A new product development study using intelligent data analysis algorithm based on KE theory

Abstract

The Artificial Neural Networks (ANN) are more widely used in the New Product Development (NPD) process in recent years. The product data generation process is a prerequisite for the application of the ANN algorithm. In the development of new products, the Kansei Engineering (KE) method is an effective emotion-based data generation method. The Semantic Difference (SD) method is usually used to obtain data to apply to design idea generation. Facing the data demand of product creativity, it is important to establish the relationship between consumer perception and product expression. Numerical relationships are not linear and several methods are required for solving these problems. The method of the Back Propagation (BP) neural network is simple and effective to be used in this case. This paper proposes an innovative data modeling method using digital coding and KE. This model explores a rational design method of perceptual intention and builds an intelligent model. Compared with traditional method, the modified model can quickly and accurately reflect the users’ perceptual needs, make the design more scientific, improve the design efficiency, and reduce design costs. This method is used in the design of electric welding machines, and this process can effectively provide technical support for NPD process in small and medium-sized enterprises.

Keywords

New product development KE semantic difference ANN BP

1 Introduction

In the process of New Product Development, the most difficult stage is usually in the concept generation stage. The concept generation is generally considered to have two modes. One mode is derived from the designer’s experience, and the other is driven by the elements of the product itself. With the advent of the data age, especially the development of Big Data (BD) and Artificial Intelligence (AI), studying the driving mechanism of the product’s own elements has increasingly become a new way of developing new products.

In the research of this paper, the SD method of Kansei Engineering constructs the data required for analysis. The input and output of these data are non-linear, and it is not easy to obtain results by conventional linear methods. A relatively simple BP neural network is used to calculate these data, and the valuable guiding rules obtained from the data, play an auxiliary role in product development.

Data analysis of product design has become an effective way of new product development, and the method of applying BP neural network modeling is widely used in new product development. The Back Propagation (BP) neural network is one of the most widely used neural network models. This kind of network is a multi-layer feed forward network trained by error back propagation algorithm. The BP network can learn and store a large number of input-output pattern mapping relationships without revealing the mathematical equations describing this mapping relationship in advance. Its learning rule is to use the steepest descent method to continuously adjust the weights and thresholds of the network through backpropagation to minimize the sum of squared errors of the network.

In the recent years, Wang et al. [1] use some uncertainty techniques including rough sets and fuzzy sets are applied to capture more accurate emotion knowledgeand obtain the weight of the core product design items. In order to explore the non-linear relationship between design elements and kansei images, the fuzzy weighted association rule mining method was applied to obtain the set of frequent fuzzy weighted association rules based on evidence theory’s reliability indices of minimum support and confidence so as to realize user demand-driven product design. Chen et al. [2] propose a fuzzy-logic-based emotional evaluation scheme for color pattern design in aircraft cockpits. In traditional Miryoku engineering, Wu et al. [3] construct the product Kansei factors based on the qualitative analysis method. they proposed an evaluation-fuzzy-quantification model based on users’ Kansei.

In the process of using artificial neural network to develop new products, data acquisition is a key point, and data acquisition using the method of the KE is a valuable method for conceptual product development. The KE is widely used in product ideas, Roy, et al. [4] explored a method of applying Kansei as part of a product design creative process. Yong-Feng, et al. [5] studied on the product design method based on KE was adopted. the results were analyzed by using mathematical methods, and the relationship that guided the product design between Kansei words and design properties was established. Shen, et al. [6] and Xu, et al. [7] deduced Design mode of product form on combining KE with Ergonomics. KE is used as the main method to transform consumers’ feelings and images of shape, size, material, operability of product into design form futures. Xiao, et al. [8] established an universality evaluation standard system of product design on basis of KE and virtual reality. Chen, et al. [9] aimed to explore the application of KE in design of life products.

The KE and BP combined can be useful in the product design creative process. Yan, et al. [10] studied the relationship between the design elements of the head form of home service robots and the perceptual evaluation of users,analyzed quantitatively by KE and BP neural network. Su, et al. [11] established the relationship between the product form design elements and the consumers’ emotional image perception by using BP three-layer Neural Network and improved BP four-layer Neural Network. Akita, et al. [12] propose a new evaluation method and how to visualize its result in the field of furniture development for indoor public spaces through the application of a quantitative method. Kondo, et al. [13] studied one of the important Kansei values of automobiles is engine sound. in Kou, et al. [14] studied the visual impression observed in the front grills of Sport Utility Vehicles have been investigated by using many adjectives and compared the effect of design types. The Semantic Differential (SD) method was used to quantify the consumers’ visual impression for our experiment.

For the field of product innovation and design, traditional design methods and processes are not scientific to a certain extent, which leads to the products designed by designers not meeting the requirements of users, and it is difficult to quickly and accurately target the market. The application of BP neural network in the development and design of new products can solve these problems. Lin, et al. [15] presented a new fuzzy logic approach to determine the best combination of mobile phone form elements for matching a given product image. Bo, et al. [16] and Zhi-Gang et al. [17] realized the automation of product color matching design. It takes the Kansei engineer methods as the theoretical foundation, uses the neural network as the key technology to study product color-matched.

