Construction method and device of bucket load spectrum generation model, equipment and medium

Abstract

The invention discloses a construction method and device of a bucket load spectrum generation model, equipment and a medium. The method comprises the steps that a physical test bed is controlled to apply multi-dimensional dynamic loads to all stress point positions of the hydraulic excavator, and the hydraulic excavator is controlled to execute simulation operation according to working condition types; instantaneous loads acting on the bucket after the loads are transmitted to all the stress point positions are recorded in real time, and all the instantaneous loads are gathered to form a three-dimensional load spectrum of the bucket; collecting dynamic response data generated by each stress point position, associating the three-dimensional load spectrum with the dynamic response data to form training sample pairs, aggregating the training sample pairs to form a training data set, performing characterization processing, inputting the training data set into an original deep learning model for iterative training, and obtaining a three-dimensional load spectrum; and taking the deep learning model meeting the iteration termination condition as a hydraulic excavator bucket load spectrum generation model. According to the embodiment of the invention, the virtual rack is constructed by a physical test rack of a small number of excavator samples, and a basis is provided for efficiently generating reliable bucket load spectrums for batch excavators.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of Internet of Things, and in particular to a construction method and device of a bucket load spectrum generation model, equipment and a medium. BACKGROUND

[0002] The bucket stress load spectrum of a hydraulic excavator is a key boundary input for structure life simulation analysis, and its accuracy directly affects the reliability of excavator structure fatigue life evaluation. The traditional method of obtaining the bucket stress load spectrum mainly relies on physical testing or theoretical calculation: physical testing requires deploying a large number of stress patches on the boom, dipper arm and other structural parts, and combining with the fatigue test bench to obtain the load spectrum through load application and strain matching; the theoretical calculation is based on the establishment of a torque balance equation to solve the equation based on the cylinder pressure, displacement and spatial position relationship.

[0003] However, the existing technology has significant limitations in obtaining the bucket stress load spectrum: ① the physical testing scheme based on the fatigue test bench usually requires deploying more than tens of stress patches, and needs to build a special test bench for load application and strain matching, which has the problems of high test cost and long test cycle, and is difficult to meet the rapid testing needs of batch models; ② the theoretical calculation scheme based on the torque balance equation is convenient to operate and has high efficiency, but generally does not consider the bucket partial load working condition, and the idealized assumption conditions and simplified parameters are introduced in the equation establishment process, resulting in a large deviation between the solving result and the actual load, which cannot meet the requirements of high-precision life simulation. With the increasing requirements of engineering on the efficiency and cost control of load spectrum acquisition, there is an urgent need for a bucket stress load spectrum generation method that takes into account accuracy and economy. SUMMARY

[0004] Therefore, the present application provides a construction method, device, equipment and medium of a bucket load spectrum generation model to solve the problems of high cost and long cycle of physical test bench and large deviation of theoretical calculation in traditional bucket load spectrum acquisition.

[0005] In a first aspect, the present application provides a construction method of a bucket load spectrum generation model, comprising:

[0006] In response to a test instruction issued by the management end for the target hydraulic excavator, the target physical fatigue test bench is controlled to apply a known multi-dimensional dynamic load to each force point of the target hydraulic excavator, and the target hydraulic excavator is controlled to perform simulated operation according to the working condition type; wherein the test instruction includes: excavator model, a plurality of force points, multi-dimensional dynamic load corresponding to each force point, a plurality of working condition types and time windows corresponding to each working condition type;

[0007] The test load spectrum generation module is configured to record, in real time, the multi-dimensional dynamic load applied to each stress point, record the instantaneous load acting on the bucket after the multi-dimensional dynamic load is transmitted, arrange all the instantaneous loads in the time window in time sequence, and form a three-dimensional load spectrum of the bucket.

[0008] The load and response correlation module is configured to synchronously collect dynamic response data corresponding to the working condition type generated by each stress point due to the execution of the simulation operation and the multi-dimensional dynamic load borne by each stress point, correlate the three-dimensional load spectrum with the dynamic response data, form a training sample pair containing a mapping relationship between the load and the response, and aggregate the training sample pairs of each hydraulic excavator sample in the test system to form a training data set.

[0009] The load spectrum generation model construction module is configured to perform feature processing on the training data set, input the training data set after the feature processing into an original deep learning model for iterative training, and use the deep learning model meeting an iterative termination condition as a hydraulic excavator bucket load spectrum generation model.

[0010] In a second aspect, an embodiment of the present application further provides a construction device of a bucket load spectrum generation model, including:

[0011] The load application and simulation operation module is configured to, in response to a test instruction for a target hydraulic excavator issued by a management end, control a target physical fatigue test bench to apply a known multi-dimensional dynamic load to each stress point of the target hydraulic excavator, and control the target hydraulic excavator to execute a simulation operation according to a working condition type. The test instruction includes: a model of the excavator, a plurality of stress points, a plurality of multi-dimensional dynamic loads corresponding to the stress points, a plurality of working condition types, and a time window corresponding to each working condition type.

[0012] The test load spectrum generation module is configured to record, in real time, the multi-dimensional dynamic load applied to each stress point, record the instantaneous load acting on the bucket after the multi-dimensional dynamic load is transmitted, arrange all the instantaneous loads in the time window in time sequence, and form a three-dimensional load spectrum of the bucket.

[0013] The load and response correlation module is configured to synchronously collect dynamic response data corresponding to the working condition type generated by each stress point due to the execution of the simulation operation and the multi-dimensional dynamic load borne by each stress point, correlate the three-dimensional load spectrum with the dynamic response data, form a training sample pair containing a mapping relationship between the load and the response, and aggregate the training sample pairs of each hydraulic excavator sample in the test system to form a training data set.

[0014] The load spectrum generation model construction module is configured to perform feature processing on the training data set, input the training data set after the feature processing into an original deep learning model for iterative training, and use the deep learning model meeting an iterative termination condition as a hydraulic excavator bucket load spectrum generation model.

[0015] In a third aspect, an electronic device is provided, and the electronic device includes at least one processor, and a memory connected with the at least one processor in communication; the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to perform the method for constructing a bucket load spectrum generation model according to any one of the embodiments of the present application.

[0016] In a fourth aspect, a computer readable storage medium is provided, and the computer readable storage medium stores computer instructions for enabling a processor to implement the method for constructing a bucket load spectrum generation model according to any one of the embodiments of the present application.

[0017] According to the embodiments of the present application, the physical fatigue test bench and the excavator sample are used to cooperatively collect the associated sample of the dynamic response data-three-dimensional load spectrum, and the deep learning is combined to train the generation model, so that the accuracy of the physical test is retained, the high cost and long period of repeated test of each device in the traditional bench test are avoided, the deviation caused by the idealized assumption in the theoretical calculation is overcome, the bucket load spectrum generation model with high precision is efficiently constructed at a low cost, and reliable support is provided for the load spectrum acquisition and structure life simulation of the batch excavator.

