Building material carrying optimization system for multiple types of unmanned transportation forklifts

Abstract

The invention discloses a building material carrying optimization system for multi-type unmanned transportation forklifts, relates to the technical field of data processing, and solves the problems that safety and efficiency are difficult to balance and faults are difficult to position in building material carrying. The system comprises a data acquisition and preprocessing module for acquiring and preprocessing multi-dimensional data of a forklift and outputting standardized transportation data; the parameter calculation module is used for calculating a dumping index, an efficiency index and a balance factor based on the data; the primary speed regulation module is used for judging the risk type according to the dumping index and executing primary speed regulation in combination with the balance factor; the dynamic speed regulation module is used for updating parameters, mapping balance factors, calculating target speed regulation values in combination with risk types and dynamically adjusting strategies; the iterative optimization module is used for updating parameters to generate candidate speed regulation populations and screening optimal speed regulation parameters through crossover variation; and the fault diagnosis module is used for constructing a fault feature vector based on the speed regulation historical data and diagnosing a road surface problem or a fault of the forklift. The system dynamically balances safety and efficiency, accurately diagnoses faults, adapts to multiple types of unmanned transportation forklifts, and improves the carrying reliability and efficiency.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and more specifically, to an optimized system for handling building materials using various types of unmanned forklifts. Background Technology

[0002] With the rapid development of industrialized construction and intelligent warehousing logistics, unmanned forklifts, with their advantages of automation and high efficiency, have become core equipment for handling building materials and transferring warehouse goods. To address safety and efficiency issues during transportation, existing technologies have made numerous attempts, but there are still significant shortcomings in the in-depth exploration and comprehensive utilization of tilt-related indices, specifically in the following aspects: Chinese patent CN119706685B discloses a method and system for intelligent collaborative scheduling of transport forklifts for warehouse management. This patent obtains vibration amplitude and frequency through vibration sensors under the fork arms, calculates the tipping risk index (TRI) based on the relative height of the object's centroid, and constructs a constraint function based on TRI and the forklift's operating status (speed, acceleration, etc.), using a genetic algorithm to optimize the individual forklift's state. However, this patent only uses the tipping risk index as a safety constraint, failing to couple it with transportation efficiency indicators (such as actual transportation progress and unit energy consumption). It cannot dynamically balance safety redundancy and transportation efficiency based on tilt-related indices. Furthermore, the application of TRI is limited to individual forklift parameter optimization and does not extend to fault diagnosis, making it impossible to distinguish between road bumps and forklift malfunctions based on the trend of TRI changes.

[0003] Chinese patent CN120876602A proposes an external vision system for forklift operation. This system obtains cargo masks through a segmentation network, calculates the cargo tilt angle based on the geometric features of the masks, and triggers a forced deceleration command when the tilt angle is greater than 10°. However, in this patent, the tilt angle is only used as a single basis for risk alarm. It does not formulate differentiated speed adjustment strategies for different tilt angle ranges, nor does it link the tilt angle with efficiency indicators (such as task completion time) to construct a balancing factor. The application of tilt-related data is fragmented, only staying at the early warning level, and is not integrated into a closed-loop system of "exponential calculation - parameter optimization - status feedback", making it difficult to adapt to dynamic load scenarios.

[0004] In the Chinese patent with patent number CN115129068B, the disclosed intelligent positioning and navigation system based on AGV forklifts focuses on path adjustment and obstacle avoidance optimization based on the lateral area occupied by obstacles and the path overlap rate. It uses a weighing device to obtain the weight of the goods to calculate the transport power value to adjust the speed. However, this patent does not involve the calculation and application of any tilt-related indices (such as centroid offset angle and tipping risk index). It relies entirely on parameters such as weight and path to formulate scheduling strategies and cannot identify the potential tipping risk caused by the tilt of the goods. This poses a safety hazard in scenarios where the goods are stacked irregularly or the road surface is tilted.

[0005] Chinese patent CN116449853B proposes a path planning method for forklift AGVs. This method calculates the centroid offset angle of the moving polygon to obtain the overall offset degree and adjusts the path pheromone concentration based on the degree of dispersion. However, this patent only uses the centroid offset angle for path risk assessment, without coupling it with efficiency indices (such as transport volume per unit time) to construct a multi-objective decision-making model. Furthermore, it does not utilize historical tilt-related data for fault tracing; it only determines the impact range by matching the current degree of dispersion with historical images. It cannot determine whether the fault originates from uneven road surfaces or a forklift suspension system malfunction based on the magnitude of tilt angle changes, thus limiting the accuracy of fault diagnosis.