Traditionally, data for product development is usually obtained through surveys. Cheng, et al. [18] analyzed the research status of current instrument design, and summarize the types of instruments which look beautiful and can bring users good using experience. Ge, et al. [19] deduced a brand-new market model that urges enterprises and designers to make KE as an important indicator of future development. Guo, et al. [20] obtained a new method on the effective acquisition of Kansei words for assessing product design features that plays a decisive role in KE research, whereas previous studies rarely give a full understanding of how to effectively grasp the related Kansei words. Wang et al. [21] built a knowledge transfer model that capitalizes on hypercycle theory and proposes the concept of modularizing information carriers and information processing methods. Su, et al. [22] studied the Kansei theory based on Convolutional Neural Networks, and their study was provided to validate the proposed architecture. the results proved the feasibility of using computer vision to mimic human vision for the automatic evaluation of Kansei attributes. Chen, et al. [23] studied the style design methods of professional female vests that meet the emotional needs of consumers, using the theory of KE as a guide to screen representative samples of female professional vests and relevant emotional vocabularies of styles, through morphological analysis, style design elements of female professional vests are extracted, the fifth-order semantic difference questionnaire was used to establish the perceptual assessment matrix for design elements. Wei [24] proposed an innovative method based on the semantic differential analysis of KE and the correlation analysis of machine learning in the case of small-size samples, which could achieve the evaluation of users’ subjective feelings to form factors of modeling language and color schemes in the process of product design.Chen, et al. [25] considers KE concept is widely used in all kinds of product design. Designers can use KE system to develop new product forms with regional cultural characteristics, combine consumers’ perceptual needs with design strategies of cultural and creative industries, accurately grasp consumers’ preferences, and determine the relationship between design themes and design elements. Xiao-Yun, et al. [26] aims to explore with the design method based on KE, to enable the design of children intellectual products to better meet the needs of children and their parents, regarding the problem of low participation of children users in the design and research of children intellectual products.

This paper mainly aims at the application of intelligent data analysis algorithm of KE theory in new product development, the intelligent data analysis algorithm refers to the BP neural network mapping model based on the relationship between the evaluation of product Kansei intention and the shape of product modeling, the purpose of this study is to establish a predictive model between product modeling image design and product modeling quantification, and to link product design elements with consumer demand and perception.

There is a case in this paper, it takes an electric welding machine as an example to combine the BP neural network and KE theory into the process of product development, and the Matlab is used as the platform, the intelligent measurement model is established. In the research, the various index elements of product design are coded, and these codes are used as the input elements of the neural network. By establishing the relationship between the input elements and the output elements, a neural network model based on KE analysis is established and used this model makes predictions. In the new product development stage, this model is more suitable for market demand, it can guide product design and development, and provide a more convenient and convenient way for new product development and design. Faster and more in line with market demand products.

2 BP neural network

Typically, Artificial Neural Networks are computed and modeled in a way that mimics the biological nervous system. It is a system formed by a large number of dots. In the entire associated system, its basic Elements can be simply understood as a series of nodes, and the lines connecting nodes to nodes. Each node can be used as a neural unit to perform neural operations. The connection between nodes can be used for transmission and can be weighted for connection signals. In recent years, the use of neural network algorithms to assist product design and development is well researched. In the latest research, Chen, et al. [27] aims to construct a predictive model between product modeling imagery design and product modeling quantification to associate product design elements with consumer needs and perceptions. Feng, et al. [28] studied to reduce the dependence of product color matching on professional knowledge effectively and improve the efficiency of product color matching design, a complex color matching method based on BP neural network was proposed. Min-Na, et al. [29] is provided which includes principal component analysis, clustering methodology and BP neural network technique, based on the research of domestic and foreign product evaluation systems in recent years.

2.1 BP neural network modeling

The BP neural network is one of the practical application modes of artificial network models, and is used in various fields as the main form of neural network, such as identifying information, analyzing data, and predicting in other aspects, the BP neural network has the advantages of simple structure, stable working state, and easy hardware implementation.

Figure 1 shows a typical BP neural network topology. The BP neural network model can be composed of three or more layers of neural networks. The structure can be divided into input, implicit and output parts. The input neuron is used as a node to receive external signals. After the external signal or data value is affected by the hidden layer through the network, it is output through the output neuron, and then the output value is compared with the expected value, and the weight is corrected by the accuracy of the comparison value. The training is repeated until the minimum error function is reached, and the accuracy reaches the requirement.

Fig. 1

Typical BP neural network topology.

In this way, a good neural network model can be used to predict the design plan or the specific data process can be represented by the neural network operation flowchart in Fig. 2. In Fig. 2, MSE is the performance function of the network. The Mean Square Error of the network is called “Mean Square Error”.

Fig. 2

Neural network operation flowchart.

2.2 Derivation of the BP neural network algorithm

As Fig. 3, Typical three-layer BP neural network structure, $\bar{α} = (x_{1}, x_{2})$ As input sample, neu₁, neu₂, neu₃, neu₄, neu₅ and neu₆. are neurons, w, b, $\hat{y}$ respectively indicate weight, bias, and output function value.

Fig. 3

A simple three-layer neural network.