[0018] It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of the present application, nor is it used to limit the scope of the present application. Other features of the present application will become apparent from the following description. BRIEF DESCRIPTION OF DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0020] Figure 1 is a flow chart of a method for constructing a bucket load spectrum generation model according to an embodiment of the present application;

[0021] Figure 2 is a flow chart of another method for constructing a bucket load spectrum generation model according to an embodiment of the present application

[0022] Figure 3 is a structural schematic diagram of a construction device for a bucket load spectrum generation model according to an embodiment of the present application;

[0023] Figure 4It is a structural schematic diagram of an electronic device for implementing a method for constructing a bucket load spectrum generation model according to an embodiment of the present application. DETAILED DESCRIPTION

[0024] In order for those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should fall within the scope of protection of the present application.

[0025] It should be noted that the terms "first", "second", and the like in the specification and claims of the present application and the above-described drawings are used to distinguish similar objects, and do not necessarily have to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0026] Embodiment one

[0027] Figure 1 A flowchart of a method for constructing a bucket load spectrum generation model according to Embodiment One of the present application is provided, which can be applicable to the research and development or production stage of a hydraulic excavator. In the case of needing to quickly obtain the bucket load spectrum of different models of excavators under various working conditions to support the fatigue life simulation analysis of structural parts, the method can be executed by a construction device for a bucket load spectrum generation model. The device can be realized in the form of hardware and / or software, and can be configured in a test system including at least one physical fatigue test bench and at least one hydraulic excavator sample. As shown in the figure, the method comprises: Figure 1

[0028] S110, in response to the test instruction for the target hydraulic excavator issued by the management end, controlling the target physical fatigue test bench to apply a known multi-dimensional dynamic load to each stress point of the target hydraulic excavator, while controlling the target hydraulic excavator to perform simulated operation according to the working condition type; wherein the test instruction includes: excavator model, multiple stress points, multi-dimensional dynamic load corresponding to each stress point, multiple working condition types, and time window corresponding to each working condition type.

[0029] ​In the present embodiment, the test system refers to a hardware environment comprising at least one physical fatigue test bench and at least one hydraulic excavator sample for collecting training data of the load spectrum generation model. The physical fatigue test bench refers to a test device uniquely matched with a specific model of hydraulic excavator, which can apply a preset dynamic load to the force point of the excavator to simulate the stress state in actual operation. The model-specific characteristics ensure the loading accuracy and device matching. The target hydraulic excavator refers to a specific model of excavator for which the load spectrum generation model needs to be constructed. The test instruction is a control instruction issued by the management end, which contains the core parameters required for the test and is the input signal to drive the test process. The force point refers to the key structural position of the hydraulic excavator that bears the load (such as the connection point of the boom and stick, the hinge point of the stick and bucket, etc.), and the force transmission at these positions directly affects the load state of the bucket. The multi-dimensional dynamic load refers to the load applied along the three-dimensional space direction that changes with time, simulating the complex force system action in the excavator operation. The working condition type refers to the typical operation mode of the excavator, and different working conditions correspond to different operation actions and load characteristics. The time window refers to the continuous test duration set for a certain working condition type, which is used to limit the time range of data collection under a single working condition and ensure the time sequence integrity of the data.

[0030] The management end generates a test instruction containing the excavator model, force point, multi-dimensional dynamic load parameter, working condition type, and time window according to the model and test requirements of the target hydraulic excavator. After receiving the instruction, the test system first matches the physical fatigue test bench corresponding to the target excavator model, controls the loading device to accurately dock with each force point, and simultaneously starts two core actions: one is to apply load to each force point according to the preset multi-dimensional dynamic load parameter; the other is to control the target excavator to perform simulated operation actions according to the working condition type in the instruction. The coordinated execution of the two ensures that the simulation scenario is close to the real operation, including both external load action and device self-operation action, laying a foundation for subsequent collection of "load-response" related data.

[0031] Optionally, in response to the test instruction issued by the management end for the target hydraulic excavator, the target physical fatigue test bench applies known multi-dimensional dynamic load to each force point of the target hydraulic excavator, and the target hydraulic excavator performs simulated operation according to the working condition type, which can include:

[0032] The test instruction is analyzed to extract the excavator model, multiple force points, multi-dimensional dynamic load corresponding to each force point, multiple working condition types, and time window corresponding to each working condition type;

[0033] Based on the coordinates of each force point and the multi-dimensional dynamic load parameter, the loading device of the target physical fatigue test bench is configured to accurately dock the force application end of the loading device with each force point;

[0034] Based on the extracted working condition type, a job action library matching the working condition type is called from the pre-stored test system, and the action sequence and timing parameters that the actuators of the target hydraulic excavator need to complete are determined based on the job action library;

[0035] Based on the excavator model, a load time sequence curve matching the time window is extracted from the pre-stored historical load spectrum library, and the loading device of the target physical fatigue test bench is triggered to apply multi-dimensional dynamic load to each force point based on the load time sequence curve;

[0036] Based on the time window, the actuators of the target hydraulic excavator are triggered to perform simulated operations according to the action sequence and timing parameters.

[0037] The loading device refers to the execution component on the physical fatigue test bench for applying load to the force point of the excavator. It can output multi-dimensional dynamic load according to pre-set parameters, and its force application end needs to be mechanically connected to the force point to ensure the accuracy of load transmission. The job action library is a pre-stored standardized action data set in the test system corresponding to different working condition types, including the action sequence (such as extension, rotation angle) and timing parameters (such as action start and stop time, speed) of the actuators, which are used to standardize the action logic of the simulation operation. The actuator refers to the core component (such as the boom, stick, bucket, and its driving cylinder) of the hydraulic excavator that completes the job action, and its action directly determines the realization of the working condition. The action sequence is a combination of continuous actions that the actuator needs to complete in a certain working condition, such as "bucket down → into the soil → lift → rotate", which is the specific action flow of the simulation operation. Timing parameters are time parameters used to control the execution rhythm of the action sequence, such as the duration of each action and the interval time between adjacent actions, to ensure the time synchronization of the simulation operation and the load application. The historical load spectrum library is a collection of load curves recorded over time for the same model of excavator in past tests stored in the test system, including load time sequence characteristics in different working conditions, which can be used as a reference template for this load. The load time sequence curve is a curve describing the change of multi-dimensional dynamic load over time in a certain working condition, which is used to guide the force application rhythm of the loading device, making the applied load more close to the real operation characteristics.

[0038] The embodiment is a specific landing process of "responding to test instructions, performing loading and simulation work", and the core logic is "instruction analysis → parameter configuration → collaborative execution", which ensures the accurate matching of load application and simulation work. Specifically, after the test system receives the test instructions of the management end, the key parameters are extracted through the analysis module to determine the target excavator model, stress point position, multi-dimensional dynamic load of each point position, working condition type and corresponding time window, providing data basis for subsequent operations. The loading device configuration is based on the stress point position coordinates and multi-dimensional dynamic load parameters obtained by analysis, and the loading device of the target physical fatigue test bench is controlled to perform mechanical alignment, that is, the spatial posture of the force applying end is adjusted through a servo motor or hydraulic drive to ensure accurate contact with each stress point position, avoid deviation or loss during load transmission, and ensure that the applied dynamic load acts on the structure. The action planning is to call the matching action template from the action action library of the test system according to the analyzed working condition type, generate the action sequence and timing parameters of the execution mechanism, and ensure that the action logic of the simulation work is consistent with the real working condition. In order to make the applied load more consistent with the timing characteristics of the actual work, the system retrieves the load timing curve matched with the current time window from the historical load spectrum library based on the target excavator model, and uses it as the control instruction of the loading device. The loading device applies multi-dimensional dynamic load to each stress point position according to the timing rule of the curve, realizing the precise control of "time-force-direction". The simulation work collaborative execution is to trigger the execution mechanism of the target excavator to perform the work according to the planned action sequence and timing parameters based on the time window, for example, at the 0 moment when the loading device applies the initial load, the execution mechanism starts to act according to the sequence of "entering the soil → excavating → lifting", and the start and stop time of each action is aligned with the key nodes of the load timing curve, ensuring that the external load application and the device itself action are highly coordinated in the time dimension, simulating the coupling effect of load and action in the real work, and providing a scene basis for subsequent collection of high-quality related data.

