Intelligent Process Design System for Human–Robot Collaboration in Helicopter Assembly ^†

Abstract

Traditional manual assembly is limited in terms of both efficiency and quality. In contrast, robots are characterized by rapidness and accuracy and can cooperate with humans to perform complex tasks. Human–robot collaboration may hold the potential to enhance the manufacturing capacity of the helicopter industry. However, the traditional assembly process design methods based on personal experience can hardly adapt to the transformation of manufacturing mode, which makes deploying human–robot collaborative assembly inefficient. In this paper, we systematically analyze applications of human–robot collaboration in helicopter fuselage assembly. Concretely, an automatic drilling and riveting process based on human–robot collaboration is designed and verified. Moreover, we develop an intelligent process design prototype system that is specifically designed for human–robot collaborative assembly by modeling and integrating process knowledge. It can effectively assist human designers by means of recommending equipment selection, process parameters, and numerical control programs. Taking a fuselage assembly process design as an example, we verify that the prototype system can improve both the management of process knowledge and the efficiency of process design.

1. Introduction

Fuselage assembly is a crucial procedure in helicopter manufacturing. The workload of fuselage assembly accounts for about 30% of the total workload of a helicopter due to the complex shape of helicopters and the large number of components and connectors. Improving the quality and efficiency of fuselage assembly is the key to enhancing the helicopter manufacturing capacity. Currently, domestic helicopter manufacturing enterprises still mostly adopt traditional manual drilling and riveting assembly methods, which results in high labor costs and overhead. In contrast, collaborative robots have advantages in cost effectiveness, simple deployment, and flexible usage [1]. By integrating vision sensors, they can smartly identify and locate objects in complex environments, thereby cooperating with workers to accomplish dynamic and complex assembly tasks. Human–robot collaboration technology has been widely applied in several industries, such as electronics and automotive [2]. And, it is also promising to leverage human–robot collaboration to improve the manufacturing capacity of the helicopter industry.

As illustrated in Figure 1, process design is the bridge between products and manufacturing. In the case of helicopters, multiple product models are often derived from the same prototype, which leads to highly frequent and similar process design activities. However, the traditional process design method based on personal experience can hardly meet the demands of multi-variety and small-batch production mode [3], which has limited the improvement of efficiency in helicopter manufacturing. Firstly, the process design methods used by enterprises are mainly geared towards traditional manual operations, with limited support for human–robot collaborative assembly. The key to human–robot collaboration lies in determining the appropriate selection of devices and compiling corresponding numerical control programs, which involves knowledge and skills significantly different from the traditional manual assembly. Secondly, various types of information required during the process design have not been integrated. For example, manufacturing resource management, process information management, and production planning management are maintained separately in different departments, making it difficult to effectively correlate and utilize them. Thirdly, the existing process design platforms cannot assist technicians effectively. Technicians still need to manually determine parameters and equipment based on personal experience, which influences efficiency and standardization.

Addressing the aforementioned issues, this paper explores the applications of human–robot collaboration in helicopter fuselage assembly and preliminarily verifies its feasibility in processes, such as drilling and riveting. We construct an intelligent process design prototype system specifically for human–robot collaborative assembly, which integrates manufacturing resources, process information, and process design. By effectively reusing historical process knowledge, it is capable of suggesting appropriate parameters and devices. This capacity contributes to alleviating the dependence of process design on personal experience and enhances the efficiency and quality of process design activities.

2. Human–Robot Collaboration in Helicopter Assembly

Traditional industrial robots primarily work within safety barriers that prevent contact with humans [4]; thus, they cannot meet the requirements for human–robot collaboration. Consequently, collaborative robots have emerged, which can significantly enhance the rapid response capability in flexible manufacturing. By integrating various advanced sensors, their adaptability to unstructured scenarios is further improved. The production mode based on collaborative robot technology combines the precision of robots with the flexibility of humans and will play an important role in helicopter assembly.

2.1. Collaborative Assembly Framework

Currently, there are several applications of human–robot collaborative assembly in the aviation industry abroad. For instance, Airbus [5] has applied a Human–Robot Collaborative Automatic Drilling Unit (ADU) for the assembly of the outboard flaps on the A350 aircraft. The A350’s outboard flap assembly requires drilling at eight different locations. Previously, workers had to manually hold a 4kg ADU to perform tasks such as positioning, clamping, and drilling. Airbus adopted a seven degrees of freedom collaborative robot from KUKA, weighing 14kg, to carry the ADU and accomplish the drilling operation. Airbus has also developed a human–robot collaborative self-piercing rivet (SPR) device based on the UR10 robot, which is used for the installation of the tail cone wall on the A320 aircraft. The robot absorbs the impact and vibration generated during the self-piercing riveting operation, thus protecting the operators. The Fokker Company [6] has employed a UR10 collaborative robot equipped with a glue dispensing end effector, using a manual teaching method for robot operation programming, to achieve the gluing operation of the landing gear bushings.

