Intelligent control method for heading machine based on hybrid heterogeneous computing

Abstract

The present application relates to a kind of based on mixed heterogeneous computing's intelligent control method of heading machine, comprising: edge controller obtains heading machine task information, and according to heading machine task information and the resource scheduling strategy set in advance, obtain task scheduling information;Edge controller receives the instruction of starting execution task sent by host computer, obtains task scheduling sequence according to the instruction and task scheduling information;According to the task scheduling sequence, edge controller sends the instruction of acquisition data information to data acquisition equipment, and receives the data information fed back by data acquisition equipment;According to the data information fed back by data acquisition equipment, edge controller carries out data processing using the model optimized in advance, obtains model calculation result, and sends model calculation result to host computer platform.The present application can realize the collaborative control of heading machine logic control, motion control and other real-time tasks and intelligent application non-real-time tasks, improve the execution efficiency of task and the efficiency of data interaction.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of tunnel construction, and particularly relates to an intelligent control method for a tunneling machine based on hybrid heterogeneous computing. BACKGROUND

[0002] With the rise of artificial intelligence, big data and other technologies and the promotion of engineering digital transformation, intelligent technology has gradually penetrated into the tunneling construction industry. However, due to the large scale, numerous units and complex structure of shield equipment, and the complex geological environment during construction, intelligent tunneling still needs to be continuously developed and improved to ensure the efficiency and safety of tunneling construction.

[0003] With the increasing demand for intelligent tunneling, more and more intelligent systems have been deployed on tunneling equipment, such as belt conveyor slag volume flow monitoring, rock slag analysis, automatic measurement of shield tail gap, automatic spraying and mixing, etc. Generally, the intelligent control calculation task is directly executed on the industrial controller. However, due to the limited resources of the industrial controller, this method cannot guarantee the execution efficiency of the algorithm. In order to better complete the intelligent task and achieve the expected effect, another method is to additionally configure different hardware support based on the industrial control, which will lead to the need to carry multiple sets of controller hardware of different manufacturers and models on the tunneling equipment, resulting in waste of controller resources and complex data interaction communication of multiple hardware systems, making it difficult to realize resource sharing and coordinated control of multiple tasks.

[0004] Therefore, there is an urgent need for an intelligent control method for a tunneling machine based on hybrid heterogeneous computing. SUMMARY

[0005] (I) Technical problems to be solved

[0006] In view of the above-mentioned shortcomings and deficiencies of the prior art, the present application provides an intelligent control method for a tunneling machine based on hybrid heterogeneous computing, which solves the technical problems of low execution efficiency, waste of controller resources, complex data interaction communication of multiple hardware systems, and difficulty in realizing resource sharing and coordinated control of multiple tasks in the prior art to achieve the intelligent requirements of the tunneling machine.

[0007] (II) Technical solutions

[0008] In order to achieve the above-mentioned purposes, the main technical solutions adopted by the present application include:

[0009] The embodiment of the application provides a kind of based on hybrid heterogeneous computing's intelligent control method of heading machine, the intelligent control method is applied to heading machine intelligent control system, the system includes: for sending the instruction of execution task to edge controller and showing model calculation result upper computer platform, for optimizing algorithm model cloud, for collecting data information data acquisition equipment and edge controller, the intelligent control method of heading machine includes:

[0010] S1, edge controller obtains heading machine task information, and according to the heading machine task information and the resource scheduling strategy set in advance, obtains task scheduling information;

[0011] S2, edge controller receives the instruction of starting execution task sent by upper computer, obtains task scheduling sequence according to the instruction and task scheduling information;

[0012] S3, according to the task scheduling sequence, edge controller sends the instruction of collecting data information to data acquisition equipment, and receives the data information fed back by data acquisition equipment;

[0013] S4, according to the data information fed back by data acquisition equipment, edge controller carries out data processing using the algorithm model optimized in advance, obtains model calculation result, and sends the model calculation result to upper computer platform.

