwidth bottlenecks and network jitter risks. However, after completing frame-segmented FFT, frequency band decomposition, and grid aggregation at the edge, only a lightweight group inertial motion feature vector (containing structured parameters such as inferred procedures, confidence levels, and personnel density) needs to be uploaded, reducing the uplink data volume by several orders of magnitude while ensuring millisecond-level real-time response capabilities. The frequency band energy decoupling and template matching logic executed internally by this module is completely consistent with that in Example 2, and will not be repeated here.

[0061] The instruction parsing and matching evaluation module 30 is used to acquire the supervision instruction text to be evaluated and perform structured parsing on the supervision instruction text, extract the target process and target construction section, map the target construction section to the corresponding construction grid set, calculate the consistency degree between the actual dominant process and the target process within the construction grid set based on the group inertial motion feature vector, and generate instructions by combining the personnel density of the construction grid set. On-site matching degree. Specifically, this module integrates a natural language processing engine and a spatial indexing service. The natural language processing engine is responsible for extracting key semantic elements from unstructured text, while the spatial indexing service maintains a dynamic mapping table between construction section names, axis numbers, and physical grid IDs. When calculating the matching degree, this module reuses the weighted statistical and hierarchical assignment mechanism described in Example 3, that is, determining the actual dominant process by weighting by personnel density, and generating a score based on the logical chain relationship of the process and the saturation of personnel in place. This design enables the system to understand the adaptation relationship between management semantics and physical state, rather than simply performing a literal comparison.

[0062] The damping prediction module 40 is used to calculate the execution damping coefficient based on the inferred difficulty of the process switching between the construction process and the target process, the current inertial momentum of the working face in the construction grid set, and the historical instruction performance data of the corresponding construction unit. As the core calculation unit of this system, the unique feature of this module is the introduction of the working face inertial momentum factor derived purely from kinematics. Specifically, this module directly extracts the mid-frequency average energy, actual personnel density, and standard deviation of the movement direction angle from the characteristic vector of the group inertial motion, and calculates the physical quantity characterizing the interruption cost by combining the formulas described in Example 4, and then normalizes it into a damping factor through a Sigmoid-like function. At the same time, this module also maintains a historical performance database, recording the ratio of the actual closing time to the standard time for each construction unit for the same type of instruction in the past, thereby constructing a correction factor for the credit dimension. The three factors are dynamically weighted and synthesized to output the final execution damping coefficient.

[0063] The optimized decision output module 50 is used to optimize the decision output module according to the instructions. The combined results of on-site matching degree and execution damping coefficient generate and output optimized suggestions for supervision instructions. Specifically, this module incorporates a two-dimensional decision strategy matrix, dividing the matching degree and damping coefficient into different intervals and mapping them to differentiated handling actions. For example, when the matching degree is high and the damping is low, the module outputs an "Immediately Issue" suggestion; when the matching degree is low and the damping is high, the module outputs an "Escalate Notification or Delay Issuance" suggestion. This module also supports integration with various downstream systems, such as supervision management platforms, mobile apps, or on-site audible and visual alarms, to achieve multi-channel delivery of optimized suggestions. Through this joint decision-making mechanism, the system can transform abstract evaluation indicators into concrete management action guidelines, effectively avoiding interference with the construction schedule by invalid instructions.

[0064] The closed-loop feedback correction module collects actual execution time data after the supervisor's instruction is executed and dynamically corrects the calculation weights of various components in the damping prediction module 40 based on this data. Specifically, when on-site personnel provide feedback via mobile terminal that the instruction has been rectified or the system detects an expected change in the work status, the module automatically records the time stamp difference between the instruction issuance and closure, using it as the truth value label. Subsequently, using Bayesian update or gradient descent algorithms, the deviation between the predicted damping and the actual execution time is compared, and the weight coefficients of the process switching damping factor, the work surface inertial momentum factor, and the historical performance damping factor are adjusted in reverse. This correction process is typically performed offline every night to avoid consuming real-time computing resources during the day. Through this continuous learning mechanism, the system can gradually adapt to the construction characteristics and management style of specific projects, overcome the generalization problem of static models under different working conditions, and achieve long-term adaptive improvement in evaluation accuracy.

[0065] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, which may include: read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk, etc.