The above human–robot collaboration cases are sporadic applications in certain aircraft assembly scenarios, primarily serving as proof of technical feasibility, and have yet to evolve into large-scale technological application systems. Moreover, the positional accuracy attainable by the manual teaching method of collaborative robots is limited, which fails to meet the precision requirements for helicopter assembly. A more effective approach would be to integrate intelligent sensors into the robots, allowing for automatic robot guidance and control through visual recognition and positioning, thereby improving precision and intelligence. Collaborative robots can also be mounted on Automatic Guided Vehicles (AGVs), gantry frames, or movable rails installed on the ground to further extend their working range.

As shown in Figure 2, we sort out the application scenarios of human–robot collaboration in helicopter fuselage assembly and design corresponding implementation schemes, systematically conducting research on human–robot collaboration applications. The aim is to improve and solve practical production issues such as operating in confined spaces with awkward postures, performing single repetitive tasks, enduring long periods of weight bearing, and exposure to harmful substances. The initial achievement of human–robot collaborative assembly is enhancing production efficiency and reducing labor intensity. Subsequently, we will continuously carry out in-depth applications and gradually increase the work proportion of robots, realizing low-cost, low-risk, highly efficient automation for green and intelligent manufacturing.

2.2. Collaborative Drilling and Riveting

At the current stage, robots still cannot achieve the flexibility and adaptability of humans and do not possess the capability of autonomous decision making and motion planning. Therefore, in human–robot collaboration, it is crucial to reasonably delineate the interface between humans and robots and to compile numerical control programs that specify the robots’ movement paths. This paper designs an automated drilling and riveting process based on human–robot collaboration, as shown in Figure 3, in the context of a typical assembly scene for helicopter fuselage assembly. We detail several important steps as follows:

(1)

The preparation phase includes the installation and calibration of the robot, vision sensor, and end effectors.

(2)

The image recognition unit is used to detect the datum hole and calculate position compensation for the drilling operation. The laser device is used to measure the normal angle offset at the drilling position. If the offset is greater than 0.5 degrees, the robot’s posture is automatically adjusted. If the adjustment fails to achieve the correct posture after three attempts, the system will issue an error alert and pause the operation.

(3)

After moving to the drilling position, the drilling spindle mounted at the robot end begins to rotate and feed according to the inputted drilling parameters, and then the chip suction and dust removal unit is activated. Once the spindle reaches the pre-set extension, it begins to retract. When the drill has completely exited the workpiece surface, the chip suction and dust removal unit stop. After the spindle retracts to a proper position, the drilling operation is complete.

(4)

A worker manually inserts the rivet into the hole and firmly presses it against the workpiece surface with a rivet hitter held in hand. Guided by the vision unit, the robot moves so that the elastic anvil at the end touches the rivet. The human–robot interaction system sends a start signal to the worker, who then begins the riveting operation. During riveting, the robot’s end effector continuously provides feed force, as measured in real time by the pressure sensor. After several hammerings, the pose adjustment unit measures the position of the anvil and compares it with the initial position to determine the change in rivet height. If the rivet head is up to standard, then the riveting operation at this location is complete; otherwise, riveting continues for several more hammerings until the criterion for the rivet head is reached.

(5)

The robot equipped with the end effector moves to the next drilling position calculated by the vision unit, preparing for the next cycle of drilling and riveting.

Figure 3. (a) Human–robot collaborative drilling and riveting process. (b) Human–robot collaboration assembly process design.

Figure 3. (a) Human–robot collaborative drilling and riveting process. (b) Human–robot collaboration assembly process design.

3. Intelligent Process Design System

Most existing process design systems [7,8,9,10] are developed for the machining process of workpieces. According to the similarity of workpiece features, typical process templates are invoked for process design. However, there is a significant difference between assembly processes and machining processes. Assembly processes use a large number of fasteners to permanently connect multiple components together. The connection methods and connection positions involved are various, which leads to the complexity of the feature extraction and similarity measurement for the assembly processes. From another perspective, the assembly process routes are quite fixed [11] and generally consist of several main operations including drilling, riveting, and sealing. We are able to select devices and compile numerical control programs based on typical templates and several process parameters, which we call query parameters in this paper. Therefore, it is suitable to develop an intelligent assembly process design system at the operation level.