[0039] S120, record the multi-dimensional dynamic load applied to each stress point position in real time after transmission, and the instantaneous load acting on the bucket. Arrange all the instantaneous loads in the time window in time sequence to form a three-dimensional load spectrum of the bucket.

[0040] The instantaneous load refers to the load acting on the bucket after the force point position is transmitted at a certain moment, including three-dimensional directional component values (X, Y, Z directions) and corresponding action time, and is a basic component unit of the load spectrum. The three-dimensional load spectrum of the bucket refers to a data set formed by arranging all instantaneous loads in time sequence within a time window of a certain working condition, including time dimension and spatial dimension (three-dimensional component), and completely reflects the stress change law of the bucket under the working condition. In the process of applying load on the test bench and the excavator performing simulated operation, through the sensors deployed on the force transmission path of the bucket, the instantaneous load (including X, Y, Z directional components and time stamp) transmitted to the bucket at each moment is captured in real time; all instantaneous loads collected within the same working condition time window are sorted and integrated in time sequence, ensuring that each time node corresponds to a unique three-dimensional component value; the finally formed three-dimensional load spectrum is not only the real record of the stress of the bucket under the working condition, but also the result to be learned and predicted by the subsequent deep learning model.

[0041] Further, the instantaneous load acting on the bucket after the multi-dimensional dynamic load applied to each force point position is transmitted is recorded in real time, and all instantaneous loads within the time window are arranged in time sequence to form a three-dimensional load spectrum of the bucket, which can include:

[0042] While the target physical fatigue test bench applies multi-dimensional dynamic load to each force point position, the process of multi-dimensional dynamic load transmitted from each force point position to the bucket through the structure of the target hydraulic excavator is tracked;

[0043] The load sensor deployed at the force transmission position of the bucket is used to collect the instantaneous load acting on the bucket after transmission in real time; the instantaneous load includes three-dimensional directional component values based on the working posture of the bucket and the corresponding action time, wherein the X direction is the horizontal working direction of the bucket, the Y direction is the horizontal side direction perpendicular to the X direction, and the Z direction is the vertical direction;

[0044] The collected instantaneous loads are time-sequentially aligned in the order of action time based on the time window, ensuring that each time node corresponds to a unique three-dimensional component value;

[0045] The instantaneous load set after time-sequential alignment is integrated in a predetermined format to form a three-dimensional load spectrum of the bucket including time dimension and three-dimensional spatial load dimension.

[0046] The structural transmission refers to the physical process of multi-dimensional dynamic load transmitted from the force point to the mechanical structure of the excavator and then to the bucket, which reflects the conduction path of force in the structure (e.g., from the bucket rod hinge point to the bucket connecting shaft to the bucket body). The force transmission position of the bucket refers to the key position on the bucket that bears and transmits the load, which is the core position where the load finally acts on the bucket. The load sensor is a force measuring device deployed at the force transmission position of the bucket, which can collect the magnitude and direction of the force acting on the bucket in real time and is a direct hardware carrier for obtaining instantaneous load. The bucket operation posture refers to the spatial posture of the bucket during operation (e.g., inclination angle, orientation), and defining three-dimensional directions based on this can ensure that the load direction description is consistent with the actual operation scene. Time sequence alignment refers to sorting the instantaneous loads collected at different times according to the action timestamp, ensuring that the load data is continuous and non-overlapping in the time dimension, forming a complete time sequence. The predetermined format refers to the standardized data format of the three-dimensional load spectrum, which includes the timestamp column and the X / Y / Z directional force column, facilitating data storage, calling, and analysis during model training.

[0047] The embodiment is a specific implementation process for generating a three-dimensional load spectrum of a bucket, and the core logic is "tracking transmission → real-time collection → time sequence integration", which ensures that the load spectrum can truly reflect the stress law of the bucket in simulated operation. Specifically, while the target physical fatigue test bench applies loads to each force point, the system indirectly tracks the transmission path of the load through a structural mechanics model or a sensor network. For example, the transmission process of the load from "boom force point → dipper arm → bucket connecting hinge point → bucket body" is recorded, ensuring that the subsequently collected bucket load is the true stress after complete structural transmission, rather than isolated local load, and avoiding load distortion due to missing transmission path. Using the three-dimensional load sensor deployed at the force transmission position of the bucket, the force value acting on the bucket after structural transmission is captured in real time: the sensor synchronously outputs three-dimensional force values (X-direction horizontal operation force, Y-direction horizontal lateral force, and Z-direction vertical force); and the action timestamp corresponding to each force is recorded, forming an instantaneous load data point of "time-three-dimensional force value".

[0048] S130, synchronously collecting dynamic response data corresponding to the working condition type generated by each force point subjected to multi-dimensional dynamic load during simulated operation, associating the three-dimensional load spectrum with the dynamic response data, and forming a training sample pair containing the mapping relationship between load and response. The training data set is formed by aggregating the training sample pairs of each hydraulic excavator sample in the test system.

[0049] The dynamic response data refers to the physical quantity change data generated by the structure itself when the excavator is subjected to multi-dimensional dynamic load and performs simulated operation, which is a response signal reflecting the load of the device. The training sample pair refers to a pair of associated data composed of three-dimensional load spectrum and dynamic response data in the same working condition and the same time window, wherein the dynamic response data is the input feature, the three-dimensional load spectrum is the output label, and the mapping relationship between the two is reflected. The training data set refers to a large-scale data set formed by aggregating the training sample pairs of multiple hydraulic excavator samples in multiple working conditions in the test system, which is used for training of the deep learning model to improve the generalization ability of the model.

[0050] In the above loading and simulated operation process, a data acquisition device is started synchronously, and through sensors deployed at each force point, dynamic response data generated by the structure under the action of load and operation is collected; the collected dynamic response data is time-aligned according to a time window to ensure that it corresponds to the three-dimensional load spectrum in the time dimension; the aligned dynamic response data and the three-dimensional load spectrum are bound to form a single training sample pair (each sample pair corresponds to complete data of one working condition); the sample pairs of multiple excavator samples (such as different samples of the same model) in multiple working conditions in the test system are aggregated to form a training data set covering multiple scenarios and multiple devices, providing data support for model learning of general rules.