[0014] Optionally, the edge controller includes: virtual controller control module, K CPU virtual controllers, M NPU virtual controllers, N FPGA virtual controllers;

[0015] The virtual controller control module controls the K CPU virtual controllers, M NPU virtual controllers and N FPGA virtual controllers to distribute.

[0016] Optionally, in S1, the heading machine task information includes:

[0017] Task type, task demand for controller resource size and task demand for controller real-time performance.

[0018] Optionally, in S1, the resource scheduling strategy set in advance includes:

[0019] When the demand for controller real-time performance of task is 50ms level, it is divided into CPU virtual controller to execute;

[0020] When the demand for controller real-time performance of task is 1-10ms level, it is divided into FPGA virtual controller or CPU virtual controller to execute;

[0021] When the task is image processing task, it is divided into NPU virtual controller to execute.

[0022] Optionally, in the S1, the task scheduling information comprises:

[0023] the virtual controllers corresponding to different tasks, and the scheduling priority and time slice information of the virtual controllers.

[0024] Optionally, in the S2, generating the task scheduling sequence according to the instruction and the task scheduling information comprises:

[0025] for the virtual controllers not in the same priority, determining the task scheduling sequence according to the priority order;

[0026] for the virtual controllers in the same priority, determining the task scheduling sequence according to the time slice information of the virtual controllers by using a time slice round-robin algorithm.

[0027] Optionally, the method further comprises:

[0028] S0, the edge controller receives an optimized algorithm model sent by the cloud;

[0029] the algorithm model comprises a shield tail gap measurement algorithm, a rock slag identification model, and an automatic spraying and mixing trajectory planning model.

[0030] Optionally, the system further comprises an executor;

[0031] the executor is in communication connection with the edge controller, configured to receive a model calculation result sent by the edge controller, and perform a corresponding action according to the model calculation result.

[0032] Optionally, the system further comprises a plurality of PLC sub-stations;

[0033] the PLC sub-stations are connected between the edge controller and the data acquisition device, and between the edge controller and the executor, configured to transmit data information and model calculation results.

[0034] Optionally, the method further comprises:

[0035] the edge controller saves the task scheduling information to the local end, and generates a local end configuration file.

[0036] (Three) beneficial effects

[0037] The beneficial effects of the present application are: the intelligent control method of the heading machine based on mixed heterogeneous computing provided by the present application can concentrate the intelligent system of the heading machine on a single controller by using the edge computing controller with the mixed heterogeneous architecture of multi-core CPU+FPGA+NPU, thereby reducing the redundancy of the controller hardware; meanwhile, the mixed heterogeneous computing is used to realize the collaborative control of the multi-system tasks of the logic control and intelligent control of the heading machine, thereby guaranteeing the actual effect of the multiple intelligent systems carried by the heading machine; in addition, the data collected by the field terminal device and the processing results of each task are uniformly stored and managed, the interconnection of the underlying data is realized, the terminal data can be conveniently called on the application level, and the data processing efficiency is greatly improved. BRIEF DESCRIPTION OF DRAWINGS

[0038] Figure 1 A flowchart of the intelligent control method of the heading machine based on mixed heterogeneous computing of the embodiment of the present application is shown in the figure.

[0039] Figure 2 A system architecture diagram of the intelligent control system of the heading machine based on mixed heterogeneous computing in the embodiment 2 of the present application is shown in the figure.

[0040] Figure 3 A configuration diagram of the edge controller in the embodiment 2 of the present application is shown in the figure. DETAILED DESCRIPTION

[0041] In order to better explain the present application and facilitate understanding, the present application is described in detail by specific implementation manners in combination with the accompanying drawings.

[0042] The intelligent control method of the heading machine based on mixed heterogeneous computing provided by the embodiment of the present application uses the edge controller with the mixed heterogeneous architecture of multi-core CPU+FPGA+NPU, determines the scheduling information of each task according to the task information and resource scheduling strategy of different tasks, and solves the technical problems in the prior art, such as low execution efficiency, waste of controller resources, complex data interaction communication of multiple hardware systems, and difficulty in realizing resource sharing and coordinated control of multiple tasks, thereby realizing the collaborative control of real-time tasks such as logic control and motion control of the heading machine and intelligent application non-real-time tasks, and improving the resource scheduling and task execution efficiency and data interaction efficiency.

