Research on optimisation of transformer manufacturing process and construction of smart factory based on digital twin technology

In the enterprise production process, the large consumption of manufacturing resources not only aggravates the enterprise production costs or causes environmental pollution is one of the main reasons, in the transformer manufacturing process involved in the resource consumption category has diversified, so the transformer manufacturing and processing process resource environmental attributes are analysed as shown in Figure 1. The resource consumption attributes contained in the material, energy, manpower, and three types of resource consumption are analysed in detail for transformer parts of the machining process. The precise content is as follows: 1)

Material consumption

Transformer manufacturing process material consumption mainly includes the consumption of raw materials and auxiliary materials. The raw materials required in the manufacturing process of transformer components are mainly silicon steel sheets, insulation materials, steel, and so on. Auxiliary materials include grid cloth, molds, thermostats, welding wire, and other similar items. The material consumption process is usually expressed by the material flow process shown in Figure 2, that is, the raw materials through the equipment, operators and auxiliary facilities of the mutual co-operation into finished or semi-finished products.

In the process of enterprise production and manufacturing parts, the material flow mainly includes production and processing, packaging, transport, inspection, assembly, and other components.

Processing can change the structure and properties of raw materials, allowing them to be transformed into parts. Transfer is the process of moving workpieces between work centers to achieve process transfer. Storage refers to the waiting of parts before processing and the storage in the warehouse after processing, while inspection mainly refers to the quality control of the material flow.

Figure 1.

Resource and environmental attributes of transformer manufacturing process

Figure 2.

Material flow process

Transformer components are used in the mechanical manufacturing process to input raw materials, produce and process output products, while generating environmental impacts. In a sense, the environmental attributes are a kind of worthless waste, but in fact, the process can also be optimised to improve recycling so that the environmental attributes can be into other systems of valuable resources. The environmental attributes of the transformer manufacturing process are shown in Figure 3, which mainly include the emission of pollutants such as sulfur dioxide, soot, nitrogen oxides, and so on. 1)

Gas waste

The process of transformer manufacturing will bring a certain amount of air pollution at the same time, and when it reaches a certain level, it will affect the health of the body and destroy the ecological balance. The main emissions in the processing of waste gas are dust, welding gas, and organic gases. Dust is mainly produced during the processing of insulating parts, welding gas is mainly produced during the winding process of the coil using welding rods to weld the renewal line, and organic gas is mainly produced during the curing process of the coil, casting a small amount of VOCs (the main component is non-methane total hydrocarbons).

Figure 3.

Environmental properties of transformer manufacturing process

In the manufacturing process of transformers, the key elements involved in the production activities include production line equipment, products in process, staff, etc., while the production environment will have an impact on the manufacturing process of the product. Taking into consideration the above factors, this paper adopts a formal modeling language to model the key elements of the transformer manufacturing process, and the model is defined as follows: 1PS::=PE∞PP∞PW Where PS represents the physical space of the key elements of the production line manufacturing process, PE represents the set of production line equipment in the physical space. PP represents the set of products under production in the physical space, and PW represents the set of staff in the physical space. ∞ represents the natural connection between PE, PP, and PW, indicating the natural interaction between the three. PE, PP, PW are all dynamic sets, the set of elements and their states are constantly updated with the dynamic operation of the manufacturing process. 1)

Production line equipment entity modelling

A formal modelling language is used to model the physical equipment of the intelligent production line, and the model is defined as follows: 2PE={ PE1,⋯,PEi } 3PEi={ EPi,SDSet,ELog } 4SDSet={ SDi1,⋯,SDim,⋯,SDin },∀m∈[1,n] 5SDim={ Stype,DSet } Where PE_i denotes the i nd automation device in the physical space, EP_i denotes the model and technical parameters of the i th automation device, SDSet denotes the set of n sensor data configured during the operation of the i th automation device, and ELog denotes the log information during the execution of the automation device’s actions. SD_im denotes the model of the mth sensor in the i th automation device and the set of data collected in real time, and S_type denotes the model of the sensor. DSet represents the set of data collected in real time by the m th sensor, which covers energy, vibration, power, etc., depending on the sensor model.