[0051] Optionally, dynamic response data corresponding to the working condition type generated by the execution of the simulated operation and the multi-dimensional dynamic load borne by each force point is synchronously collected, the three-dimensional load spectrum is associated with the dynamic response data to form a training sample pair containing the mapping relationship between the load and the response, comprising:

[0052] When the actuator of the target hydraulic excavator is triggered to start the simulated operation according to the action sequence and the time sequence parameter, strain sensors arranged at each force point collect strain signals generated by each force point under the action of multi-dimensional dynamic load in the simulated operation as strain data in the dynamic response data;

[0053] A pressure sensor arranged at the oil cylinder oil circuit interface collects real-time pressure values of each oil cylinder in the simulated operation as oil cylinder pressure data in the dynamic response data;

[0054] A displacement sensor arranged at the end of the oil cylinder piston rod collects the change amount of the extension length of the piston rod in the simulated operation as oil cylinder displacement data in the dynamic response data;

[0055] The strain data, oil cylinder pressure data, and oil cylinder displacement data are marked with time stamps according to the time window to form a dynamic response data set with time sequence marks;

[0056] The dynamic response data set is matched and associated with the bucket three-dimensional load spectrum in the same time window according to the time stamp, to form a training sample pair containing the mapping relationship between the load and the response.

[0057] The strain sensor is a sensing device (such as a strain gauge) arranged on the surface of each force point structure of the excavator, which can convert the small deformation of the structure caused by the load into an electrical signal, and is used to collect strain data in the dynamic response and reflect the stress deformation state of the structure. The pressure sensor refers to a measuring device installed at the oil cylinder oil circuit interface, which is used to monitor the pressure change of hydraulic oil on the oil cylinder inlet and return oil circuit in real time, and the output pressure value is directly related to the stress state of the oil cylinder, which is an important part of the dynamic response data. The displacement sensor refers to a linear measuring device arranged at the end of the oil cylinder piston rod, which can accurately collect the length change of the piston rod during the extension and retraction process, reflect the action amplitude and speed of the actuator, and provide data support for the action dimension of the dynamic response. The dynamic response data set with time sequence marking refers to the data set formed by adding time stamps to the strain data, oil cylinder pressure data and oil cylinder displacement data according to the collection time, which ensures that each type of response data corresponds in the time dimension, and constitutes a complete "time-response" signal set. Time stamp matching and association refers to binding the dynamic response data set with the load data at the same time in the three-dimensional load spectrum through the same time stamp, establishing the mapping relationship of "response signal at a certain time → load state at the corresponding time", which is the core logic of forming the training sample pair.

[0058] The core landing process of generating the training sample pair is to "synchronously collect multiple dimensions → time sequence marking → load-response association", which ensures that the response data at each time can correspond to a unique load state, provides high-quality samples for model learning mapping relationship, and specifically: activate the strain sensors arranged at each force point while the excavator actuator starts the simulation operation. These sensors are in close contact with the structure surface, and when the force point bears multi-dimensional dynamic load and deforms with the operation action, the sensors convert the deformation into real-time strain signals and record the collection time synchronously, forming a continuous strain data sequence. Synchronously start the pressure sensors installed at the oil cylinder oil circuit interface to monitor the hydraulic oil pressure change of the oil cylinder during the extension and retraction action in real time. The sensor converts these pressure changes into electrical signals and marks the time stamp to form the oil cylinder pressure data. Through the displacement sensors arranged at the end of each oil cylinder piston rod, the length change of the piston rod during the simulation operation is captured in real time. The displacement sensor converts the length change into a digital signal and adds a time stamp to form the oil cylinder displacement data.

[0059] S140, feature processing is performed on the training data set, and the training data set after the feature processing is input to the original deep learning model for iterative training, and the deep learning model meeting the iterative termination condition is taken as a hydraulic excavator bucket load spectrum generation model.

[0060] The characteristic processing refers to pre-processing the dynamic response data and the three-dimensional load spectrum in the training data set, converting the original data into a format suitable for the input of the deep learning model, and improving the training efficiency and accuracy of the model. The original deep learning model refers to an initial neural network model that has not been trained and has the ability to learn nonlinear mapping relationships, which is the algorithm carrier of the embodiment. The iterative training refers to inputting the characteristic training data set into the original model, and gradually reducing the deviation between the predicted load spectrum output by the model and the real load spectrum by continuously adjusting the model parameters. The iteration termination condition refers to the judgment standard for stopping the model training, which ensures that the model has sufficient accuracy and avoids overfitting. The hydraulic excavator bucket load spectrum generation model refers to the deep learning model after training, which can receive new dynamic response data as input and directly output the corresponding bucket three-dimensional load spectrum, realizing the rapid generation of the load spectrum.

[0061] The training data set is characterized, for example, the strain, pressure, displacement signals in the dynamic response data are standardized, and the spatial distribution features and time sequence features are extracted. The processed data set is divided into a training set and a validation set in proportion, and input into the original deep learning model. The model calculates the predicted load spectrum through forward propagation, compares it with the real load spectrum to calculate the loss value, and then adjusts the network parameters through back propagation for repeated iteration and optimization. When the model loss value meets the preset termination condition, the training is stopped, and the model at this time has the ability to accurately predict the bucket three-dimensional load spectrum from the dynamic response data, and can be used as the final load spectrum generation model for application.

[0062] In the embodiment of the application, the "dynamic response data-three-dimensional load spectrum" associated samples are collected by the physical fatigue test bench and the excavator sample in cooperation, and the model is trained and generated in combination with deep learning, which not only retains the accuracy of physical testing, but also avoids the high cost and long cycle problem of repeated testing of each device in traditional bench testing, and overcomes the deviation caused by idealized assumptions in theoretical calculation, realizes efficient construction of high-precision bucket load spectrum generation model with low cost, and provides reliable support for batch excavator load spectrum acquisition and structure life simulation.

[0063] Embodiment two

[0064] Figure 2 The flowchart of another hydraulic excavator bucket load spectrum generation model construction method provided by the embodiment two of the application is based on the embodiment one and is refined, specifically, as shown in the embodiment two, the method comprises the steps of: Figure 2

[0065] ​S210, in response to the test instruction for the target hydraulic excavator issued by the management end, control the target physical fatigue test bench to apply a known multi-dimensional dynamic load to each force point of the target hydraulic excavator, and control the target hydraulic excavator to perform simulated operation according to the working condition type; wherein the test instruction includes: excavator model, multiple force points, multi-dimensional dynamic load corresponding to each force point, multiple working condition types and time window corresponding to each working condition type.

[0066] S220, record the instantaneous load acting on the bucket after the multi-dimensional dynamic load applied to each force point is transmitted in real time, arrange all the instantaneous loads in the time window in time sequence to form a three-dimensional load spectrum of the bucket.

[0067] S230, synchronously collect dynamic response data corresponding to the working condition type generated by each force point from performing simulated operation and bearing multi-dimensional dynamic load, associate the three-dimensional load spectrum with the dynamic response data to form a training sample pair containing the mapping relationship between load and response, and form a training data set by aggregating the training sample pairs of each hydraulic excavator sample in the test system.

[0068] S240, extract the three-dimensional load spectrum and the dynamic response data from each training sample pair in the training data set, separate the spatial features and the time sequence features after standardization processing; wherein the spatial features include the directional distribution of the three-dimensional load spectrum and the spatial distribution features of the dynamic response data, and the time sequence features include the dynamic change features of the load and the response with the time window.