[0043] In order to better understand the above technical solutions, the exemplary embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to enable a clearer, more thorough understanding of the present application and to convey the complete scope of the present application to those skilled in the art.

[0044] DETAILED DESCRIPTION

[0045] Embodiment 1

[0046] Referring to Figure 1 The intelligent control method of the tunneling machine based on hybrid heterogeneous computing in the embodiment is applied to an intelligent control system of the tunneling machine, and the system comprises an upper computer platform for sending an instruction of executing a task to an edge controller and displaying a model calculation result, a cloud end for optimizing an algorithm model, a data acquisition device for acquiring data information, and the edge controller. The intelligent control method of the tunneling machine comprises the following steps.

[0047] In step S1, the edge controller acquires task information of the tunneling machine, and acquires task scheduling information according to the task information and a resource scheduling strategy set in advance.

[0048] In step S2, the edge controller receives an instruction of starting to execute a task sent by the upper computer, and acquires a task scheduling sequence according to the instruction and the task scheduling information.

[0049] In step S3, the edge controller sends an instruction of acquiring data information to the data acquisition device according to the task scheduling sequence, and receives data information fed back by the data acquisition device.

[0050] In step S4, the edge controller performs data processing by using the algorithm model optimized in advance according to the data information fed back by the data acquisition device, acquires a model calculation result, and sends the model calculation result to the upper computer platform.

[0051] The intelligent control method of the tunneling machine based on hybrid heterogeneous computing in the embodiment can deploy an intelligent system of the tunneling machine on a single edge controller, thereby reducing the redundancy of the controller hardware. Meanwhile, the collaborative control of the logical control, intelligent control and multiple system tasks of the tunneling machine is realized by the calculation based on hybrid heterogeneity, thereby guaranteeing the actual effect of the multiple intelligent systems carried by the tunneling machine. In addition, the data information acquired by the data acquisition device is fed back to the edge controller in the embodiment, thereby realizing the interconnection and intercommunication of the underlying data, enabling the upper computer to conveniently call the data information, and greatly improving the data processing efficiency.

[0052] Embodiment 2

[0053] The intelligent control method of the tunneling machine based on hybrid heterogeneous computing in the embodiment is applied to an intelligent control system of the tunneling machine, and the system comprises an upper computer platform for sending an instruction of executing a task to an edge controller and displaying a model calculation result, a cloud end for optimizing an algorithm model, a data acquisition device for acquiring data information, and the edge controller. The intelligent control method of the tunneling machine comprises the following steps.

[0054] Step S1: The edge controller obtains the task information of the roadheader, and obtains the task scheduling information based on the task information of the roadheader and the resource scheduling strategy set in advance;

[0055] Step S2: The edge controller receives the instruction to start executing the task from the host computer, and obtains the task scheduling order according to the instruction and task scheduling information;

[0056] Step S3: According to the task scheduling order, the edge controller sends a data collection instruction to the data collection device and receives data information fed back by the data collection device;

[0057] Step S4: Based on the data information fed back by the data acquisition device, the edge controller uses the pre-optimized algorithm model to process the data, obtain the model calculation results, and send the model calculation results to the host computer platform.

[0058] In the specific implementation process, see Figure 2 The data acquisition equipment consists of various sensors, including temperature sensors, pressure sensors, distance sensors, and cameras. Each system of the tunnel boring machine incorporates data acquisition equipment based on its functional requirements and design. For example, the rock slag analysis system's data acquisition equipment includes a camera, the shield tail gap measurement system's data acquisition module consists of multiple cameras, and the automatic spray mixing system's data acquisition module comprises distance sensors, inclination sensors, and travel sensors.

[0059] The edge controller includes: a virtual controller control module, K CPU virtual controllers, M NPU virtual controllers, and N FPGA virtual controllers;

[0060] The virtual controller control module controls K CPU virtual controllers, M NPU virtual controllers, and N FPGA virtual controllers for allocation.