The digital twin model needs to be highly modular, well scalable and dynamically adaptable, and the construction of the model can be done in the information space using parametric modelling methods. A virtual model of physical entities is built in software such as Tecnomatrix, Demo3D, Visual Components, etc. The virtual model contains not only a description of the geometric information and topological relationships of the intelligent production line, but also a complete dynamic engineering information description of each physical object [20]. Then, multiple dimensional attributes of the model are parametrically defined to achieve real-time mapping of the manufacturing process of the smart production line. A formal modelling language is used to model the digital twin model in the information space, and the model is defined as follows: 9CS::=DE∞DP∞DW where CS represents the information space of key elements of the production line manufacturing process, DE represents the set of digital twin models of production line equipment in the information space. DP represents the set of digital twin models of in-process products in the information space, and DW represents the set of digital twin models of staff in the information space. ∞ denotes the natural connection between DE , DP and DW , which shows the natural interaction between the three. DE , DP, and DW are all dynamic collections, where the collection elements and their states in the information space are updated synchronously with the dynamic operation of the manufacturing process in the physical space.

On the basis of establishing the physical space entity model PS and the information space twin model CS , the virtual-reality mapping association between them is further established, and their virtual-reality mapping association relationship is modelled using a formal modelling language, and the model is defined as follows: 10PS::=PE∞PP∞PW 11CS::=DE∞DP∞DW 12PS↔1:1CS Where ↔1:1 denotes the bidirectional real mapping between the physical space entity model and the information space twin model, and ∞ denotes the natural connection between different models. As a result, it can be concluded that entity device PE and twin device DE , entity product PP and twin product DP, entity person PW and twin person DW should keep synchronous bidirectional real mapping with each other, as shown in Fig. 4.

Figure 4.

Virtual real mapping association modeling

The difficulties in realising a digital twin system for transformer manufacturing process optimisation are mainly in the following aspects [21]: 1)

Adopting fusion technology based on model definition (MBD) and data-driven modelling and simulation methods to construct a digital twin model with high fidelity in the information space that is consistent with the physical space, and to achieve full visualization of the manufacturing process.

The design of optimisation algorithms in the optimisation module of the green manufacturing process of transformers begins with the selection of suitable optimisation algorithms and the determination of optimisation algorithm variables, and then completes the design of the optimisation algorithm model and other tasks. In the process of machine tool manufacturing, machine tool feed rate and spindle power are important parameters that affect the performance and quality of machine tool machining as well as tool wear, and the machine tool feed rate is mainly affected by the feed rate. Machine tool spindle power changes with changes in depth of cut, spindle speed, and other parameters. In general, an increase in the depth of cut will often lead to an increase in machine tool spindle power, this time by reducing the machine tool feed rate to protect the tool. While the reduction of the depth of cut will lead to a decrease in the spindle power of the machine tool, at this time can be improved by increasing the machine tool feed rate to enhance the efficiency of the machine tool. Constant power adaptive control of the machine tool manufacturing process can be achieved by using the machine tool spindle power and feed rate as the control variable and controlled parameter of the optimisation algorithm.

The fuzzy control algorithm has the characteristics of good flexibility, high stability, fast response, high control accuracy, and strong nonlinear computing ability, combined with the characteristics of the processing parameters in the machine tool manufacturing process and the algorithm performance requirements. In this paper, the fuzzy control algorithm is used to achieve constant power adaptive control of the machine tool manufacturing process, and the research flow of the fuzzy control algorithm in the machining process optimisation module is shown in Figure 5. According to the overall design analysis of the system, the main function of the human-computer interaction interface of the machining process optimization module is to provide the method of optimization algorithm calling and algorithm parameter adjustment in order to realize the system’s adaptive intelligent control of the machine tool manufacturing process, and to improve the system interactivity and parameter adjustment capability of the optimization process, etc.

Figure 5.

Research flow of fuzzy control algorithm

Referring to the design and development scheme of the Manufacturing Process Monitoring Module HMI, the Manufacturing Process Optimisation Module HMI development is implemented by using the UGUI module of Unity3D software together with the C# programming language and Visual Studio programming environment.

The manufacturing process optimisation module achieves constant power adaptive control of the machine tool manufacturing process by obtaining the spindle power in real-time, outputting the target feed rate of the machine tool using the optimisation algorithm, and transmitting the target feed rate of the machine tool to the machine tool numerical control system in real-time. Therefore, in order to ensure the real-time transmission of the optimisation parameters, a stable and efficient data transmission link must be established between the manufacturing process optimization module and the CNC machine tool.

With reference to the data transmission method of the manufacturing process monitoring module, the transmission of optimisation parameters in the manufacturing process optimisation module is performed by using computer software to establish the server and client based on the OPC communication protocol, and the values of the optimisation parameters are transmitted in real time to the CNC system via the Ethernet interface of the machine tool CNC.