The calculation Equation of the K-layer neural unit can be expressed as: the input of K is equal to the output of the previous layer multiplied by the weight of K, plus the bias of K. Next, we use Fig. 3 as a sample to demonstrate the algorithm derivation of the BP neural network by bringing in actual data. It mainly includes forward propagation calculation and back propagation calculation. In the process of algorithm derivation, define the Logistic function. $f_{k} (x) = \frac{1}{1 + e^{x}}$ (1)

The Equation (1) is the activation function. $L (y, \hat{y}) = E {(y, y)}^{2} .$ (2)

Simplified to: $L (y, \hat{y}) = \frac{1}{2} Σ {(y, \hat{y})}^{2}$ (3)

The Equation (3) is the loss function.

3 Innovative design of BP neural network products based on KE

From ove, it can be understood the principle and techniques of BP neural network can be combined to build an intelligent data analysis algorithm. Product development and design innovation research can take advanta of this algorithm. Through the combination of the two, a new scheme or innovative design concept is proposed. Through user’s feedback to the market, quantifying the perceived intention of the product, obtaining the corresponding data, inputting the calculation of BP neural network, establishing the calculation model of intelligent data algorithm. This model can connect the perceptual feature of product with product design and predict the product appearance feature, so that new product development can predict the product which is more in line with users’ psychological expectation in the process of product innovation design.

3.1 The establishment of perceptual semantic space

Take the electric welding machine as an example to explain the application of Kansei theory in detail, establish the morphological semantic space of the electric welding machine, and pave the way for the design of the Aotai electric welding machine (see Fig. 4). Through questionnaires or online collection, a large number of semantic vocabulary related to modeling can be obtained. In order to establish an effective semantic space, it should not be established with any pertinence or bias.

Fig. 4

Decomposition of welder morphological elements.

Through collection and sorting, there are 150 commendatory words and 15 derogatory words that can describe the semantic vocabulary of the product shape. Because derogatory words do not have a guiding role in product design and development, the derogatory words are removed. Then use the KE method to analyze and obtain 10 vocabulary related to the shape of the electric welding machine. They are: high-tech, economical and practical, unique, fashionable, masculine, simple, harmonious, powerful, quality, and reliable. See Table 1 for details.

Table 1

Semantic space establishment

Main vocabulary	Include content
High tech	Advanced, high-speed, advanced mechanical, conceptual, difficult to operate, metallic, electric
Economical	Reliable, trustworthy, realistic, popular, practical, easy to use, safe, durable, and labor-saving
Unique	Individual, unique, avant-garde, surreal, novel, exaggerated
Stylish	Open, lively, interesting, vivid, novel, innovative, high-end
Masculine	Square, sharp, straight, rough, strong, powerful, bold, hard
Concise	Compact, bright, restrained, complete, linear
Harmonious	Integral, environmentally friendly, atmospheric, stable, harmonious, harmonious, generous, satisfactory, and coordinated
powerful	Thick, stable, angular, tension
Quality	Expensive, exquisite, delicate, meticulous
Reliable	Integrated, environmentally friendly, atmospheric, reliable, harmonious and coordinated

3.2 Establishment of morphological space

The establishment of the product form space is done by first selecting a large number of form samples through investigation and data collection, then classifying the samples with similar shapes into one category, and finally carrying out the decomposition of the form elements of the product. Through the collection and analysis of the data of the 8 famous brands of welding machines at home and abroad, we can build the background knowledge related to electric welding machines and understand its modeling characteristics. As shown in Table 2.

Table 2
Comparative analysis of brand product forms (Foreign brands)

Brand Analysis

The overall shape, a strong sense of professionalism. Emphasize the industrial sense of the product and the overall sense of family. But the shape is simple and breakthrough is not enough.

The form is simple and modern, integrated, with a strong sense of professionalism, focusing on serialization. The overall shape is relatively traditional, and there is little change between the series.

The form is modern and smooth, with a strong sense of consumption. It breaks through the shape of the traditional electric welding machine, with more detailed considerations, and the arc also increases product affinity.

The shape is simple, the industrialization feels strong, the shape is mainly rectangular, and the ergonomics is not considered enough.

High-end image, modern and smooth form, strong sense of innovation and consumption. Trying different shapes to increase the aesthetics of electrical equipment. Modern sense, strong sense of technology.

Brand	Analysis

The overall shape, a strong sense of professionalism. Emphasize the industrial sense of the product and the overall sense of family. But the shape is simple and breakthrough is not enough.

The form is simple and modern, integrated, with a strong sense of professionalism, focusing on serialization. The overall shape is relatively traditional, and there is little change between the series.

The form is modern and smooth, with a strong sense of consumption. It breaks through the shape of the traditional electric welding machine, with more detailed considerations, and the arc also increases product affinity.

The shape is simple, the industrialization feels strong, the shape is mainly rectangular, and the ergonomics is not considered enough.

High-end image, modern and smooth form, strong sense of innovation and consumption. Trying different shapes to increase the aesthetics of electrical equipment. Modern sense, strong sense of technology.

In the long course of development of welding machines, different brands have formed a series of products. The simple and generous appearance is designed by superimposing, wrapping, cutting and combining simple geometric shapes. Many brands also have established a certain style. In recent years, products with better aesthetics and more suitable for working environment are being developed. Table 3 is a comparative analysis of Chinese welding machine shapes.