[0061] During the specific implementation process, the virtual controller module can run and control multiple virtual controllers configured with virtual CPUs, NPUs, and FPGAs. The physical controller resources corresponding to the virtual controllers are determined through the mapping relationship between the physical controller resources of the tunnel boring machine and the virtual controllers. In addition, control modules for various tasks of the tunnel boring machine are deployed on each virtual device.

[0062] Control modules related to the tunnel boring machine tasks are deployed on each virtual controller. These modules are responsible for executing specific control logic. By deploying these control modules on the virtual controllers, parallel processing of tasks can be achieved, thereby improving overall efficiency.

[0063] In this embodiment, each virtual controller uses virtualization technology to isolate interference between tasks, thereby ensuring stability and security.

[0064] In this embodiment, the tunneling machine task information in step S1 includes:

[0065] the task type, the demand of the task on the controller resource size, and the demand of the task on the controller real-time performance.

[0066] Optionally, in step S1, the pre-set resource scheduling strategy includes:

[0067] when the demand of the task on the controller real-time performance is 50 ms level, the task is divided into the CPU virtual controller for execution;

[0068] when the demand of the task on the controller real-time performance is 1-10 ms level, the task is divided into the FPGA virtual controller or the CPU virtual controller for execution;

[0069] when the task is an image processing task, the task is divided into the NPU virtual controller for execution.

[0070] In this embodiment, the tasks with low real-time performance requirements in the logical control and the process control are divided into the CPU virtual controller for execution, so as to realize the control requirements with medium and low time delay (50 ms level); the tasks such as the automatic spray mixing system and the trajectory planning task of the mechanical arm are divided into the FPGA virtual controller or the CPU virtual controller for execution, so as to meet the processing requirements with low time delay (1 ms-10 ms); the image processing tasks such as the shield tail gap measurement system, the rock slag analysis system, and the belt conveyor slag volume flow monitoring system which have large demand on the computing resources are divided into the NPU virtual controller for execution.

[0071] In step S1, the task scheduling information includes:

[0072] the virtual controller corresponding to different tasks, and the scheduling priority and the time slice information of the virtual controller.

[0073] Optionally, in step S2, according to the instruction and the task scheduling information, the task scheduling sequence is generated, including:

[0074] for the virtual controllers not in the same priority, the task scheduling sequence is determined according to the priority order;

[0075] for the virtual controllers in the same priority, the time slice round algorithm is adopted, and the task scheduling sequence is determined according to the time slice information of the virtual controller.

[0076] In this embodiment, the time slice round algorithm is a classical algorithm for scheduling multiple tasks, which divides events into several time slices, each task runs for a period of time in a time slice, and then switches to the next model.

[0077] For example, referring toFigure 3 , assuming that the task of slag analysis is matched to the virtual NPU0, the priority thereof is 11, and the time slice information thereof is 10, and the task of shield tail gap measurement is matched to the virtual NPU1, the priority thereof is 10, and the time slice information thereof is 8; that is, the priority of the virtual NPU0 is higher than the priority of the virtual NPU1, and therefore the task scheduling order is determined according to the priority order, that is, the task of slag analysis is performed first;

[0078] Assuming that the logical control task 1 is matched to the virtual CPU0 in the physical CPU0, the priority thereof is 6, and the time slice information thereof is 5, and the logical control task 2 is assigned to the virtual CPU1 in the physical CPU0, the priority thereof is 6, and the time slice information thereof is 5; that is, the priorities of the virtual CPU0 and the virtual CPU1 are the same, and at this time, the logical control task 1 is run for 1 time slice first, the logical control task 2 is run for 1 time slice second, and the two tasks are alternately run until the end of the time slice.

[0079] In this embodiment, before step S1, the following steps are further included:

[0080] Step S0, the edge controller receives the optimized algorithm model sent by the cloud;

[0081] The algorithm model includes: a shield tail gap measurement algorithm, a slag identification model, and an automatic mixed spraying trajectory planning model.