[0006] Furthermore, existing speed control optimization technologies often rely on fixed thresholds or single parameters. For example, the aforementioned patents may only rely on weight-based speed adjustments or only trigger warnings based on tilt angles. They lack a dynamic balancing mechanism that combines tilt indices and efficiency metrics, and do not introduce iterative optimization algorithms to continuously adapt speed control parameters. This makes it difficult to simultaneously meet the dual requirements of "low tilt risk" and "high transportation efficiency" in complex scenarios such as building material handling. Therefore, there is an urgent need for an unmanned forklift handling optimization method that fully utilizes tilt-related indices, integrates safety and efficiency goals, and achieves dynamic speed control and accurate fault diagnosis to overcome the shortcomings of existing technologies. Summary of the Invention

[0007] To address the aforementioned technical problems, this invention provides an optimized system for handling building materials using various types of unmanned forklifts, comprising: The data acquisition and preprocessing module is used to collect multi-dimensional data from unmanned transport forklifts and preprocess it to output standardized transport data. The parameter calculation module calculates the dumping index, efficiency index, and balance factor based on the transportation data; the initial speed adjustment module determines the risk type based on the dumping index and a preset threshold, and performs the initial speed adjustment in conjunction with the balance factor. The dynamic speed control module re-collects data to update parameters, maps the balance factor to a preset range, and calculates the target speed control value in combination with the risk type, and dynamically adjusts the speed control strategy. The iterative optimization module updates parameters based on real-time data, generates a population of candidate speed regulation values, and iteratively selects the optimal speed regulation parameter with the highest fitness through crossover and mutation. The fault diagnosis module constructs a fault feature vector based on historical data during the speed adjustment process, and diagnoses the fault type as road surface problems or faults of the unmanned transport forklift itself.

[0008] Furthermore, the data acquisition and preprocessing module collects multi-dimensional data including the real-time speed, load, tilt angle, and vibration data of the unmanned transport forklift. It performs data cleaning, normalization, outlier removal, and feature extraction preprocessing. The average speed, maximum tilt angle, and vibration standard deviation are obtained through feature extraction. The preprocessed multi-dimensional data is then combined to output standardized transport data.

[0009] Furthermore, the parameter calculation module extracts the center of gravity offset, tilt angle, and vibration amplitude from the transportation data, assigns preset weights to each parameter, and performs a weighted summation to obtain the tipping index.

[0010] Furthermore, the parameter calculation module extracts the average speed, energy consumption, and task completion time from the transportation data, calculates the ratio of average speed to energy consumption, and then calculates the logarithm of the reciprocal of the task completion time plus 1. The two are multiplied to obtain the efficiency index.

[0011] Furthermore, the parameter calculation module retrieves the dynamic adjustment coefficient from historical transportation data, calculates the ratio of the dumping index to the efficiency index, and multiplies it by the adjustment coefficient to obtain the balance factor.

[0012] Furthermore, the initial speed adjustment module presets low-risk and high-risk thresholds, compares the tipping index with the thresholds to determine the low, medium, and high risk types, and performs the initial speed adjustment in conjunction with the balance factor: low risk maintains the original speed, medium risk reduces the speed, and high risk stops the speed.

[0013] Furthermore, the dynamic speed regulation module re-collects transportation data to update relevant parameters and maps the balance factor to a preset range; it presets speed regulation benchmarks corresponding to three types of risks, and calculates the target speed regulation value by combining the balance factor range boundary value and proportional coefficient; it sets preset efficiency requirements and a safe range for the tipping index, and performs speed-up, further speed-down, or emergency braking operations based on whether the efficiency meets the standard and whether the tipping index is within the safe range. If the efficiency requirements are not met, the balance factor is recalculated and iteratively executed.

[0014] Furthermore, the iterative optimization module collects real-time transportation data to update relevant parameters, constructs a Gaussian distribution with the current target speed adjustment value as the mean and a preset standard deviation, and randomly generates an initial candidate speed adjustment value population; by calculating the sum of the reciprocal of the difference between the balance factor and the target benchmark value and the weighted value of the efficiency index, the fitness of each candidate value is obtained.

[0015] Furthermore, the iterative optimization module performs single-point crossover and Gaussian mutation operations on the initial population, calculates the fitness, and retains the optimal candidate value for iteration; when the termination condition is met, it outputs the optimal speed regulation parameter, controls the forklift operation and updates the parameter, and maintains the current parameter if the target is met.

[0016] Furthermore, the fault diagnosis module extracts fault-related data from historical data to construct a fault feature vector, assigns preset weights to each dimension, and inputs it into a pre-trained random forest model to output the probability of road surface problems and forklift malfunctions; when the probability exceeds a preset threshold, a corresponding alarm is triggered and fault information is recorded.

[0017] The positive and progressive effects of this invention are as follows: This invention achieves a dynamic balance between risk and efficiency by constructing a tipping index and an efficiency index and combining them with a dynamic adjustment coefficient to obtain a balance factor. This is further optimized by initial speed adjustment based on risk classification and interval mapping. A Gaussian distribution candidate population is constructed using the target speed adjustment value. Through cross-mutation and iterative screening and combined with real-time data updates, the accuracy of the speed adjustment parameters in adapting to dynamic working conditions is improved. At the same time, a multi-dimensional fault feature vector is constructed based on historical speed adjustment data. A pre-trained random forest model is used to accurately distinguish between road surface faults and forklift faults, reducing the misjudgment rate and clarifying maintenance directions, thus comprehensively ensuring the safety and efficiency of material handling operations. Attached Figure Description

[0018] Figure 1 This is a system schematic diagram of the present invention. Detailed Implementation

[0019] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0020] Reference Figure 1A building material handling optimization system for various types of unmanned forklifts includes a data acquisition and preprocessing module. This module collects and preprocesses multi-dimensional data from the unmanned forklift, outputting standardized transportation data. In practical applications, this module uses sensors integrated into key locations on the forklift for data acquisition, such as speed sensors mounted on the drive axle, weight sensors deployed on the fork carriage, and tilt sensors fixed at the vehicle's center of gravity. It also works with GPS and road surface detection equipment to obtain path parameters, ensuring multi-dimensional data coverage of the entire transportation process. The preprocessing stage also requires selecting appropriate filtering algorithms to remove noise based on data characteristics, such as using moving average filtering to process vibration frequency data, avoiding invalid information interfering with subsequent calculations. Through the precise deployment of multiple sensors at key locations, core data affecting transportation safety and efficiency can be comprehensively captured. Combined with targeted filtering and preprocessing operations, data quality is effectively improved, ensuring the accuracy of subsequent calculations of parameters such as tipping index and efficiency index. This provides reliable data support for the precise formulation of speed adjustment strategies, guaranteeing the effectiveness of risk assessment and efficiency optimization.