Table 3

Comparison and analysis of welding machines (Domestic brands)

Brand	Analysis

Strong professional sense of form. Diversified handle handling. The overall shape is more traditional. There are different attempts in color.

There are different attempts in the shape. The old series of welding machines have traditional overall shape, and the new welding machines have a smooth and high-end design.

The shape is simple, the traditional welding machine is highly industrialized, and the shape is mainly rectangular. The change of the handle and the use of different materials increase the sense of design.

Comparison of Chinese and other electric welding machine products: Compared with Chinese welding machine products, foreign electric welding machine products were developed earlier, with more mature product lines and stable performance, emphasizing family-oriented products, industry sense and professionalism. It is not difficult to see from the above Table 3 that foreign products are overall more superior than domestic products. The reasons for this phenomenon are: First of all, the development time of foreign industries is earlier than that of domestic industries, and the overall industrial development level is higher than that of domestic industries. Secondly, foreign material processing technology is better than domestic, and advanced material processing technology can better promote the development of electrical equipment. Third, the productivity of foreign countries is more advanced than that of China, which provides a material basis for the development of electrical equipment, and have higher capital and technology investment. Finally, foreign countries pay more attention to the design of product appearance and invest more in design, while domestically, enterprises invest less in design.

3.3 Analysis of the development trend of electric welding machine

Due to cultural differences at home and abroad, users have different aesthetics, thus, there are also differences in the appearance of welding machine products. For example, the appearance of electrical products in Europe and the United States pays more attention to professionalism and functionality; while the development of electrical products in my country is short of time and learns from different foreign design styles, so there are various styles in the shape of electrical equipment, and common characteristics.

On the basis of the above research data, we distinguished the different shapes of electric welding machines. A total of 30 electric welding machine modeling samples were collected, and then these product samples were subdivided, divided into different design elements, and summarized the welding of different design styles. The samples can be divided into 5 categories through analysis. After comparing the morphological elements of the electric welding machine, it is found that the main points of the electric welding machine product design are (see Table 4): control panel design, ventilation plate design, handle design, side heat dissipation hole design, and overall body shape design. The influence of the detail design of each part on the overall design cannot be underestimated. The good detail design will make the whole design reflect the sense of exquisiteness and high-end, and it also shows the designer’s intention to design.

Table 4
Shape analysis of welding machine

Part analysis

Inclined concave control panel: Compared with the straight concave control panel, it is more ergonomic, convenient to operate and easier to observe.

Ventilation plate: From the perspective of the entire front face, it occupies a large part of the area. The existing welding machine ventilating plates are designed in various ways, all in order to improve the decoration and have the characteristics of design assistance to the overall design.

Side radiating holes: Because the shape of the radiating holes on the sheet metal shell is subject to many restrictions, they are all stamped. The design of the heat dissipation holes on the side of the large area can play a finishing touch.

Handle: The design of the handle should not only run through the whole, but also consider the ergonomics and texture of the hand. The design is not only good-looking but also convenient. It can improve the grade, quality and professional sense of the product.

Part analysis

Inclined concave control panel: Compared with the straight concave control panel, it is more ergonomic, convenient to operate and easier to observe.

Ventilation plate: From the perspective of the entire front face, it occupies a large part of the area. The existing welding machine ventilating plates are designed in various ways, all in order to improve the decoration and have the characteristics of design assistance to the overall design.

Side radiating holes: Because the shape of the radiating holes on the sheet metal shell is subject to many restrictions, they are all stamped. The design of the heat dissipation holes on the side of the large area can play a finishing touch.

Handle: The design of the handle should not only run through the whole, but also consider the ergonomics and texture of the hand. The design is not only good-looking but also convenient. It can improve the grade, quality and professional sense of the product.

After 5 types of product samples can be obtained through the KJ method, these 5 representative samples are abstracted and decomposed. Each design element can be quantitatively analyzed for future design research and design process. The analysis diagram is shown in Table 5.

Table 5

Sample abstract shapes

Type	Sample 1	Sample 2	Sample 3	Sample 4	Sample 5

Overall shape
Control panel
Ventilation panel
Side cooling holes
Handle

3.4 The relationship between semantic space and morphological space

After roughly classifying the welding machine samples, it is necessary to clarify how users feel about the samples differently. Then establish the connection between product semantic space and form space. Through a questionnaire survey of 50 different working environments and welders, 10 semantic words and 5 types of welding machine model samples will be scored according to their own feelings. Preliminarily get the score value of each sample about perceptual image words. The score level is divided into 1 to 5 according to the severity of the association. 1 is severely unrelated, and 5 is severely related. Then calculate the average score of each item and express it in the form of a line graph, which can clearly summarize the final conclusion.

The Fig. 5 horizontal coordinates represent the 10 semantic words of the welding machine model semantic space, namely high-tech, economical and practical, unique, fashionable, masculine, simple, harmonious, powerful, quality, reliable, the ordinate indicates the semantic word score of the sample of welder model.

Fig. 5

Sample score.

From the Fig. 5 it can be concluded that sample 1 has the highest score among the harmonious, simple, economical and practical emotional words; sample 2 has the most sense of quality; sample 3 is reliable; and has the highest score for technological, masculine, and power Is sample 4; sample 5 is unique and fashionable. In order to refine the impact of the design elements of the welding machine on the user’s perceptual image, 50 welders and 10 industrial designers respectively scored the 5 design elements of the welding machine sample, and the score level is divided into 1∼5, 1 is severely unrelated, 5 is severely related, and then calculate the average score to reach a conclusion. Each perceptual vocabulary is analyzed by this method.