[0069] The spatial features refer to the distribution characteristics of data in the spatial dimension, including the directional distribution features of the three-dimensional load spectrum and the spatial distribution features of the dynamic response data. The time sequence features refer to the change law of data in the time dimension, including the dynamic change features of the load and the response with the time window. The embodiment is the core algorithm process of model training, and the core logic is "feature separation → separate extraction → fusion learning → iterative optimization". The data features are fully mined through the hybrid network, specifically: from each training sample pair of the training data set, the dynamic response data and the three-dimensional load spectrum are extracted, and the two types of data are standardized. The spatial features are the numerical distribution of different sensor points extracted from the dynamic response data and the proportional relationship of X / Y / Z directional forces extracted from the three-dimensional load spectrum. The time sequence features are the change curve with time extracted from the dynamic response data and the load peak time sequence in the time window extracted from the three-dimensional load spectrum.

[0070] Optionally, the original deep learning model is a hybrid deep learning model containing a convolutional neural network and a recurrent neural network.

[0071] The hybrid deep learning model refers to a model combining the advantages of two or more neural network structures. In this embodiment, it specifically refers to a combined model of "convolutional neural network (CNN) + recurrent neural network (RNN)", where CNN is good at extracting spatial features, and RNN is good at capturing time dependence. The combination of the two can simultaneously process the spatial distribution and temporal variation characteristics in the data. By explicitly using the "convolutional neural network + recurrent neural network" hybrid structure in the original deep learning model, the core purpose is to match the "spatial-temporal" dual characteristics of the training data: dynamic response data (strain, pressure, displacement) is distributed in different sensor points (spatial dimension) and continuously changes over time (temporal dimension); the three-dimensional load spectrum also contains spatial direction distribution (X / Y / Z direction) and time sequence characteristics. The hybrid deep learning model can extract these two types of features through CNN and RNN respectively, and can more accurately learn the mapping relationship between "response-load" than a single network structure.

[0072] S250, input the extracted spatial features into the convolutional neural network part of the original deep learning model, and output a spatial feature vector representing the spatial correlation relationship through convolution operation and feature mapping.

[0073] Convolution operation and feature mapping are the core operations of convolutional neural network. The separated spatial features are input into the CNN part (such as a network containing 2 convolution layers + 1 pooling layer); the convolution layer performs sliding calculation on the spatial features through a 3x3 convolution kernel (such as calculating the strain correlation of adjacent 3 force points); the pooling layer preserves key features through downsampling; finally, a spatial feature vector is output, which is a one-dimensional vector output after processing by the convolutional neural network, and condenses all spatial correlation information in the data (such as the correlation strength of each point response and load direction), which is a digital expression of spatial features. For example, it contains abstract features such as "the correlation degree of strain at each point and X-direction load" and "the spatial coupling coefficient of cylinder pressure and Z-direction load".

[0074] S260, input the extracted time sequence features into the recurrent neural network part of the original deep learning model, and output a time sequence feature vector representing the dynamic change law through time sequence dependence capture and sequence modeling.

[0075] Time sequence dependence capture and sequence modeling are the core functions of recurrent neural network. The separated time sequence features are input into the recurrent neural network part to capture long-term time sequence dependence (such as the influence of displacement change at 5 seconds on load at 10 seconds); finally, a time sequence feature vector is output, which is a one-dimensional vector output after processing by the recurrent neural network, and condenses all time sequence change laws in the data, which is a digital expression of time sequence features, such as containing "the time difference between load peak value and pressure peak value" and "the periodic frequency of strain signal".

[0076] S270, fuse the spatial feature vector and the time sequence feature vector to generate a joint feature vector containing the load-response mapping relationship, and output a prediction result based on the joint feature vector.

[0077] The joint feature vector is a comprehensive feature vector formed by fusing the spatial feature vector and the time sequence feature vector, and simultaneously contains spatial correlation and time sequence dependence information, which fully describes the dual characteristics of the "response-load" mapping. The prediction result refers to the predicted value of the three-dimensional load spectrum of the bucket output by the deep learning model based on the joint feature vector. By comparing the real load spectrum in the training sample, the error can be calculated. After the joint feature vector is mapped through the full connection layer, the predicted three-dimensional load spectrum of the bucket (including the X / Y / Z direction force prediction value at each time in the time window) is output.

[0078] S280, adjust the layer parameters of the convolutional neural network and the recurrent neural network according to the prediction result, and iterate the training again using the training data set until the loss value reaches the preset convergence condition or the iteration number meets the set threshold, stop the training, and the deep learning model that meets the iteration termination condition is used as the hydraulic excavator bucket load spectrum generation model.

[0079] The layer parameters refer to the learnable parameters such as weights and biases of each layer of the recurrent neural network. Adjusting the parameters can optimize the prediction accuracy of the model. The loss value is an index for measuring the difference between the prediction result and the true value. The smaller the loss value, the more accurate the model prediction. The preset convergence condition refers to the judgment standard for stopping the model training, which ensures that the model terminates training after reaching the preset accuracy. The loss value between the prediction result and the true three-dimensional load spectrum in the training sample is calculated, and the loss value is fed back to each layer of the network through the back propagation algorithm to adjust the convolution kernel weight of the convolutional neural network, the recurrent unit parameter of the recurrent neural network, and the full connection layer weight; repeat the above "input data→extract features→prediction→calculate loss→adjust parameters" process, and iterate multiple times using the training data set; when the loss value decreases to the preset convergence condition or the iteration number reaches the set threshold, stop the training. At this time, the model has fully learned the mapping rule of "dynamic response→bucket load spectrum", and can be used as the final load spectrum generation model.

[0080] In the embodiment of the present application, the mixed deep learning model takes into account the spatial correlation and time sequence rule of the test data, and through targeted training of the deep learning real load transmission logic, it not only continues the cost reduction advantage of a small amount of bench test, but also improves the accuracy of batch generation of bucket load spectrum, providing high-precision and economical load input for excavator structure life simulation.

[0081] Embodiment three

[0082] Figure 3 A structure schematic diagram of a bucket load spectrum generation model construction device provided in embodiment three of the present application. As shown inFigure 3 The device comprises:

[0083] The load application and simulation operation module 310 is configured to control the target physical fatigue test bench to apply known multi-dimensional dynamic loads to each force point of the target hydraulic excavator in response to a test instruction issued by the management terminal, and control the target hydraulic excavator to perform simulation operation according to the working condition type; wherein the test instruction comprises: excavator model, multiple force points, multi-dimensional dynamic loads corresponding to each force point, multiple working condition types, and time windows corresponding to each working condition type.

[0084] The test load spectrum generation module 320 is configured to record the instantaneous load acting on the bucket after the multi-dimensional dynamic load applied to each force point is transmitted, arrange all the instantaneous loads in the time window in time sequence, and form a three-dimensional load spectrum of the bucket.

[0085] The load and response correlation module 330 is configured to synchronously collect dynamic response data corresponding to the working condition type generated by each force point due to the execution of the simulation operation and the multi-dimensional dynamic load borne, correlate the three-dimensional load spectrum with the dynamic response data, form a training sample pair containing the mapping relationship between the load and the response, and form a training data set by aggregating the training sample pairs of each hydraulic excavator sample in the test system.