[0021] The parameter calculation module calculates the dumping index, efficiency index, and balance factor based on the transportation data. During the calculation process, the parameter weights need to be dynamically adjusted according to the specific transportation scenario. For example, when transporting fragile cement products, the weight of the center of gravity offset in the dumping index can be increased to 0.5, while when transporting sand and gravel with higher stability, this weight can be reduced to 0.3. The adjustment coefficients also need to refer to historical data of similar tasks to ensure that the balance factor can truly reflect the current balance between safety and efficiency. By dynamically adjusting the parameter weights according to the scenario, the dumping index can accurately match the safety risk characteristics of different materials, the efficiency index can objectively reflect the efficiency level of the current transportation conditions, and the design of retrieving adjustment coefficients based on historical data allows the balance factor to adapt to the dynamic needs of safety and efficiency in real time, thereby achieving a dynamic balance between risk and efficiency and avoiding the problem of excessive safety or insufficient efficiency caused by a single weight setting.

[0022] The initial speed adjustment module determines the risk type based on the tipping index and a preset threshold, and performs the initial speed adjustment in conjunction with the balance factor. The preset threshold needs to be set according to the forklift's load capacity and industry safety standards. For example, the low-risk threshold for a 1-ton forklift can be set to 0.3, and the high-risk threshold can be set to 0.7. The speed reduction operation under medium-risk conditions should adopt a gradient speed reduction method to avoid aggravating material shaking due to sudden deceleration. The speed reduction range can refer to the size of the balance factor; the larger the balance factor, the larger the speed reduction range. By setting the risk threshold according to the forklift's load capacity and industry standards, the scientific and safe nature of the risk type determination is ensured. The gradient speed reduction design not only avoids the secondary risks caused by sudden deceleration, but also achieves a balance between safety and efficiency through the correlation between the balance factor and the speed reduction range. This allows the initial speed adjustment to quickly respond to the risk level while reserving reasonable space for subsequent dynamic optimization, thus initially achieving a basic balance between safety and efficiency.

[0023] The dynamic speed control module re-acquires data and updates parameters, maps the balance factor to a preset range, and calculates the target speed adjustment value based on the risk type, dynamically adjusting the speed control strategy. It is recommended to set the data re-acquisition frequency to 5-10Hz to ensure timely parameter updates. The boundary values ​​of the preset balance factor range need to be determined through historical data statistics, for example, dividing the range into [0,0.5), [0.5,1.0), [1.0,1.5), and [1.5,+∞). The proportional coefficient corresponding to each range needs to be calibrated by forklift power testing. After the emergency braking protocol is triggered, in addition to sending an alarm, the current forklift position and status should be recorded simultaneously for subsequent fault tracing. Through high-frequency data updates of 5-10Hz, changes in operating conditions (such as sudden changes in road slope or load center of gravity shift) can be captured in real time. Combined with the range divisions and proportional coefficients calibrated by historical data statistics and power testing, the calculation of the target speed adjustment value is more closely aligned with the actual operating characteristics of the forklift, achieving dynamic adaptation of the speed control strategy. Simultaneously, the status recording design after emergency braking provides key scenario data for fault diagnosis, aiding in the accurate location of the fault's root cause.

[0024] The iterative optimization module updates parameters based on real-time data, generating a population of candidate speed regulation values. Through crossover and mutation iterations, the module selects the optimal speed regulation parameter with the highest fitness. It is recommended to set the number of candidate speed regulation values ​​to 20-50 to ensure population diversity while avoiding excessive computation. The reduction rate of crossover and mutation rates needs to be flexibly adjusted according to the iteration progress. For example, the crossover rate can be reduced by 0.1 every 10 iterations to ensure that the optimal solution is explored in the early stages and the stable solution is focused on in the later stages. The efficiency weight coefficient in the fitness calculation can be adjusted according to project priority; the weight is set to 0.3 in safety-first scenarios and 0.5 in efficiency-first scenarios. By reasonably setting the number of candidate populations, computational costs are controlled while ensuring search breadth. The dynamically adjusted crossover and mutation rates realize the optimization logic of "exploration in the early stages and convergence in the later stages." The adjustable efficiency weight coefficient in the fitness calculation allows the optimization process to accurately match the priority requirements of different scenarios. The final optimal speed regulation parameter can adapt to dynamically changing working conditions (such as load fluctuations and road surface changes) and balance the goals of safety and efficiency, significantly improving the adaptation accuracy of the speed regulation parameter.