Sample 1 scored the highest among the harmonious, simple, economical and practical emotional words. From the analysis in Fig. 6, it can be seen that in sample 1, the most harmonious is the ventilation panel. The material properties of the front panel are sheet metal parts. The ventilation panel is formed by punching openings of the sheet metal parts; the simplest design element is the overall shape, The analysis shows that the entire welder is a shell composed of sheet metal parts and fixed by screws and nuts, and the shape has not changed much. The most economical and practical is the design of the side panel.

Fig. 6

Sample 1 design element score.

From the above the Fig. 7, Sample 2 scored the highest among the perceptual words of quality. The analysis shows that the designer of the control panel has the most sense of quality. Comparing with other samples, it can be concluded that the design of sample 2 ventilation plate is meticulous.

Fig. 7

Sample 2 design element score.

From the above Fig. 8, it can be concluded that the overall shape of sample 3 has the highest score among reliable perceptual words.

Fig. 8

Sample 3 design element score.

Through the analysis of the above Fig. 9, it can be concluded that sample 4 has the most technological sense of the control panel, and its control panel is an irregular pentagonal design. Compared with the quadrilateral design, the pentagonal design has also affected the change of the ventilation panel and the overall shape, and the whole is more dynamic. The most masculine is the overall look. The most powerful sense is the handle.

Fig. 9

Sample 4 design element score.

Through the analysis of the above Fig. 10, The overall shape of sample 5 is an important factor that affects users’ evaluation of product fashion and unique sensibility, and can be used as a guide for subsequent product development. From the data comparison, it can be concluded that the harmonious and simple sense of sample one, through the analysis of the design elements of the series one welding machine sample, the overall shape is a square block, the outer profile of the front panel is a small round and chamfered rectangle, a ventilation plate and The control panel is on the sheet metal part on the same vertical surface, and the opening is more square, without much change.

Fig. 10

Sample 5 design element score.

3.5 Establishment of semantic space and morphological space of Aotai welding machine

Welding machine sample acquisition: through the welding machine pictures collected from the welding machine user’s investigation and analysis and sales records, four samples of Aotai welding machine are determined, as shown in the Fig. 11, the morphological elements are decomposed, as shown in Table 6. All modeling forms are decomposed and classified according to different components. The modeling of these components has an impact on the final product form. The shape of this device has very obvious characteristics, and through the analysis of the characteristics of different structures, coding and data can be obtained.

Fig. 11

The product modeling analysis.

Table 6

Aotai welding machine sample

Type	S 1	S 2	S 3	S 4

Overall shape
Control panel
Ventilation panel
Side cooling holes
Handle

Table 7

Adjective pairs

Adjective pair	Evaluation factor
High-tech-traditional	Sense of technology
Economical-expensive	Economical and practical
Unique-popular	Uniqueness
Stylish-ordinary	Fashion sense
Masculine-feminine	Masculine
Simple-complex	Simplicity
Harmonious-unbalanced	Sense of coordination
Powerful-weak	Sense of power
Quality-rough	Sense of quality
Reliable-unreliable	Reliability

Through the comparative analysis of the above, it is easy to see: ding172 The shapes of different models of Aotai electrical equipment are similar or can be said to be the same, and there is no obvious distinction between different models, different functions and use environments. ding173 The shape is simple, the shape is traditional, and the industrialization feels strong. ding174 The color of the product is mainly blue and gray, the shell is mainly sheet metal, and the panels of some models use plastic parts. The product is not considered enough in details. ding175 The operation area is not clearly distinguished, and the ergonomics is not considered enough.

4 Semantic space research

4.1 Perceptual intent vocabulary analysis

Perform user semantic analysis based on the above sample of Aotai welding machine. Under the background and cultural environment of Aotai company, combined with the research and analysis of Aotai electric welding machine users and the analysis of domestic and foreign welding machines, the emotional vocabulary semantics obtained by using SD method was used by 50 welders and 10 R&D personnel Aotai welding machine samples are evaluated at level 7 of the perceptual image vocabulary, As the Tables 8and 9, shown, which can obtain the intuitive expectations of the users of the Aotai welding machine for the welding machine, and then guide the development of new products of the Aotai welding machine based on the results.

Table 8
Seven levels of scale

Adjective Quantification Adjective

Progressive 3 2 1 0 –1 –2 –3 Conservative

Light 3 2 1 0 –1 –2 –3 Heavy

Soft 3 2 1 0 –1 –2 –3 Stiff

Novel 3 2 1 0 –1 –2 –3 ordinary

emotional 3 2 1 0 –1 –2 –3 rational

Ordered 3 2 1 0 –1 –2 –3 messy

Open 3 2 1 0 –1 –2 –3 closed

Coordinated 3 2 1 0 –1 –2 –3 Unbalanced

bright 3 2 1 0 –1 –2 –3 dark

Relaxed 3 2 1 0 –1 –2 –3 nervous

Adjective	Quantification	Adjective
Progressive	3	2	1	–1	–2	–3	Conservative
Light	3	2	1	–1	–2	–3	Heavy
Soft	3	2	1	–1	–2	–3	Stiff
Novel	3	2	1	–1	–2	–3	ordinary
emotional	3	2	1	–1	–2	–3	rational
Ordered	3	2	1	–1	–2	–3	messy
Open	3	2	1	–1	–2	–3	closed
Coordinated	3	2	1	–1	–2	–3	Unbalanced
bright	3	2	1	–1	–2	–3	dark
Relaxed	3	2	1	–1	–2	–3	nervous