[0066] Furthermore, the embodiments of the present invention have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for evaluating and optimizing construction supervision instructions, characterized in that, The evaluation and optimization methods include: Real-time acquisition of inertial measurement unit data and ultra-wideband positioning data from each smart safety helmet at the construction site, while simultaneously obtaining the text of the supervisor's instructions to be evaluated; Based on the data from the inertial measurement unit, the frequency band energy decoupling method is used to decompose the group motion spectrum of each smart safety helmet into low-frequency energy proportion, mid-frequency energy proportion, and high-frequency energy proportion, combined with pre-established standard procedures. A frequency band template matrix is ​​used to generate a group inertial motion feature vector for each construction grid. The group inertial motion feature vector includes at least the inferred construction procedure for each construction grid, the confidence level corresponding to the inferred construction procedure, and the personnel density. The supervision instruction text is structured and parsed to extract the target process and target construction section. The target construction section is mapped to the corresponding construction grid set. Based on the group inertial motion feature vector, the consistency degree between the actual dominant process and the target process within the construction grid set is calculated. Combined with the personnel density of the construction grid set, instructions are generated. On-site matching degree; The execution damping coefficient is calculated based on the difficulty of switching between the inferred construction process and the target process, the current working surface inertial momentum of the construction grid set, and the historical instruction performance data of the corresponding construction unit. According to the instructions The combination of the on-site matching degree and the execution damping coefficient generates an optimized scheme for the supervision instructions and outputs the optimized supervision instructions.

2. The construction supervision instruction evaluation and optimization method as described in claim 1, characterized in that, Based on the data from the inertial measurement unit, the frequency band energy decoupling method is used to decompose the group motion spectrum of each smart safety helmet into low-frequency energy proportion, mid-frequency energy proportion, and high-frequency energy proportion, including: The composite acceleration modulus of all smart safety helmets in each construction grid is processed by short-time frame division. Fourier transform is performed on each frame to obtain the power spectral density. The power spectral density is decomposed into low-frequency energy, mid-frequency energy and high-frequency energy according to the preset frequency band range. The low-frequency band corresponds to slow personnel movement and work surface transfer, the mid-frequency band corresponds to rhythmic limb movements, and the high-frequency band corresponds to tool impact and high-frequency vibration. The proportions of the low-frequency energy, mid-frequency energy, and high-frequency energy in the total energy are calculated to obtain the proportions of low-frequency energy, mid-frequency energy, and high-frequency energy.

3. The construction supervision instruction evaluation and optimization method as described in claim 2, characterized in that, The combination of pre-established standard procedures The frequency band template matrix generates the group inertial motion feature vectors for each construction grid, including: Pre-determine the expected proportions of low-frequency bands, mid-frequency bands, and high-frequency bands for different construction procedures to form standard procedures. Frequency band template matrix; The real-time feature triplet, composed of the low-frequency energy proportion, mid-frequency energy proportion, and high-frequency energy proportion of each construction grid obtained in real time, is compared with the standard procedure. The expected proportion of each construction process in the frequency band template matrix is ​​calculated for similarity. The construction process with the highest similarity is taken as the inferred construction process of the construction grid, and the value of the similarity is output as the confidence level corresponding to the inferred construction process.

4. The construction supervision instruction evaluation and optimization method as described in claim 1, characterized in that, The evaluation and optimization method further includes: The entire construction area is divided into multiple construction grids according to the preset grid size. The grid size is matched with the construction section division and axis positioning. Each construction grid is an independent minimum analysis unit. The ultra-wideband positioning data of each smart safety helmet is mapped to the corresponding construction grid. The inertial measurement unit data of all smart safety helmets in the same construction grid are aggregated and processed to output the group inertial motion feature vector of the construction grid.

5. The construction supervision instruction evaluation and optimization method as described in claim 1, characterized in that, The process involves calculating the consistency between the actual dominant process and the target process within the construction grid set based on the group inertial motion feature vector, and generating instructions by combining this with the personnel density of the construction grid set. On-site matching includes: All construction grids corresponding to the target construction section in space are grouped into a construction grid set. The inferred construction procedure, the confidence level corresponding to the inferred construction procedure, and the personnel density of each construction grid in the construction grid set are extracted from the group inertial motion feature vector. The inferred construction procedures are weighted and statistically analyzed based on the personnel density of each construction grid to obtain the actual dominant procedures and overall confidence level of the construction grid set. The process consistency score is obtained by comparing the process logic relationship between the target process and the actual dominant process. The personnel saturation is calculated based on the ratio between the actual personnel density of the construction grid set and the baseline personnel density corresponding to the actual dominant process. Instructions are generated by comprehensively calculating the process consistency score, the overall confidence level, and the personnel on-site saturation level. On-site matching degree.