[0025] The fault diagnosis module constructs fault feature vectors based on historical data during speed adjustment to diagnose fault types as road surface issues or forklift malfunctions. Historical data extraction must cover at least three months of speed adjustment records to ensure sufficient data to support feature vector construction. The random forest fault classification model needs to be trained in advance with a large amount of historical fault data, including typical cases such as road bumps, sensor malfunctions, and suspension system failures. The model accuracy must reach over 90% before deployment. After an alarm is triggered, fault information must be pushed to the corresponding maintenance personnel, and fault details must be stored in the database for subsequent analysis. By extracting historical data covering nearly three months, the fault feature vectors are ensured to comprehensively include typical features of various fault scenarios. The high-accuracy random forest model trained on a large number of labeled fault cases can accurately identify the feature differences between road surface issues and forklift malfunctions, effectively reducing the false positive rate. Real-time push and detailed storage of fault information provide clear guidance for subsequent maintenance and accumulate data for model iteration and optimization, enabling precise location and rapid response to the root cause of the fault.

[0026] Furthermore, the specific working process of the data acquisition and preprocessing module is as follows: The acquired multi-dimensional data includes the real-time speed, load weight, road slope, turning radius, vibration frequency, and battery power of the unmanned transport forklift; during acquisition, it is necessary to ensure that the sensor sampling frequency is consistent, usually set to 10Hz, to avoid data timestamp misalignment. Preprocessing operations such as data cleaning, normalization, outlier removal, and feature extraction are performed on the acquired multi-dimensional data; during the data cleaning stage, null values ​​and garbled characters caused by instantaneous sensor failures are removed; normalization uses min-max standardization to transform the data to the [0,1] interval; outlier removal uses the 3σ criterion to identify extreme data; during feature extraction, the average speed is calculated using a 5-second time window, the maximum tilt angle is taken as the peak value during transportation, and the vibration standard deviation is derived using the variance formula; the average speed, maximum tilt angle, and vibration standard deviation are calculated through feature extraction, and combined with the preprocessed multi-dimensional data, standardized transportation data is output. The standardized data needs to be stored in a preset format for easy retrieval by subsequent modules. By unifying the sensor sampling frequency to avoid data timing misalignment, and combining multi-step preprocessing operations to remove invalid interference and unify data units, the standardized output data is ensured to be accurate, consistent, and usable. Targeted feature extraction design (such as calculating the average speed within a time window and taking the peak value as the maximum tilt angle) can accurately extract the core features that affect safety and efficiency, providing high-quality data input for subsequent parameter calculation and speed adjustment optimization, and indirectly improving the overall system adaptation accuracy and balance effect.

[0027] Furthermore, the specific process of calculating the tipping index by the parameter calculation module is as follows: extract the center of gravity offset, the tilt angle of the unmanned transport forklift, and the vibration amplitude from the transportation data; the center of gravity offset needs to be calculated collaboratively using data from the weight sensor and the tilt angle sensor to reflect the distance of the cargo's center of gravity from the forklift's center, and assign preset weights to the center of gravity offset, the tilt angle of the unmanned transport forklift, and the vibration amplitude respectively; the weight allocation needs to be determined by experts in conjunction with historical failure cases, for example, the weight of the center of gravity offset is 0.4-0.6, the weight of the tilt angle is 0.3-0.4, and the weight of the vibration amplitude is 0.1-0.2, with a total weight sum of 1. Multiply each parameter by its corresponding weight and sum them to obtain the tipping index. After the calculation is completed, it needs to be compared with historical data from the same period to verify the rationality of the index. By using the combined calculations of weight and tilt sensors, the accuracy of the center of gravity offset is ensured. The weighting based on expert experience and historical failure cases allows the tilt index to highlight the impact of key safety risk factors (such as center of gravity offset). The comparison and verification with historical data further ensures the reliability of the index, making the risk type determination accurate and laying a solid foundation for the dynamic balance between safety and efficiency.

[0028] Furthermore, the specific process of calculating the efficiency index by the parameter calculation module is as follows: Average speed, energy consumption, and task completion time are extracted from the transportation data; energy consumption data is converted from the forklift battery management system; task completion time is accumulated from the task start time, and the ratio of average speed to energy consumption is calculated; this ratio reflects the transportation efficiency per unit of energy consumption, with a higher value indicating higher efficiency. Then, the logarithm of the reciprocal of the task completion time plus one is calculated; logarithmic processing smooths the impact of fluctuations in task completion time, avoiding extreme time values ​​that could cause index anomalies. The two calculation results are multiplied to obtain the efficiency index, which needs to be calibrated periodically to ensure consistency with actual transportation efficiency. By integrating three core efficiency indicators—average speed, energy consumption, and task completion time—and combining logarithmic processing to smooth the impact of extreme values, the efficiency index can objectively and stably reflect the actual transportation efficiency level. The periodic calibration mechanism ensures the long-term effectiveness of the index, providing an accurate efficiency reference for the calculation of the balance factor and helping to achieve a dynamic balance between safety and efficiency.