3.1. System Architecture

The process design prototype developed in this research adopts Browser–Server architecture. The system is divided into an interaction layer, a function layer, and a data resource layer, as shown in Figure 4.

In the data resource layer, we adopt an object-oriented approach to establish a process information model for human–robot collaborative assembly. By analyzing the process characteristics of human–robot collaborative assembly, we abstractly represent the assembly process as objects, including manufacturing resources, process information, parameter information, and decision-making knowledge. Each type of object has specific attributes and data types, and a complete description of the assembly information is formed through the integrated association between objects. Among them, the manufacturing resource object covers equipment such as robots, sensors, end effectors, and toolings. The parameter information object is compatible with various data types such as numerical parameters, enumeration parameters, and character parameters. The decision-making knowledge object is compatible with common decision-making forms such as formula decision relationships, rule decision relationships, and query decision relationships. The process information object integrates multiple equipment, parameters, and decision rules, and, therefore, is represented by a dictionary data structure.

The function layer is the core of system functionality implementation. To reduce the coupling between different functions within the system, we divide the prototype system into two major subsystems: process knowledge management and intelligent process design. The process knowledge management subsystem includes modules for managing manufacturing resources, process information, parameter information, decision knowledge, and process design tasks. The intelligent process design subsystem is responsible for implementing the intelligent process design based on process templates. The pipeline mainly includes the creation of design tasks, the determination of process templates, the reasoning of process information, and design task submission. Based on the query parameters and decision knowledge of each procedure in human–robot collaborative assembly, reasoning is conducted for the required equipment selection and process parameters of the procedure, providing intelligent assistance to technicians in the preparation of numerical control programs.

3.2. Intelligent Process Design

The process design system for human–robot collaborative assembly is driven by query parameters and decision knowledge. It makes automatic decisions at the operation level through reasoning rules, generating structured process information and numerical control programs. Then, it navigates the revision of automatically generated process information and numerical control programs by interaction. The detailed pipeline is shown in Figure 3b and mainly includes the following contents:

(1)

Technicians determine the process route based on CAD models and extract query parameters from the operation features.

(2)

The query parameters are used as screening conditions and are compared with the process templates to determine whether there are similar cases to the assembly process. If so, an appropriate process template is selected as the mathematical model for process reasoning; otherwise, it needs to be redesigned.

(3)

A new assembly process design task is created based on the process template, and the process content is decided according to the query parameters. As shown in Figure 5, each object in the process information model generates its attributes through the parameter relationship in the process template. The process design system recommends the versions of devices and provides examples of numerical control programs for technicians to refer to.

(4)

On the interactive editing interface, the process inferred by the intelligent design system is revised in a guided manner. After confirmation by the technicians, the structured process data and NC programs are submitted to the process design task database for unified data management.

Figure 5. A demonstration of intelligent assembly process design.

Figure 5. A demonstration of intelligent assembly process design.

4. Results and Analysis

In order to verify the effectiveness of the proposed intelligent process design system, we invite five process personnel to participate in the simulation experiment. Each of them completes the assembly process, respectively, by manually consulting the documents and using our intelligent process design system. The total time cost is divided into consulting time and modification time, as presented in Figure 6. The consulting time includes the time for preparing the input process parameters and retrieving similar process instances. The modification time refers to the time taken to modify the process items to a reasonable process. As the number of assembly process documents grows, the intelligent process design system constructed in this work can retrieve similar process instances more efficiently. Meanwhile, the modification time of our system is the same as that of the traditional method, which reflects that the process design results suggested by our system can meet the requirements.

5. Conclusions and Future Work

This paper systematically sorts out the application framework of human–robot collaboration in helicopter fuselage assembly and designs a human–robot collaborative automated drilling and riveting process. Additionally, we develop an intelligent process design prototype system for human–robot collaborative assembly, achieving the integration of manufacturing resources, process information, and process design data. Through the reuse of historical process knowledge, it can realize the reasoning of device selection, process parameters, and numerical control programs, thereby intelligently assisting technicians in accomplishing process design. Future work includes promoting human–robot collaborative assembly into practice and further expanding the scope of the process design system in helicopter manufacturing.