[0082] When the cloud optimizes the algorithm model, the training data set is input into the initialized algorithm model, a cross-entropy function and an Adam optimizer are used, the learning rate is 0.001, and the optimized slag analysis model is obtained until the cross-entropy regression loss function is minimized. The training data set is historical data; for example, when the slag analysis model is optimized, the training data set is a slag historical image, a slag classification result corresponding to the slag historical image, and a surrounding rock grade identification result.

[0083] The tunneling machine intelligent control system in this embodiment further includes an actuator;

[0084] The actuator is in communication connection with the edge controller, and is configured to receive the model calculation result sent by the edge controller and perform a corresponding action according to the model calculation result.

[0085] The tunneling machine intelligent control system in this embodiment further includes a plurality of PLC sub-stations;

[0086] The PLC sub-stations are connected between the edge controller and the data acquisition device and between the edge controller and the actuator, and are configured to transmit data information and model calculation results.

[0087] For example, referring to Figure 2The sensors such as temperature sensors, pressure sensors, distance measuring sensors, etc. are responsible for collecting data information, and are connected to the edge controller through a PLC substation to transmit data signals, as shown in FIG. Figure 2

[0088] When the data acquisition device is a camera, the PLC substation can be omitted, and the camera can be directly connected to the edge controller to transmit images of rock slag, images of a shield tail gap, and video data.

[0089] In addition, when the data acquisition device is a camera, a switch can be arranged between the camera and the edge controller to manage and optimize network communication, thereby ensuring smooth transmission of image or video data and avoiding delay and packet loss.

[0090] In the embodiment, the intelligent control method of the heading machine further includes:

[0091] The edge controller saves the task scheduling information to the local end to generate a local end configuration file.

[0092] The intelligent control method of the heading machine based on the hybrid heterogeneous computing can also carry a risk identification system, an intelligent heading system, and other systems for intelligent control of the heading machine.

[0093] The intelligent control method of the heading machine based on the hybrid heterogeneous computing, by using the edge computing controller with the multi-core CPU+FPGA+NPU hybrid heterogeneous architecture, can centrally deploy the intelligent system of the heading machine on a single controller, reducing the redundancy of the controller hardware; the hybrid heterogeneous computing is used to realize collaborative control of multiple system tasks of the logic control and intelligent control of the heading machine, ensuring the actual effect of the multiple intelligent systems carried by the heading machine; in addition, the application uniformly stores and manages the data collected by the field terminal device and the processing results of each task, realizes the interconnection and intercommunication of the underlying data, enables the application layer to conveniently call the terminal data, and greatly improves the data processing efficiency.

[0094] In the description of the present application, it should be understood that the terms "first" and "second" are only for the purpose of description, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first" and "second" can explicitly or implicitly include one or more of the features. In the description of the present application, the meaning of "multiple" is two or more, unless otherwise specifically limited.

[0095] ​In the present application, unless specifically defined otherwise, the terms "mounting", "connected", "connection", "fixed", "unfixed", and the like should be construed broadly, for example, can be fixed connection, can be detachable connection, or integral; can be mechanical connection, can be electrical connection; can be direct connection, or indirect connection through an intermediate medium; can be internal communication of two elements, or interaction relationship between two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0096] In the present application, unless specifically defined otherwise, the first feature is "on" or "under" the second feature, which can be direct contact between the first and second features, or indirect contact between the first and second features through an intermediate medium. Moreover, the first feature is "above", "over" and "on" the second feature, which can be directly above or obliquely above the first feature, or only indicates that the horizontal height of the first feature is higher than that of the second feature. The first feature is "below", "under" and "under" the second feature, which can be directly below or obliquely below the first feature, or only indicates that the horizontal height of the first feature is lower than that of the second feature.

[0097] In the description of the present application, the description of the terms "one embodiment", "some embodiments", "embodiment", "example", "specific example" or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In the present application, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in any one or more embodiments or examples. In addition, those skilled in the art can combine and combine the different embodiments or examples described in the present application and the features of the different embodiments or examples, without contradiction.