[0029] Furthermore, the specific process of obtaining the balance factor by the parameter calculation module is as follows: Dynamically selectable adjustment coefficients are retrieved from historical transportation data. These adjustment coefficients need to be determined based on the current task priority and scenario complexity; for example, the adjustment coefficient for emergency tasks is 1.2-1.5, and for routine tasks it is 1.0-1.2. The ratio of the dumping index to the efficiency index is calculated. This ratio directly reflects the current relationship between safety and efficiency; the larger the ratio, the higher the relative safety risk. This ratio is multiplied by the retrieved adjustment coefficients to obtain the balance factor. After calculation, the balance factor needs to be judged to see if it is within a reasonable range; if it exceeds the range, the adjustment coefficients need to be retrieved again. By dynamically retrieving adjustment coefficients based on task priority and scenario complexity, the balance factor can accurately adapt to the core needs of different scenarios (e.g., emergency tasks prioritize efficiency, while complex scenarios prioritize safety). The design of the ratio of the dumping index to the efficiency index directly quantifies the current trade-off between safety and efficiency. The balance factor obtained by multiplying the two provides a clear balancing basis for subsequent speed adjustment strategies, ensuring that speed adjustment operations can respond to safety risks without neglecting efficiency goals.

[0030] Furthermore, the specific working process of the initial speed adjustment module is as follows: Preset low-risk and high-risk thresholds; the threshold setting needs to refer to the forklift design parameters and industry safety standards. Different types of forklifts have different thresholds; for example, the low-risk threshold for a stacker forklift is 0.2, and the high-risk threshold is 0.6. Compare the calculated tipping index with the low-risk and high-risk thresholds to determine the risk type: a tipping index below the low-risk threshold indicates low risk, between the low-risk and high-risk thresholds indicates medium risk, and above the high-risk threshold indicates high risk. During the determination process, attention must be paid to boundary value handling; for example, an index equal to the low-risk threshold is classified as low risk. Combine the calculated balance factor and the determined risk type to perform the initial speed adjustment: in the low-risk state, control the unmanned transport forklift to maintain its original operating speed; at this time, the efficiency index needs to be monitored in real time to ensure that efficiency meets the standard. In the medium-risk state, perform a speed reduction operation; the speed reduction is usually 20%-30% of the original speed, specifically fine-tuned according to the balance factor. In the high-risk state, perform a speed stop operation; after stopping, check the condition of the goods and the forklift, and restart transport only after eliminating the risk. By referencing forklift design parameters and industry standards to set differentiated thresholds, the professionalism and safety of risk type determination are ensured, and the clear handling of boundary values ​​avoids ambiguity in the determination. Efficiency monitoring in low-risk conditions, fine-tuning speed reduction combined with balancing factors in medium-risk conditions, and speed stop investigation in high-risk conditions realize the initial speed adjustment logic of "risk-level response and dynamic efficiency consideration", which can quickly control safety risks and minimize unnecessary efficiency losses.

[0031] Furthermore, the specific working process of the dynamic speed adjustment module is as follows: The transport data of the unmanned forklift is re-collected, and the balance factor, tipping index, efficiency index, and risk type are updated based on this transport data. The re-collection time interval needs to be adjusted according to the road surface complexity; for complex roads, it is shortened to 5 seconds, and for simple roads, it is extended to 10 seconds. Multiple balance factor interval boundary values ​​are preset, and the updated balance factor is mapped to the predefined interval. The interval division must ensure coverage of all possible balance factor values. Each interval corresponds to a different adjustment tendency. Three types of risk types are preset with corresponding speed adjustment benchmarks: low risk corresponds to a regular speed adjustment benchmark, which is usually 90% of the forklift's rated speed; medium risk corresponds to a moderate speed reduction benchmark, which is 60%-80% of the rated speed; and high risk corresponds to a deep speed reduction benchmark, which is 30%-50% of the rated speed. The adjustment amount is calculated based on the boundary value of the interval to which the balance factor belongs, combined with a preset proportional coefficient. The proportional coefficient needs to be determined through forklift power response testing and is usually 0.1-0.3. The speed adjustment benchmark and adjustment amount are then combined. The values ​​are summed to obtain the target speed adjustment value. The target speed adjustment value must be within the forklift speed adjustment range. If it exceeds the range, the boundary value is taken. Preset requirements and preset safety ranges are set: the preset requirement is a preset percentage where the efficiency index is not lower than the historical average; the preset percentage is usually 80%-90% and can be adjusted according to project needs. The safety range is the range where the tipping index is lower than the low-risk threshold. If the updated efficiency index does not meet the preset requirements but the updated tipping index is within the safety range, the speed increase operation is performed according to the target speed adjustment value. The speed increase adopts a gradient method to avoid sudden speed changes. If the updated tipping index exceeds the safety range, the speed decrease operation is performed according to the target speed adjustment value. After the speed decrease, the tipping index needs to be continuously monitored until it returns to the safety range. If the risk type is high risk and the balance factor exceeds the preset critical value, the emergency braking protocol is triggered and an alarm is sent to the monitoring system. The emergency braking protocol needs to cut off the forklift power output and activate the parking brake at the same time. If the efficiency index still does not meet the preset requirements after the speed increase operation, the balance factor is recalculated and the above operation is iterated. The number of iterations usually does not exceed 5 times to avoid falling into an infinite loop. By dynamically adjusting the data acquisition interval based on road surface complexity, parameter updates are ensured to accurately match the pace of changing operating conditions. The proportional coefficient and interval division calibrated through power testing make the calculation of the target speed adjustment value more closely match the actual power response characteristics of the forklift. Operations such as gradient acceleration and continuous deceleration monitoring avoid the safety risks caused by sudden speed changes. Based on the dual-dimensional judgment of efficiency achievement and safety range, a dynamic speed adjustment logic of "not breaking the safety bottom line and not abandoning the efficiency target" is realized, allowing the speed adjustment strategy to adapt to changes in operating conditions in real time and continuously optimize the balance between safety and efficiency.