6. The construction supervision instruction evaluation and optimization method as described in claim 5, characterized in that, The step of comparing the target process with the actual dominant process to obtain a process consistency score includes: When the target process is the same as the actual dominant process, the process consistency score is assigned the first score. When the target process and the actual dominant process are either the preceding or succeeding processes in the preset standard process logic chain, the process consistency score is assigned a second score, which is less than the first score. When the target process and the actual dominant process are not the same process and are not the immediate predecessor or successor of each other, the process consistency score is assigned a third score, which is less than the second score.

7. The construction supervision instruction evaluation and optimization method as described in claim 5, characterized in that, The calculation of personnel saturation based on the ratio between the actual personnel density of the construction grid set and the baseline personnel density corresponding to the actual dominant process includes: Retrieve the baseline personnel density corresponding to the actual dominant process, where the baseline personnel density is the minimum personnel density threshold required to perform the normal operation of the process. Calculate the ratio of the actual personnel density within the construction grid set to the baseline personnel density; When the ratio is greater than or equal to 1, the personnel saturation is assigned a value of 1; when the ratio is less than 1, the personnel saturation is assigned the value of the ratio.

8. The construction supervision instruction evaluation and optimization method as described in claim 1, characterized in that, The calculation of the execution damping coefficient based on the inferred difficulty of switching between the construction process and the target process, the current working surface inertial momentum of the construction grid set, and the historical instruction performance data of the corresponding construction unit includes: Calculate the process switching damping factor based on the theoretical minimum transition time required to switch from the actual dominant process to the target process; The inertial momentum factor of the work surface is calculated based on the mid-frequency average energy, actual personnel density, and consistency of personnel movement direction within the construction grid set. The ratio of the actual closing time to the standard closing time of the construction unit for the same type of supervision instructions in the past preset time period is retrieved to calculate the historical performance damping factor. The process switching damping factor, the working surface inertial momentum factor, and the historical performance damping factor are weighted and synthesized to obtain the execution damping coefficient, wherein the weight of each factor is dynamically updated according to the historical instruction execution result data.

9. The construction supervision instruction evaluation and optimization method as described in claim 8, characterized in that, The calculation of the inertial momentum factor of the working face based on the mid-frequency average energy, actual personnel density, and consistency of personnel movement direction within the construction grid set includes: The horizontal movement angles of all smart safety helmets within the construction grid set are statistically analyzed, and the standard deviation of the movement angles is calculated. The smaller the standard deviation, the more consistent the movement direction of the personnel. The inertial momentum value is obtained by combining the mid-frequency average energy, the actual personnel density, and the standard deviation of the direction of movement angle. When the inertial momentum value is higher than the preset inertial threshold, the inertial momentum factor of the working surface tends to be high, indicating that the current working surface is in a high-intensity orderly operation state, and interrupting the current operation will generate a large execution resistance. When the inertial momentum value is lower than or equal to the preset inertial threshold, the inertial momentum factor of the working surface tends to be low, indicating that the current working surface is in a low-intensity or disordered state, and the execution resistance to interrupting the current operation is small.

10. A construction supervision instruction evaluation and optimization system, characterized in that, The evaluation and optimization system includes: Data acquisition module: used to collect inertial measurement unit data and ultra-wideband positioning data of each smart safety helmet at the construction site in real time, and to obtain the text of the supervision instructions to be evaluated at present; Group state feature extraction module: Based on the data from the inertial measurement unit, this module uses a frequency band energy decoupling method to decompose the group motion spectrum of each smart safety helmet into low-frequency energy proportion, mid-frequency energy proportion, and high-frequency energy proportion, combined with pre-established standard procedures. A frequency band template matrix is ​​used to generate a group inertial motion feature vector for each construction grid. The group inertial motion feature vector includes at least the inferred construction procedure for each construction grid, the confidence level corresponding to the inferred construction procedure, and the personnel density. The instruction parsing and matching degree evaluation module is used to perform structured parsing of the supervision instruction text, extract the target process and target construction section, map the target construction section to the corresponding construction grid set, calculate the consistency degree between the actual dominant process and the target process within the construction grid set based on the group inertial motion feature vector, and generate instructions by combining the personnel density of the construction grid set. On-site matching degree; Damping prediction module: used to calculate the execution damping coefficient based on the difficulty of switching between the inferred construction process and the target process, the current working surface inertial momentum of the construction grid set, and the historical instruction performance data of the corresponding construction unit; Optimize decision output module: used to perform the following based on the instructions. The combination of the on-site matching degree and the execution damping coefficient generates an optimized scheme for the supervision instructions and outputs the optimized supervision instructions.

Images