[0086] The load spectrum generation model construction module 340 is configured to perform feature processing on the training data set, input the feature-processed training data set into the original deep learning model for iterative training, and take the deep learning model meeting the iterative termination condition as the hydraulic excavator bucket load spectrum generation model.

[0087] In the embodiments of the present application, the dynamic response data-three-dimensional load spectrum correlation sample is collected by the physical fatigue test bench and the excavator sample in cooperation, and the deep learning training generation model is combined, which not only retains the accuracy of physical testing, but also avoids the high cost and long period problem of repeated testing of each device in traditional bench testing, and overcomes the deviation caused by idealized assumptions in theoretical calculation, so as to efficiently construct a high-precision bucket load spectrum generation model at a low cost, and provide reliable support for batch excavator load spectrum acquisition and structure life simulation.

[0088] Optionally, on the basis of each of the above embodiments, the load application and simulation operation module 310 can comprise:

[0089] The test instruction analysis unit is configured to analyze the test instruction and extract the excavator model, multiple force points, multi-dimensional dynamic loads corresponding to each force point, multiple working condition types, and time windows corresponding to each working condition type.

[0090] A test bench configuration unit is configured to configure a loading device of a target physical fatigue test bench based on each force point coordinate and multi-dimensional dynamic load parameter, so that the force applying end of the loading device is accurately connected with each force point;

[0091] A working condition and action matching unit is configured to call a working action library matched with the working condition type in the test system based on the extracted working condition type, and determine a sequence of actions and timing parameters of an execution mechanism of the target hydraulic excavator based on the working action library;

[0092] A load applying unit is configured to extract a pre-stored load timing curve matched with the time window in the historical load spectrum library based on the excavator model, and trigger the loading device of the target physical fatigue test bench to apply multi-dimensional dynamic load to each force point based on the load timing curve;

[0093] A simulated working execution unit is configured to trigger the execution mechanism of the target hydraulic excavator to execute the simulated working based on the sequence of actions and timing parameters with the time window as a reference.

[0094] Optionally, on the basis of each of the above embodiments, the test load spectrum generation module 320 can include:

[0095] A load transmission state tracking unit is configured to track the process of multi-dimensional dynamic load transmitted from each force point to the bucket through the structure of the target hydraulic excavator while the target physical fatigue test bench applies multi-dimensional dynamic load to each force point;

[0096] An instantaneous load collecting unit is configured to collect instantaneous load acting on the bucket after transmission in real time by using a load sensor arranged at the force transmission position of the bucket; the instantaneous load includes three-dimensional directional component values and corresponding action time based on the bucket working posture, wherein the X direction is the horizontal working direction of the bucket, the Y direction is the horizontal lateral direction perpendicular to the X direction, and the Z direction is the vertical direction;

[0097] A load timing alignment unit is configured to align all the collected instantaneous loads in the order of action time based on the time window, so as to ensure that each time node corresponds to a unique three-dimensional component value;

[0098] A bucket three-dimensional load spectrum generation unit is configured to integrate the set of instantaneous loads after timing alignment in a predetermined format, to form a bucket three-dimensional load spectrum including time dimension and three-dimensional space load dimension.

[0099] Optionally, on the basis of each of the above embodiments, the load and response correlation module 330 can include:

[0100] The strain data acquisition unit is configured to collect strain signals of each force point caused by multi-dimensional dynamic loads during the simulation operation through strain sensors arranged at each force point, as strain data in the dynamic response data.

[0101] The pressure data acquisition unit is configured to collect real-time pressure values of each oil cylinder during the simulation operation through pressure sensors arranged at the oil cylinder oil circuit interface, as oil cylinder pressure data in the dynamic response data.

[0102] The displacement data acquisition unit is configured to collect the change amount of the extension length of the piston rod during the simulation operation through displacement sensors arranged at the end of the oil cylinder piston rod, as oil cylinder displacement data in the dynamic response data.

[0103] The dynamic response data set generation unit is configured to mark time stamps for the strain data, the oil cylinder pressure data, and the oil cylinder displacement data according to the time window, to form a dynamic response data set with time sequence marks.

[0104] The training sample pair construction unit is configured to match and associate the dynamic response data set with the bucket three-dimensional load spectrum in the same time window according to the time stamps, to form a training sample pair containing the mapping relationship between the load and the response.

[0105] Optionally, on the basis of each of the above embodiments, the original deep learning model is a hybrid deep learning model containing a convolutional neural network and a recurrent neural network.

[0106] Optionally, on the basis of each of the above embodiments, the load spectrum generation model construction module 340 can include:

[0107] The feature extraction unit is configured to extract the three-dimensional load spectrum and the dynamic response data from each training sample pair in the training data set, and separate spatial features and time sequence features after standardization processing; wherein the spatial features include direction distribution of the three-dimensional load spectrum and spatial distribution features of the dynamic response data, and the time sequence features include dynamic change features of the load and the response with the time window.

[0108] The spatial feature vector generation unit is configured to input the extracted spatial features into the convolutional neural network part of the original deep learning model, and output spatial feature vectors representing spatial correlation relationships through convolution operation and feature mapping.

[0109] The time sequence feature vector generation unit is configured to input the extracted time sequence features into the recurrent neural network part of the original deep learning model, and output time sequence feature vectors representing dynamic change laws through time sequence dependence capture and sequence modeling.

[0110] The vector fusion and result prediction unit is configured to fuse the spatial feature vector and the time sequence feature vector, generate a joint feature vector containing a load and response mapping relationship, and output a prediction result based on the joint feature vector.

[0111] The model optimization unit is configured to adjust layer parameters of the convolutional neural network and the recurrent neural network according to the prediction result, and iteratively train the model again using the training data set until a loss value reaches a preset convergence condition or an iteration number meets a set threshold, and stop training.

[0112] Optionally, on the basis of each of the above embodiments, a load spectrum generation model application unit can be further included, configured to, after using the deep learning model meeting the iteration termination condition as a hydraulic excavator bucket load spectrum generation model, in response to a load spectrum generation request for a to-be-processed hydraulic excavator that is performing a real job, the request including a real-time working condition type of the to-be-processed excavator and a corresponding time window, apply the load spectrum generation model to generate a load spectrum for the to-be-processed hydraulic excavator.

[0113] The strain sensors deployed in advance at each force point of the to-be-processed hydraulic excavator collect actual strain data generated by each force point during the real job;

[0114] The pressure sensor deployed at the oil cylinder oil circuit interface collects actual pressure data of the oil cylinder during the real job;

[0115] The displacement sensor deployed at the end of the oil cylinder piston rod collects actual extension and retraction displacement data of the piston rod during the real job;

[0116] The actual strain data, actual pressure data and actual displacement data are summarized as actual dynamic response data corresponding to the real-time working condition type, time-sequentially aligned according to the time window in the load spectrum generation request, and processed by using the same standardization processing method as the training data set to obtain input data adapted to the virtual test bench model;

[0117] The input data is input into the virtual test bench model to generate an actual bucket three-dimensional load spectrum corresponding to the real job of the to-be-processed hydraulic excavator, and the actual three-dimensional load spectrum is pushed to the management end for load characteristic analysis.