[0032] Furthermore, the specific process of generating a candidate speed adjustment value population and calculating fitness in the iterative optimization module is as follows: Real-time transportation data of the unmanned forklift is collected, and the balance factor, efficiency index, tipping index, and risk type are updated. Real-time data collection must ensure no delay to avoid affecting the timeliness of parameter updates. A Gaussian distribution is constructed using the current target speed adjustment value as the mean and a preset standard deviation is set. The preset standard deviation is typically 10%-20% of the mean to ensure that candidate values ​​are distributed around the current optimal value. Multiple candidate speed adjustment values ​​are randomly generated within this Gaussian distribution range to form the initial candidate speed adjustment value population. The number of candidate values ​​is recommended to be 20-50; too few may miss the optimal solution, while too many will increase the computational load. The preset target balance factor... The calculation includes a sub-benchmark value, a minimum constant to prevent division by zero, and an efficiency weighting coefficient. The target balance factor benchmark value needs to be determined based on industry best practices. The minimum constant is usually 1e-6 to avoid division by zero errors in the calculation. The efficiency weighting coefficient is set according to the priority of safety and efficiency, ranging from 0.3 to 0.5. The absolute value of the difference between the balance factor and the target balance factor benchmark value is calculated and the reciprocal of the minimum constant is added. This part reflects how close the candidate value is to the balance factor target. The larger the value, the better. Then, the product of the efficiency index and the efficiency weighting coefficient is calculated. This part reflects the efficiency performance of the candidate value. The two calculation results are added together to obtain the fitness of each candidate speed regulation value. The higher the fitness, the better the candidate value can balance the safety and efficiency targets. By acquiring real-time data without delay, parameter updates are ensured to be completely synchronized with the current operating conditions. A Gaussian distribution design is constructed around the current target speed adjustment value, allowing the candidate value search to focus on the vicinity of the optimal solution, balancing search efficiency and accuracy. The fitness calculation integrates "balance factor proximity" and "efficiency performance," and adapts the scenario priority through weight coefficients, so that the evaluation criteria of the candidate values ​​are fully aligned with the core objectives of the system. This provides a scientific basis for subsequent iterations to select the optimal parameters that balance safety and efficiency and adapt to dynamic operating conditions.

[0033] Furthermore, the specific process of crossover, mutation, and iterative screening performed by the iterative optimization module is as follows: A single-point crossover operation is performed on the initial candidate speed-adjustment value population, setting an initial crossover rate and gradually decreasing it with each iteration. The initial crossover rate is typically 0.8-0.9, decreasing by 0.1 every 10 iterations, and remaining around 0.5 in the later stages to ensure early exploration and later stability. A Gaussian mutation operation is then performed on the crossover population, setting an initial mutation rate and gradually decreasing it with each iteration. The initial mutation rate is 0.1-0.2, decreasing to below 0.05 in the later stages of iteration to avoid destroying the already found optimal solutions. The fitness of each candidate speed-adjustment value in the mutated population is calculated, and a preset number of candidate values ​​with the highest fitness are retained. This preset number is typically 50% of the initial population size to ensure population quality. The iterative process continues... The process involves crossover, mutation, and fitness calculation. After each iteration, the fitness change is recorded. The iteration terminates when the fitness change rate is less than a preset minimum or the preset total number of iterations is reached after two consecutive iterations. The preset minimum is typically 0.01, and the total number of iterations is 50-100. The candidate value with the highest fitness is output as the optimized speed control parameter. Before outputting, it is necessary to verify whether the parameter meets the forklift's hardware limitations. The unmanned transport forklift is controlled based on the optimized speed control parameter. Transport data is re-collected and relevant parameters are updated. After re-collection, it is necessary to quickly determine whether the parameters are suitable for the current working conditions. If the updated risk type is low risk and the efficiency index meets the preset requirements, the unmanned transport forklift continues to perform transport operations according to the current operating parameters. If the requirements are not met, the process returns to regenerate the candidate population and iterates again for optimization. By dynamically reducing the crossover mutation rate, a smooth transition from "extensive exploration" to "precise convergence" in the optimization process is achieved, avoiding getting trapped in local optima. The design of retaining the best candidate value in each iteration ensures continuous improvement in population quality. The explicit setting of the iteration termination condition balances optimization accuracy and computational efficiency. Hardware constraint verification before output and working condition adaptation judgment after iteration ensure that the optimized speed regulation parameters not only meet the actual operating capabilities of the forklift but also continuously adapt to dynamic working condition changes, ultimately achieving a significant improvement in speed regulation parameter adaptation accuracy and long-term stability in the balance between safety and efficiency.