In order to determine the main research content of the manufacturing process monitoring and optimisation system and improve the development efficiency of the system, this paper focuses on the overall design of the manufacturing process monitoring and optimisation system, and the main work completed is as follows: 1)

The overall design objectives of the manufacturing process monitoring and optimisation system are proposed, and the overall functions and development requirements of the system are studied based on the modularisation idea, so as to facilitate the guidance of the design and development of the system’s functional modules.

A catalytic cracking unit is an important part of a transformer production line, and as a typical reaction regeneration system, the process of a catalytic cracking unit consists of 5 main facilities and 40 control points, as shown in Table 1. The objective of the optimized control in this example is to enhance the unit production and ultimately reduce energy consumption in the manufacturing process of transformers. A total of 420 metrics can be collected from the production line online analysis and metering instrumentation, as well as other online systems that may affect the control objective, and the 40 controllable points also belong to the combination of these 420 metrics, so there are still 380 metrics that belong to the uncontrollable type.

According to this paper, we design a digital twin-based optimisation method for the transformer manufacturing process, using machine learning techniques to obtain data and train the digital twin model, with the goal that the model can most accurately simulate the physical environment of the actual factory.

Based on the modelling approach of key element entities, digital twin virtual and virtual-reality mapping associations, the basic formula for the construction of a digital twin model plant for transformer production control optimisation is Y_t = F(X_t±{Δ} + Z_t±{Δ}). Specifically in this example, Y is the output of the finished transformer product, X is the 40 controllable points, and Z is the other 370 uncontrollable metrics. Once the model has been trained, it will be deployed online as the information mirroring part of the digital twin in order to perform simulations based on real-time data from the production line and continuously provide real-time control optimisation recommendations. The use of PCC as a method of analysing the correlation significance of high dimensional data was used to generate the correlation matrix shown in Figure 6. The red color indicates a positive correlation between two sets of data, while the green color indicates a negative correlation. Regardless of the color, the darker the color, the stronger the correlation between the two dimensions.

Figure 6.

Correlation Matrix Map of part Catalytic Cracking Unit Indicators

The data used in this paper comes from the real-time database within the project enterprise, which contains historical and real-time data from the following systems, Manufacturing Execution System MES, Laboratory Information Management System LIMS, and Planning Optimisation Modelling System PIMS. The system is based on three points in time such as 31st January, 11th February, and 21st March 2021, respectively, and the data set of the first 7 months of the data is taken as the training set, and the data of the last 15 days as the test set, and the data are fused on a second-by-second basis.

Table 2 sets the original machining conditions and the changed working conditions. From the table, it can be seen that the new working conditions changes occur in the workpiece, tool, cutting parameters and machining features. The original machining condition is set as φ6 and the new working condition optimised by digital twin technology is φ8, in which 6mm and 10mm holes in transformer manufacturing features respectively.

After extracting the main features of the 5 main facilities and 40 control points through the digital twin model, the predictive algorithm model is trained based on the training dataset. Finally, a test set is added to the model for validation. Figure 7 shows the prediction results of some experiments, from which it can be seen that, despite the good randomness of the product burr generation, however, the production results of the established digital twin model under the new working condition are basically consistent with the actual standard of the product. And in the new working conditions, the burrs of the transformer products are significantly smaller than the original processing conditions. 20 samples in the new working conditions of manufacturing conditions have a diameter of less than 0.02mm, effectively improving the accuracy of transformer production.

Figure 7.

Test results

In the smart factory environment based on the digital twin model, for the production orders used for testing, the remaining order delay rates before and after testing are shown in Figure 8 below. The horizontal coordinate is the production order, and the vertical coordinate is the remaining order delay rate. The probability of delays in the remaining orders decreases when the delay rate is lower. Combined with the actual express production situation, orders are delayed when the remaining time is less than 24 hours, for example when the shipment is made within 24 hours. When the delay rate for the remaining orders is zero, it means that they do not need to complete production on the same day. In this test case, according to the delay rate before and after the test order, it can be seen intuitively that the delay rate of the remaining orders of the order before the test is high and has been accompanied by the period of production, and the delay rate of the remaining orders after the test is gradually decreasing, and in the middle and late stages of the production, it is reduced to zero.

Figure 8.

Comparison chart of remaining order delay rate

In summary, this paper addresses the processing characteristics of the green manufacturing process of transformers, integrates the application scenarios of each field, and verifies the proposed intelligent manufacturing process optimisation method based on the digital twin model through a processing example of a transformer enterprise. 1)

According to the machining process, information about the machine tools and products of the machining process is collected, and the digital twin model with time-varying evolution is constructed from the data of the machining process. Various factors (i.e., machining process, manufacturing features, and other elements) are considered.