[0118] The construction device of the bucket load spectrum generation model provided by the embodiment of the application can execute the construction method of the bucket load spectrum generation model provided by any embodiment of the application, and has the corresponding function modules and beneficial effects of the execution method.

[0119] Embodiment Four

[0120] Figure 4A structural diagram of an electronic device 10 that can be used to implement embodiments of the present application is shown. The electronic device is intended to represent various forms of digital computers, such as laptops, desktops, tablets, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device can also represent various forms of mobile devices such as personal digital assistants, cellular telephones, smartphones, wearable devices (e.g., headsets, glasses, watches, etc.), and other similar computing devices. The components shown here, their connections and relationships, and their functions, are meant to be examples only, and are not meant to limit implementations of the present application described and / or claimed in this document.

[0121] As shown in Figure 4 The electronic device 10 includes at least one processor 11, and a memory, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., communicatively connected to the at least one processor 11, where the memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes according to the computer programs stored in the read-only memory (ROM) 12 or loaded into the random access memory (RAM) 13 from the storage unit 18. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0122] Various components in the electronic device 10 are connected to the I / O interface 15, including an input unit 16, such as a keyboard, a mouse, etc., an output unit 17, such as various types of displays, speakers, etc., a storage unit 18, such as a magnetic disk, an optical disk, etc., and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network, such as the Internet, and / or various telecommunication networks.

[0123] The processor 11 can be various general and / or special purpose processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 performs various methods and processes described above, such as a method of constructing a shovel load spectrum generation model.

[0124] That is, in response to a test instruction for a target hydraulic excavator issued by the management end, the target physical fatigue test bench applies known multi-dimensional dynamic loads to each force point of the target hydraulic excavator, while the target hydraulic excavator performs simulated operations according to the working condition type; wherein the test instruction includes: excavator model, multiple force points, multi-dimensional dynamic loads corresponding to each force point, multiple working condition types, and time windows corresponding to each working condition type;

[0125] Real-time record the multi-dimensional dynamic loads applied to each force point after being transmitted, acting on the instantaneous load of the bucket, arrange all the instantaneous loads in the time window in chronological order to form a three-dimensional load spectrum of the bucket;

[0126] Synchronously collect the dynamic response data corresponding to the working condition type generated by each force point performing simulated operations and bearing multi-dimensional dynamic loads, associate the three-dimensional load spectrum with the dynamic response data to form a training sample pair containing the mapping relationship between load and response, and aggregate the training sample pairs of each hydraulic excavator sample in the test system to form a training data set;

[0127] Feature processing is performed on the training data set, and the training data set after feature processing is input into the original deep learning model for iterative training, and the deep learning model meeting the iterative termination condition is taken as the hydraulic excavator bucket load spectrum generation model.

[0128] In some embodiments, a method for constructing a bucket load spectrum generation model can be implemented as a computer program tangibly embodied in a computer readable storage medium, such as the storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the method for constructing a bucket load spectrum generation model described above can be performed. Alternatively, in other embodiments, the processor 11 can be configured to perform the method for constructing a bucket load spectrum generation model by any other appropriate means, for example, by means of firmware.

[0129] The various embodiments of the systems and techniques described above can be implemented in digital electronic circuitry, integrated circuitry, a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a system on a chip (SOC), a load programmable logic device (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include implementation in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which can be special or general purpose, coupled to receive data and instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device.

[0130] Computer programs used to implement the processes of the application can be written in any combination of one or more programming languages. These computer programs can be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus, such that the computer program

[0131] In the context of the present application, a computer-readable storage medium can be a tangible medium that can contain or store computer programs for use by or in connection with an instruction execution system, apparatus, or device. Computer-readable storage media can include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium can be a machine-readable signal medium. More specific examples of the machine-readable storage medium will include one or more lines of a program of instructions in a transitory signal, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0132] To provide for interaction with a user, the systems and techniques described here can be implemented on an electronic device having a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the electronic device. Other kinds of devices can be used to provide for interaction with a user as well; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form, including acoustic, speech, or tactile input.

[0133] The systems and techniques described here can be implemented in a computing system that includes a back end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front end component (e.g., a user computer having a graphical user interface or a Web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back end, middleware, or front end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0134] The computing system can include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. A server can be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system, to solve the defects of large management difficulty and weak business scalability in traditional physical host and VPS service.

[0135] It should be understood that various forms of flow shown above can be re-ordered, added to, or deleted from without departing from the spirit and scope of the present disclosure. For example, the steps recited in the present disclosure can be executed in parallel, executed in sequence, or executed in a different order, as long as the desired results of the present disclosure are achieved, and the present disclosure is not limited herein.

[0136] The specific embodiments described above are not intended to be limiting, and persons skilled in the art will appreciate that various modifications, combinations, sub-combinations and alternatives can be made to the specific embodiments without departing from the spirit and scope of the disclosure. Any alternatives, modifications, equivalents, and the like of all of the above described devices, methods, and other functional aspects of the present disclosure are intended to be encompassed by the following claims.

Claims

1. A method of constructing a bucket load spectrum generation model, executed by a test system having at least one physical fatigue test rig and at least one hydraulic excavator specimen deployed, wherein, The physical fatigue test bench is uniquely corresponding to the model of the hydraulic excavator, and the physical fatigue test bench comprises: In response to the test instruction issued by the management end for the target hydraulic excavator, the target physical fatigue test bench is controlled to apply known multidimensional dynamic load to each stress point of the target hydraulic excavator, and the target hydraulic excavator is controlled to perform simulated operation according to the working condition type; wherein the test instruction comprises: excavator model, multiple stress points, multidimensional dynamic load corresponding to each stress point, multiple working condition types and time window corresponding to each working condition type; Real-time record the instantaneous load acting on the bucket after the multidimensional dynamic load applied to each stress point is transmitted, arrange all the instantaneous loads in the time window in time sequence to form a three-dimensional load spectrum of the bucket; Synchronously collect the dynamic response data corresponding to the working condition type generated by each stress point from the simulated operation and the multidimensional dynamic load borne, associate the three-dimensional load spectrum with the dynamic response data to form a training sample pair containing the mapping relationship between load and response, and form a training data set by aggregating the training sample pairs of each hydraulic excavator sample in the test system; The training data set is processed by feature extraction, and the training data set processed by feature extraction is input into the original deep learning model for iterative training, and the deep learning model meeting the iteration termination condition is used as the hydraulic excavator bucket load spectrum generation model.

2. The method of claim 1, wherein, In response to the test instruction issued by the management end for the target hydraulic excavator, the target physical fatigue test bench is controlled to apply known multidimensional dynamic load to each stress point of the target hydraulic excavator, and the target hydraulic excavator is controlled to perform simulated operation according to the working condition type, comprising: Analyzing the test instruction to extract the excavator model, multiple stress points, multidimensional dynamic load corresponding to each stress point, multiple working condition types and time window corresponding to each working condition type; Based on the coordinates of each stress point and the multidimensional dynamic load parameters, the loading device of the target physical fatigue test bench is configured, so that the force applying end of the loading device is accurately connected with each stress point; Based on the extracted working condition type, the working action library matching the working condition type is called from the pre-stored test system, and the action sequence and timing parameters required to be completed by the execution mechanism of the target hydraulic excavator are determined based on the working action library; Based on the excavator model, the load timing curve matching the time window is extracted from the historical load spectrum library, and the loading device of the target physical fatigue test bench is triggered to apply multidimensional dynamic load to each stress point based on the load timing curve; Taking the time window as the reference, the execution mechanism of the target hydraulic excavator is triggered to perform simulated operation according to the action sequence and timing parameters.