[0034] Furthermore, the specific working process of the fault diagnosis module is as follows: Fault diagnosis-related data is extracted from historical data of the speed adjustment process, including vibration frequencies before and after speed reduction, operational stability data under different road conditions, performance data under the same load, and records of abnormal occurrences on specific road sections. Data extraction must cover at least the complete transportation cycle of the past three months to ensure the inclusion of data from various fault scenarios. A fault feature vector is constructed based on the extracted data. This fault feature vector includes the rate of change of vibration frequency before and after speed reduction, the stability difference coefficient under different road conditions, the performance deviation of each unmanned transport forklift under the same load, and the frequency of abnormal occurrences on specific road sections. Each feature dimension needs to be standardized to avoid the influence of units, and preset weights are assigned to different dimensions in the fault feature vector. The weights are based on… The distinguishability of features for faults is set, for example, the vibration frequency change rate is weighted at 0.3-0.4, and the stability difference coefficient is weighted at 0.2-0.3. The weighted fault feature vector is then input into a pre-trained random forest fault classification model. The model needs to be trained in advance with a large amount of labeled historical fault data, and can only be used when the test accuracy reaches more than 90%. It outputs the probabilities of two fault types: road surface problems and faults of the unmanned transport forklift itself. The output probabilities need to be retained to two decimal places to facilitate the determination of fault attribution. A fault probability threshold is preset, usually 85%. When the probability of any fault type exceeds this threshold, the corresponding fault alarm is triggered and the fault occurrence time, location, and feature data are recorded. The alarm information needs to be pushed to the corresponding maintenance personnel's terminal, and the recorded data needs to be stored in the fault database for subsequent analysis and model optimization. By extracting historical data covering the entire transportation cycle, the fault feature vectors are designed to comprehensively capture the differentiated characteristics of road surface problems and forklift malfunctions (e.g., road surface problems are often associated with specific road sections, while forklift malfunctions are often unrelated to load). Standardized processing and discriminative weight allocation enhance the recognizability of the feature vectors. Combined with a high-accuracy pre-trained random forest model, the core differences between the two types of faults can be accurately identified, effectively reducing the false positive rate. Real-time push notifications and database storage of fault information provide maintenance personnel with clear fault location information and accumulate valuable data for subsequent model iterations and optimizations, enabling precise fault location and continuous improvement of system performance.

[0035] Based on the above, this invention further provides a method for optimizing the handling of building materials using various types of unmanned forklifts, comprising the following steps: S1: Collecting and preprocessing multi-dimensional data of the unmanned forklift to obtain transportation data; through multi-dimensional data collection and standardized preprocessing, a high-quality data foundation is provided for subsequent parameter calculation and strategy formulation, ensuring the accuracy of system decisions. S2: Calculating the tipping index and efficiency index based on the transportation data, and obtaining a balance factor based on the tipping index and efficiency index; determining the risk type based on the tipping index and a preset threshold; the risk type includes low, medium, and high risk; adjusting the speed of the unmanned forklift initially according to the balance factor and risk type, specifically maintaining the original speed in low-risk conditions, reducing the speed in medium-risk conditions, and stopping the speed in high-risk conditions; by quantifying safety and efficiency indicators and constructing a balance factor, combined with risk classification, the initial speed adjustment is achieved, quickly establishing a basic balance between safety and efficiency, and responding to core risks. S3: After slowing down, reacquire the transportation data of the unmanned forklift, and update the balance factor and risk type based on the reacquired transportation data; divide the updated balance factor into different intervals, and calculate the specific target speed adjustment value in combination with the speed adjustment benchmark corresponding to the risk type; then adjust the speed adjustment strategy accordingly, specifically: if the updated efficiency index does not meet the preset requirements and the updated tipping index is within the preset safety range, execute the speed increase operation according to the calculated target speed adjustment value; if the updated tipping index exceeds the safety range, execute the further speed reduction operation according to the calculated target speed adjustment value; by updating parameters in real time and dynamically adjusting the speed adjustment strategy, a rapid response to changes in working conditions can be achieved, optimizing efficiency performance without exceeding the safety bottom line. S4: Collect real-time transportation data from the unmanned forklift, update the balance factor, efficiency index, tipping index, and risk type; generate a candidate speed adjustment value population centered on the current target speed adjustment value; calculate the fitness of each candidate speed adjustment value based on the deviation between the balance factor and the target balance factor benchmark value; perform crossover and mutation operations on the candidate speed adjustment value population, iteratively selecting the candidate value with the highest fitness; output the optimized speed adjustment parameters; control the unmanned forklift operation based on the optimized speed adjustment parameters, re-collect transportation data and update the balance factor, efficiency index, tipping index, and risk type; if the updated risk type is low risk and the efficiency index meets the preset requirements, the unmanned forklift continues to perform transportation operations according to the current operating parameters; select the optimal speed adjustment parameters through iterative optimization algorithms to ensure that the parameters can accurately adapt to dynamic working conditions and continuously maintain the optimal balance between safety and efficiency. S5: Based on historical data during the speed adjustment process, perform fault diagnosis to determine whether the cause of the anomaly is a road problem or a fault in the unmanned forklift itself; by constructing a multi-dimensional fault feature and a high-accuracy classification model, accurately locate the root cause of the fault, provide clear guidance for rapid maintenance, and ensure the continuity of transportation operations.