Table 9

Results of the calculation

Component score coefficient matrix
	Component
	1	2	3	4
Sense of technology	0.184	0.646	0.012	0.116
Economical and practical	0.546	0.086	0.088	0.205
Uniqueness	0.456	0.055	0.068	0.107
Fashion sense	0.111	0.082	0.005	0.610
Masculine	0.017	0.178	0.223	0.306
Simplicity	0.219	0.513	0.057	0.038
Sense of coordination	0.036	0.057	0.417	0.007
Sense of power	0.038	0.039	0.057	0.513
Sense of quality	0.212	0.004	0.706	0.012
Reliability	0.205	0.056	0.477	0.040
Extraction method: main component. Rotation method: Orthogonal rotation method with Kaiser standardization. Make up a score.

The SD method is used to divide adjectives into 7 levels to obtain more accurate feelings and images of the subjects (Table 8).

Excel is used to analyze the average of perceptual image values, and get the results as shown in Fig. 12.

It can be intuitively seen that the average difference is relatively large, and the key factors cannot be accurately focused, so SPSS is required for factor analysis.

Fig. 12

Average score.

Using SPSS software to perform factor analysis (Table 9), Readers can clearly know the perceptual vocabulary corresponding to the sample and obtain more accurate results.

Through the analysis of the user’s image result, it can be concluded that the user’s expectation of the welding machine is masculine and powerful.

4.2 Determination of the design plan

According to the above survey and data analysis, Aotai’s existing welding machine lacks a sense of strength and its form is perceived as mediocre. According to the user’s perceptual expectations for the Aotai welding machine, the design of the overall shape, control panel, ventilation plate, and side cooling holes are refined, Combined with material analysis, process analysis, and man-machine analysis for sketching, the sketch is shown in Fig. 13.

Fig. 13

Design sketch.

Based on the above design sketches, Rhino software is used for modeling, and partial details are designed, taking into account the rationality of the structure and the stability of the product. Figure 14 shows the final welding machine product rendering result.

Fig. 14

Three-dimensional model.

4.3 BP neural network model establishment and testing

Based on the analysis of all welding machine brands, establish a BP neural network model between the user’s perceptual image of the welding machine in a large environment and the product design elements. After the establishment, the design samples of the Aotai welding machine are imported into the model for training. Analyze whether the final result meets market expectations to guide product design optimization and development.

The number of nodes is determined: the design elements of the electric welding machine product are used as the input of the input layer parameters, and the analysis shows that the number of design elements is 5; the perceptual image evaluation value sorted out above is used as the output of the output layer parameters, and the perceptual vocabulary is obtained by statistics The number is 10. The number of hidden layer nodes generally uses Equation (1). $m = \frac{n + 1}{2}$ (4)

In Equation (1), m, n, and l are respectively expressed as the number of hidden layer nodes, the number of input layer nodes, and the number of output layer nodes.

Therefore, the number of hidden layer nodes in the BP network model of the electric welding machine is 8.

Because the design elements of the electric welding machine cannot be directly input as parameters, it is necessary to encode the design elements of each electric welding machine. Take the product design elements as the standard, and number the samples. The design elements of the first sample are numbered 11, 12, 13, 14, 15 for the overall shape, control panel, ventilation plate, side vents, and handles; the second one in turn It is: 21, 22, 23, 24, 25; the third is: 31, 32, 33, 34, 35; the fourth is: 41, 42, 43, 44, 45; the fifth is: 51, 52, 53, 54, 55. After unified coding, it is the parameter of the input layer.

The coding method of design elements of electric welding machine sample is:

Clear

clc

A = [11,12,13,14,15];

B = [21,22,23,24,25];

C = [31,32,33,34,35>];

D = [41,42,43,44,45];

E = [51,52,53,54,55];

for j = 1:5

for k = 1:5

for m = 1:5

for n = 1:5

a = A(5);

b = B(j);

c = C(k);

d = D(m);

e = E(n);

[a,b,c,d,e]

end

According to the requirements of the training function, the output parameters need to be in the interval [0,1], and the obtained perceptual evaluation value is not completely within this interval, so it is necessary to normalize the perceptual evaluation to [0,1]] The range of the interval. Based on the above reasons, this study uses the normalization algorithm modeling, which is simple and convenient. Among them, X is the minimum and maximum. After processing the perceptual data in the interval [0,1], it can be used as the output layer parameter of the welding machine BP neural network for network training. $X_{i} = \frac{x_{i} x_{min}}{x_{max} x_{min}}$ (5)

After the design elements of the sample are uniformly coded, the data used as the input layer is input into the BP neural network. The data is shown in Table 10.