3. The method of claim 1, wherein, Real-time record the instantaneous load acting on the bucket after the multidimensional dynamic load applied to each stress point is transmitted, arrange all the instantaneous loads in the time window in time sequence to form a three-dimensional load spectrum of the bucket, comprising: The target physical fatigue test bench applies multi-dimensional dynamic loads to each force point, and tracks the process of the multi-dimensional dynamic loads being transmitted from each force point to the bucket through the structure of the target hydraulic excavator; The load sensor deployed at the force transmission position of the bucket is used to collect the instantaneous load acting on the bucket in real time after transmission; the instantaneous load includes three-dimensional directional component values and corresponding action time based on the working posture of the bucket, wherein the X direction is the horizontal working direction of the bucket, the Y direction is the horizontal side direction perpendicular to the X direction, and the Z direction is the vertical direction; The collected instantaneous loads are time-aligned in the order of action time based on the time window, ensuring that each time node corresponds to a unique three-dimensional component value; The instantaneous load set after time alignment is integrated in a predetermined format to form a three-dimensional load spectrum of the bucket including the time dimension and the three-dimensional load dimension.

4. The method of claim 2, wherein, Synchronously collect dynamic response data corresponding to the working condition type generated by the multi-dimensional dynamic load applied to each force point during the simulation operation, associate the three-dimensional load spectrum with the dynamic response data, and form a training sample pair including the mapping relationship between the load and the response, including: When the actuators of the target hydraulic excavator are triggered to start the simulation operation according to the action sequence and time sequence parameters, the strain sensors arranged at each force point are used to collect the strain signals generated by each force point due to the multi-dimensional dynamic load applied during the simulation operation, as strain data in the dynamic response data; The pressure sensors arranged at the oil cylinder oil path interface are used to collect the real-time pressure values of each oil cylinder during the simulation operation, as cylinder pressure data in the dynamic response data; The displacement sensors arranged at the end of the oil cylinder piston rod are used to collect the extension length change of the piston rod during the simulation operation, as cylinder displacement data in the dynamic response data; The strain data, cylinder pressure data, and cylinder displacement data are marked with time stamps according to the time window to form a dynamic response data set with time sequence markers; The dynamic response data set and the three-dimensional load spectrum of the bucket in the same time window are matched and associated according to the time stamp to form a training sample pair including the mapping relationship between the load and the response.

5. The method of claim 1, wherein, The original deep learning model is a hybrid deep learning model including a convolutional neural network and a recurrent neural network.

6. The method of claim 5, wherein, The training data set is feature processed, and the feature processed training data set is input into the original deep learning model for iterative training, including: The three-dimensional load spectrum and the dynamic response data are extracted from each training sample pair in the training data set, and the spatial features and the time sequence features are separated after standardization; wherein the spatial features include the directional distribution of the three-dimensional load spectrum and the spatial distribution features of the dynamic response data, and the time sequence features include the dynamic change features of the load and the response with the time window; The extracted spatial features are input into the convolutional neural network part of the original deep learning model, and the spatial feature vectors representing the spatial correlation are output through convolution operation and feature mapping. The extracted time sequence features are input into the recurrent neural network part of the original deep learning model to capture time sequence dependence and sequence modeling, and output time sequence feature vectors representing dynamic change rules; The spatial feature vectors and the time sequence feature vectors are fused to generate joint feature vectors containing load and response mapping relationships, and a prediction result is output based on the joint feature vectors; The layer parameters of the convolutional neural network and the recurrent neural network are adjusted according to the prediction result, and the training data set is used for iterative training again until the loss value reaches a preset convergence condition or the number of iterations meets a set threshold, and the training is stopped.

7. The method according to any one of claims 1 to 6, characterized in that, After the deep learning model meeting the iteration termination condition is used as the hydraulic excavator bucket load spectrum generation model, further comprising: In response to a load spectrum generation request issued by the management end for a hydraulic excavator that is performing a real job, the request including the real-time working condition type of the to-be-processed excavator and the corresponding time window; Actual strain data generated by each force point during the real job process is collected through strain sensors pre-deployed at each force point of the to-be-processed hydraulic excavator; Actual pressure data of the oil cylinder during the real job is collected through pressure sensors deployed at the oil cylinder oil circuit interface; Actual telescopic displacement data of the piston rod during the real job is collected through displacement sensors deployed at the end of the oil cylinder piston rod; The actual strain data, actual pressure data, and actual displacement data are summarized as actual dynamic response data corresponding to the real-time working condition type, time-aligned according to the time window in the load spectrum generation request, and processed using the same standardization method as the training data set to obtain input data suitable for the virtual test bench model; The input data is input into the virtual test bench model to generate an actual three-dimensional bucket load spectrum corresponding to the real job of the to-be-processed hydraulic excavator, and the actual three-dimensional load spectrum is pushed to the management end for load characteristic analysis.

8. A shovel load spectrum generation model construction device characterized by comprising: Arranged in a test system having at least one physical fatigue test bench and at least one hydraulic excavator sample, the device comprises: A load application and simulation operation module for controlling the target physical fatigue test bench to apply known multi-dimensional dynamic loads to each force point of the target hydraulic excavator in response to a test instruction issued by the management end for the target hydraulic excavator, while controlling the target hydraulic excavator to perform simulation operation according to the working condition type; wherein the test instruction includes: excavator model, multiple force points, multi-dimensional dynamic loads corresponding to each force point, multiple working condition types, and time windows corresponding to each working condition type; A test load spectrum generation module for recording the instantaneous loads acting on the bucket after the multi-dimensional dynamic loads applied to each force point are transmitted, arranging all the instantaneous loads in the time window in chronological order to form a three-dimensional load spectrum of the bucket. The load and response association module is configured to synchronously collect dynamic response data corresponding to a working condition type generated by each force point position and a multi-dimensional dynamic load borne by the execution simulation operation, associate the three-dimensional load spectrum with the dynamic response data, and form a training sample pair containing a mapping relationship between the load and the response, and form a training data set by aggregating training sample pairs of each hydraulic excavator sample in the test system; The load spectrum generation model construction module is configured to perform feature processing on the training data set, input the feature-processed training data set to an original deep learning model for iterative training, and take a deep learning model meeting an iterative termination condition as a hydraulic excavator bucket load spectrum generation model.

9. An electronic device, comprising: The electronic device includes: at least one processor; and a memory connected to the at least one processor in communication; wherein The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to execute the construction method of the bucket load spectrum generation model according to any one of claims 1-7.

10. A computer program product, characterised in that, The computer program product includes a computer program that, when executed by a processor, implements the construction method of the bucket load spectrum generation model according to any one of claims 1-7. The computer program product includes a computer program that, when executed by a processor, implements the construction method of the bucket load spectrum generation model according to any one of claims 1-7.

Images