[0098] Although the embodiments of the present application have been shown and described above, it should be understood that the above embodiments are exemplary and should not be construed as limiting the present application, and those skilled in the art can modify, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A method for intelligent control of a heading machine based on hybrid heterogeneous computing, characterized in that, The intelligent control method is applied to an intelligent control system of a heading machine, and the system comprises an upper computer platform for sending an instruction of executing a task to an edge controller and displaying a model calculation result, a cloud for optimizing an algorithm model, a data acquisition device for acquiring data information, and the edge controller. S1, the edge controller acquires heading machine task information, and acquires task scheduling information according to the heading machine task information and a resource scheduling strategy set in advance; S2, the edge controller receives an instruction of starting to execute a task sent by the upper computer, and acquires a task scheduling sequence according to the instruction and the task scheduling information; S3, according to the task scheduling sequence, the edge controller sends an instruction of acquiring data information to the data acquisition device, and receives data information fed back by the data acquisition device; S4, according to the data information fed back by the data acquisition device, the edge controller processes data by using an algorithm model optimized in advance, acquires a model calculation result, and sends the model calculation result to the upper computer platform; The edge controller comprises a virtual controller control module, K CPU virtual controllers, M NPU virtual controllers, and N FPGA virtual controllers. The virtual controller control module controls the K CPU virtual controllers, the M NPU virtual controllers, and the N FPGA virtual controllers to be distributed, determines the physical controller resources corresponding to the virtual controllers through the mapping relationship between the physical controller resources of the heading machine and the virtual controllers, and executes specific control logic. Each virtual controller is deployed with a control module related to a heading machine task. In S1, the resource scheduling strategy set in advance comprises: When the real-time requirement of a task for a controller is 50 ms, the task is divided into a CPU virtual controller for execution; When the real-time requirement of a task for a controller is 1-10 ms, the task is divided into an FPGA virtual controller or a CPU virtual controller for execution; When the task is an image processing task, the task is divided into an NPU virtual controller for execution.

2. The intelligent control method for the boring machine based on hybrid heterogeneous computing according to claim 1, characterized in that, In S1, the heading machine task information comprises: a task type, a demand of a task for a controller resource size, and a demand of a task for a controller real-time.

3. The hybrid heterogeneous computing based intelligent control method of a heading machine according to claim 1, characterized in that, In S1, the task scheduling information comprises: a virtual controller corresponding to different tasks, and a scheduling priority and time slice information of the virtual controller.

4. The intelligent control method for the boring machine based on hybrid heterogeneous computing according to claim 1, characterized in that, In S2, according to the instruction and the task scheduling information, the task scheduling sequence is generated, which comprises: for virtual controllers not in the same priority, determining the task scheduling sequence according to the priority order; for virtual controllers in the same priority, using a time slice round-robin algorithm to determine the task scheduling sequence according to the time slice information of the virtual controllers.

5. The hybrid heterogeneous computing based intelligent control method of a heading machine according to claim 1, characterized in that, Before S1, there is also S0, the edge controller receives an optimized algorithm model sent by the cloud. The algorithm model comprises a shield tail gap measurement algorithm, a rock slag recognition model, and an automatic spray mixing trajectory planning model. The system further comprises an executor.

6. The hybrid heterogeneous computing based intelligent control method of a heading machine according to claim 1, characterized in that, ​ The executor is in communication connection with the edge controller, configured to receive a model calculation result sent by the edge controller, and perform a corresponding action according to the model calculation result.

7. The hybrid heterogeneous computing based intelligent control method of a heading machine according to claim 6, characterized in that, The system further comprises a plurality of PLC sub-stations. The PLC sub-stations are connected between the edge controller and the data acquisition device and between the edge controller and the executor, configured to transmit data information and model calculation results. 8.The hybrid heterogeneous computing based intelligent control method of a heading machine according to claim 1, wherein, The method further comprises: The edge controller saves the task scheduling information to the local end to generate a local end configuration file.

Images