[0036] The above details the specific implementation of the optimization method for handling building materials using various types of unmanned forklifts. The refinement of each step can be flexibly adjusted according to the actual operating scenario. For example, in special environments such as high temperature and high humidity, it is necessary to increase sensor protection measures and adjust the filtering intensity of data preprocessing to ensure that the method can adapt to different forklift types and building material handling needs. Through the collaborative design of various modules and steps, the system can achieve three core effects: First, by dynamically adjusting the balance factor, risk classification mechanism, and threshold and weight settings adapted to the scenario, a dynamic balance between risk and efficiency is achieved, ensuring maximum efficiency under low risk and controllable safety under medium and high risk; Second, by frequently updating data, iteratively optimizing algorithms, and calibrating parameters that fit the characteristics of forklifts, the accuracy of speed adjustment parameters in adapting to dynamic working conditions such as load changes and road surface fluctuations is significantly improved; Third, by comprehensively extracting fault features and using a high-accuracy classification model, the system can accurately distinguish between road problems and forklift malfunctions, reduce the misjudgment rate, clarify maintenance directions, and ultimately ensure the safe, efficient, and stable operation of building material handling.

[0037] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A building material handling optimization system for various types of unmanned forklifts, characterized in that, include: The data acquisition and preprocessing module is used to collect multi-dimensional data from unmanned transport forklifts and preprocess it to output standardized transport data. The parameter calculation module calculates the dumping index, efficiency index, and balance factor based on the transportation data. The initial speed adjustment module determines the risk type based on the tipping index and a preset threshold, and performs the initial speed adjustment in conjunction with the balance factor. The dynamic speed control module re-collects data to update parameters, maps the balance factor to a preset range, and calculates the target speed control value in combination with the risk type, and dynamically adjusts the speed control strategy. The iterative optimization module updates parameters based on real-time data, generates a population of candidate speed regulation values, and iteratively selects the optimal speed regulation parameter with the highest fitness through crossover and mutation. The fault diagnosis module constructs a fault feature vector based on historical data during the speed adjustment process, and diagnoses the fault type as road surface problems or faults of the unmanned transport forklift itself.

2. The building material handling optimization system for multiple types of unmanned forklifts according to claim 1, characterized in that, The data acquisition and preprocessing module collects multi-dimensional data including the real-time speed, load, tilt angle, and vibration data of the unmanned transport forklift. It performs data cleaning, normalization, outlier removal, and feature extraction preprocessing. The average speed, maximum tilt angle, and vibration standard deviation are obtained through feature extraction. The preprocessed multi-dimensional data is combined to output standardized transport data.

3. The building material handling optimization system for multiple types of unmanned forklifts according to claim 1, characterized in that, The parameter calculation module extracts the center of gravity offset, tilt angle and vibration amplitude from the transportation data, assigns preset weights to each parameter and sums them up to obtain the tipping index.

4. The building material handling optimization system for multiple types of unmanned forklifts according to claim 1, characterized in that, The parameter calculation module extracts average speed, energy consumption, and task completion time from the transportation data, calculates the ratio of average speed to energy consumption, and then calculates the logarithm of the reciprocal of the task completion time plus 1. The two are multiplied to obtain the efficiency index.

5. The building material handling optimization system for multiple types of unmanned forklifts according to claim 1, characterized in that, The parameter calculation module retrieves the dynamic adjustment coefficient from historical transportation data, calculates the ratio of the dumping index to the efficiency index, and multiplies it by the adjustment coefficient to obtain the balance factor.

6. The building material handling optimization system for multiple types of unmanned forklifts according to claim 1, characterized in that, The initial speed adjustment module presets low-risk and high-risk thresholds, compares the tipping index with the thresholds to determine the low, medium, or high risk type, and performs the initial speed adjustment in conjunction with the balance factor: low risk maintains the original speed, medium risk reduces the speed, and high risk stops the speed.

7. The building material handling optimization system for multiple types of unmanned forklifts according to claim 1, characterized in that, The dynamic speed control module re-collects transportation data, updates relevant parameters, and maps the balance factor to a preset range; The system presets speed adjustment benchmarks corresponding to three types of risks, and calculates the target speed adjustment value by combining the boundary values ​​of the balance factor range and the proportional coefficient. It sets preset efficiency requirements and a safe range for the tipping index. Based on whether the efficiency meets the standards and whether the tipping index is within the safe range, it performs speed increase, further speed reduction, or emergency braking operations. If the efficiency requirements are not met, the balance factor is recalculated and iteratively executed.

8. The building material handling optimization system for multiple types of unmanned forklifts according to claim 1, characterized in that, The iterative optimization module collects real-time transportation data to update relevant parameters, constructs a Gaussian distribution with the current target speed adjustment value as the mean and a preset standard deviation, and randomly generates an initial candidate speed adjustment value population. The fitness of each candidate value is obtained by calculating the sum of the reciprocal of the difference between the balance factor and the target benchmark value and the efficiency index weighted value.

9. The building material handling optimization system for multiple types of unmanned forklifts according to claim 1, characterized in that, The iterative optimization module performs single-point crossover (crossover rate decreases with the number of iterations) and Gaussian mutation (mutation rate decreases with the progress of iteration) on the initial population, calculates fitness and retains the optimal candidate value for iteration; when the termination condition is met, it outputs the optimal speed regulation parameter, controls the forklift operation and updates the parameter, and maintains the current parameter if the target is met.

10. The building material handling optimization system for multiple types of unmanned forklifts according to claim 1, characterized in that, The fault diagnosis module extracts fault-related data from historical data to construct a fault feature vector, assigns preset weights to each dimension, and inputs it into a pre-trained random forest model to output the probability of road surface problems and forklift malfunctions. When the probability exceeds a preset threshold, the module triggers a corresponding alarm and records the fault information.

Images