Table 10

Input data of perceptual intention evaluation value

	Sample 1	Sample 2	Sample 3	Sample 4	Sample 5
Overall shape	1 1	1 2	1 3	1 4	1 5
Control panel	2 1	2 2	2 3	2 4	2 5
Ventilation pole	3 1	3 2	3 3	3 4	3 5
Side cooling holes	4 1	4 2	4 3	4 4	4 5
Handle	5 1	5 2	5 3	5 4	5 5
Coding	1121314151	1222324252	1323334353	1424344454	1525354555

The user’s perceptual intention evaluation value of the sample and perceptual vocabulary is used as the output layer data of the BP neural network. The data is as follows:

clear;
clc;
t = [1 1 2 1 3 1 4 1 5 1]
1 2 2 2 3 2 4 2 5 2
1 3 2 3 3 3 4 3 5 3
1 4 2 4 3 4 4 4 5 4
1 5 2 5 3 5 4 5 5 5
p = [4.28 3.72 4.12 4.86 4.35]
2.72 4.24 2.12 3.52 3.68
3.36 1.34 3.55 1.56 1.56
2.56 4.29 1.98 2.08 1.76
3.28 4.34 4.56 4.04 4.24
3.08 4.06 2.52 2.28 3.02
4.44 4.28 4.52 3.22 1.78
3.04 4.54 3.35 4.24 3.26
2.92 4.26 1.64 1.28 3.46
2.76 3.96 4.26 3.42 3.98]’
P_test = [1 1 2 1 3 1 4 1 5 1]
1 2 2 2 3 2 4 2 5 2
1 3 2 3 3 3 4 3 5 3
1 4 2 4 3 4 4 4 5 4
1 5 2 5 3 5 4 5 5 5
[pn,minp,maxp,tn,mint,maxt] = premnmx(p,t);
p2 = tramnmx(P_test,minp,maxp);
net = newff(minmax(pn),minmax(tn),3,’tansig’,’tansig’,’purelin’,’
traingdm’);
net.trainparam.show = 50;
net.trainparam.mc = 0.9;
net.trainparam.lr = 0.01;
net.trainparam.epochs = 200;
net.trainparam.goal = 1e-5;
net = init(net);
[net,tr] = train(net,pn,tn);
PN = sim(net,p2);

The user’s perceptual intention evaluation value of the sample and perceptual vocabulary is used as the output layer data of the BP neural network. Newff is used to create the BP neural network, the Logsig S-type logarithmic function is used as the hidden layer activation function, the purelin linear function is used as the output layer function, and the trainlm is used for training. The net training is set number of learning times to 200, the error is 0.001, and use the traignd method to train this network until the BP neural network model of the welding machine is initially obtained. The training reached the purpose of training 142 times and stopped training, and the construction of the BP neural network model of the electric welding machine was initially completed. As shown in (Fig. 15).

Fig. 15

Convergence curve of BP neural network.

4.4 Verification of design results of Aotai welding machine

Combine the 5 design elements of the 5 representative samples of the electric welding machine, then there are a total of 5×5×5×5×5 = 3125 types. All the combinations are encoded as input layer parameters and imported into the BP neural network model. According to the design results of the Aotai welding machine, the overall shape, control panel, ventilation plate, side vents, and handles are coded corresponding to the previous sample design elements, the code is: 1424334455, and the code is imported into the BP neural network, Readers can see Whether the sense of power scores the highest in the perceptual image value of.

According to the results, the detailed design is carried out to obtain the final design plan to meet the needs of users and meet market expectations. (see Fig. 16) In this way, the design elements of the designed product scheme can be used as the input layer to input the data into the trained BP neural network, and the results according to the perceptual evaluation value of the output layer can be compared with the perceptual value of the user. Market forecasts can be made to optimize the process of new product development.

Fig. 16

Final product.

5 Conclusion

The mapping model of BP neural network based on induction engineering is proved to be feasible by the example of electric welding machine, and the application of BP neural network based on KE in practical product innovation method is illustrated by the case of electric welding machine. Under the guidance of KE theory, Matlab neural network tool is used to vectorize and extract the user’s perception meaning. On the basis of this platform, BP neural network model is established for the development and research of electric welding machine products. By establishing the relationship between Kansei analysis and BP neural network model, the Kansei intention values of products and samples are established, and a reasonable relationship model is established, get the design result which is more suitable to the market demand.

This research is mainly for the development of specialized products with clearly defined users. The fuzzy user perception is improved into the design elements of the product, which strengthens the connection between the designer and the usersand provides help for the design. At the same time, the process of traditional new product development was further improved, which brought methodological guidance and effective materials to the design of electric welding machines.

In this paper, the feasibility of KE applying BP neural network product innovation method is demonstrated. In the field of product innovation design, the problem that it is difficult for some products to meet the market demand and other defects quickly and accurately in the traditional design method and process has been solved, and the traditional design method and process have been improved, in the practical application, according to the specific application adjustment model, make the design plan that more conforms to the market demand, guide the new product development and the product innovation method development, therefore, we can apply the intelligent data analysis algorithm based on KE theory to new product development, which can provide a new research idea and design method.

Footnotes

Acknowledgments

A lot of help was provided by Jiacheng Zhang, Na Liu and Meng Yang, thanks for their hard work. The project was funded by Shandong Education Department Graduate Quality Improvement Program (SDYAL19112) and Shandong Province’s key support regions introduce urgently needed talent projects.

References

Wang

and Zhou

, Integrating rough set theory with customer satisfaction to construct a novel approach for mining product design rules, Journal of Intelligent and Fuzzy Systems (5) (2021), 1–25.

Chen

, Yu

, Chu

, Yu

and Chen

, Fuzzy emotional evaluation of color matching for aircraft cockpit design, Journal of Intelligent and Fuzzy Systems (3) (2021), 1–19.

and Kang

, Establishment of an aggregation model associated with instrument interface design based on kansei factors of electric vehicle drivers, Mathematical Problems in Engineering 2020(11) (2020), 1–15.

Roy

, Goatman

and Khangura

, User-centric design and KE, CIRP Journal of Manufacturing Science and Technology 1(3) (2009), 172–178.

Yong-Feng

L.I.

, Zhu

L.P.

Research on product design method based on KE, Packaging Engineering. (2008).

Shen

H.C.

and Wang

K.C.

, Affective product form design using fuzzy KE and creativity, Journal of Ambient Intelligence and Humanized Computing 7(6) (2016), 875–888.

Y.Q.

, Chen

, Qin

H.B.

and Wang

Z.Y.

, Study on the application of KE in product form design, Applied Mechanics & Materials 274 (2013), 513–516.

Xiao

, Cheng

, Wang

, Ye

, Le

Analysis on Universality Evaluation Standard System of Product Design on Basis of KE and Virtual Reality, International Conference on Human-computer Interaction. Springer International Publishing. (2015).

Chen

J.L.

, Zhao

, Yi

L.I.

, Zhang

Q.Y.

Product design method based on KE, Packaging Engineering (2019).

10.

Yan

, Gang

Research on the head form design of service robots based on KE and BP neural network, Seventh International Conference on Electronics and Information Engineering. International Society for Optics and Photonics (2017).

11.

, Zhang

, Peng

, Gang

Research to Compare Emotional Image Evaluation Models of Product Form Based on Neural Network.

12.

Akita

, Morita

and Shiizuka

, Evaluation of product design user satisfaction based on the systematic analysis of kansei, Journal of Immunology 154(4) (2015), 615–626.

13.

Kondo

, Hara

, Torii

, Yoshioka

, Miyazawa

Engine Sound Design Process with Utilization of Industrial Styling Design, WCX SAEWorld Congress Experience (2020).

14.

Kou

, Kawasumi

Kansei analysis of car front grill to design for thai people, Transactions of Japan Society of KE (2019).

15.

Lin

Y.C.

, Lai

H.H.

and Yeh

C.H.

, Consumer-oriented product form design based on fuzzy logic: a case study of mobile phones, International Journal of Industrial Ergonomics 37(6) (2007), 531–543.

16.

, Xi

, Zhou

New product design based target cost control with bp neural network and genetic algorithm- a case study in chinese automobile industry. (2009).

17.

Zhi-Gang

H.U.

, Na

, Qiao

X.L.

, Wang

Application of bp neural network in product color-matched design, Packaging Engineering (2016).

18.

Cheng

, Qian

, Ye

, Yang

, Shen

Research on chemical experimental instrument design mode based on KE. (2019).

19.

, Cheng

, Mao

Application of KE in product design: a case study of children’s dining chairs, Industrial Design (2019).

20.

Guo

, Qu

Q.X.

, Nagamachi

and Duffy

V.G.

, A proposal of the event-related potential method to effectively identify kansei words for assessing product design features in KE research, International Journal of Industrial Ergonomics 76(March 2020) (2020), 102940.

21.

Wang

, Chen

, Zheng

, Zhang

and Yu

, Knowledge transfer methods for expressing product design information and organization, Journal of Manufacturing Systems 58 (2021), 1–15.

22.

, Yu

, Chu

, Zhai

and Fan

, A novel architecture: using convolutional neural networks for kansei attributes automatic evaluation and labeling, Advanced Engineering Informatics 44 (2020), 101055.

23.

Chen

, Cheng

The style design of professional female vest based on KE, International Journal of Clothing Science and Technology, ahead-of-print(ahead-of-print) (2019).

24.

Wei

, Form factors of modeling language under different color schemes with grey relational analysis based on entropy method,011.1–011, International Journal of Pattern Recognition and Artificial Intelligence 33(8) (1950), 23.

25.

Chen

, Sun

Architecture-oriented Design Decision Analysis of Longchi Inkstone KE System. (2019).

26.

Xiao-Yun

X.U.

, Xie

Q.R.

, Zhang

Design method of children participatory intellectual products based on KE, Packaging Engineering (2019).

27.

Chen

S.B.

, Yang

S.J.

, Zhu

, Mei-Mei

L.I.

, Zhang

W.S.

Design and Quantification of Cars Modeling Imagery Based on BP Neural Network. (2019).

28.

Feng

, Meng-Di

W.U.

, Sui-Huai

Y.U.

, Yang

Research of tank vehicle colour matching and evaluation method based on bp neural network, Journal of Machine Design (2019).

29.

Min-Na

N.I.

, Sun

Z.H.

, Wang

, Shu-Jia

L.I.

, Huang

Applied research on bp neural network modeling for the evaluation of kansei images and elements of product design, Journal of Donghua University (2016).