In the field of economic geography, regional economic structure has become one of the core starting points for research on regional economic resilience. Early studies primarily focused on the differences in how two types of regional economic structures—specialization and diversification—affect resilience, with particular attention given to the impacts of industrial specialization and diversification (Boschma, 2017). However, the essence of industrial structure lies in technological structure; industrial configurations are the result of different combinations of technologies. Thus, the diversification and specialization of technology form the underlying logic of industrial and economic development (Teng et al., 2023; Balland et al., 2015). Scholars generally agree that regions with a high degree of technological specialization tend to exhibit lower levels of resilience when confronted with shocks, whereas regions with more diversified technological structures tend to display stronger resilience (Teng et al., 2023; Guo et al., 2018). This is because, when disruptions occur, a technologically diversified structure can spread the shock across different technological domains, thereby reducing its concentrated impact on any single sector and increasing the likelihood that each technological domain can recover (Balland et al., 2015; Brakman et al., 2015). However, excessive technological diversification may lead to a lack of connectivity between regional technologies, hindering the generation of innovation and the diffusion of positive externalities, which in turn could undermine the long-term development potential of the regional economy (Rocchetta et al., 2022). As a result, the mechanisms through which technological diversification affects regional resilience remain a matter of academic debate.

To address the limitations of the traditional concept of diversity, Frenken et al. (2007) proposed a distinction between related variety and unrelated variety within diversified regional economic structures, thereby enabling more nuanced analyses of how different types of variety influence regional economic resilience. As Frenken’s original work applied these concepts at the industrial level, subsequent scholars have largely followed this approach, employing industrial statistical data to measure related and unrelated variety and examining their respective impacts on regional economic resilience (Tan et al., 2020). More recently, this framework has been extended to the technological domain. Scholars have begun to use patent data to assess related and unrelated technological variety and analyze their effects on regional economic resilience (Tóth et al., 2022; Rocchetta et al., 2022; Chen et al., 2024). Related technological variety refers to a regional portfolio of technologies that are significantly interrelated or complementary. It emphasizes the spatial co-location of related but distinct technologies, and its mechanism operates through Jacobs’s externalities to enhance the recoverability of the regional economy in the face of shocks. Specifically, related variety reduces cognitive distance between technologies, facilitates knowledge exchange and spillovers, and promotes innovation efficiency, thereby increasing the likelihood of new industries and technologies emerging in the region, which in turn strengthens regional economic recovery (Tóth et al., 2022; Chen et al., 2025). Additionally, when a region encounters external disruptions, technologically related industries can leverage their similarities to reorganize resources and pursue mergers and restructuring, fostering industrial upgrading and facilitating rapid economic recovery (Chen, 2022). Unrelated technological variety emphasizes the low degree of interconnection among a region’s technological resources—that is, the coexistence of multiple, distinct, and unrelated technological domains. Its underlying mechanism operates through portfolio effects, which enhance a region’s economic resistance to external shocks. Specifically, when facing asymmetric shocks, a region characterized by unrelated technological variety can effectively disperse the disruptive impact across different technological sectors, thereby preventing the further spread of risk. At its core, this mechanism strengthens the region’s capacity to withstand external disturbances (Tóth et al., 2022; Boschma, 2017). Based on the above, this study proposes the following hypotheses:

Although related and unrelated variety can to some extent capture the structural characteristics of regional economies, the criteria used to determine whether technologies are “related” or “unrelated” are often based on international industrial classification systems or other technical taxonomies, rather than actual linkages among industries or technologies. In response, recent studies have begun to explore regional economic resilience based on empirically grounded representations of regional economic structures (Tóth et al., 2022; Rocchetta et al., 2022). At present, the network perspective has become a key analytical approach in economic geography for examining the real structure of regional economies. In fact, as early as 2007, Martin and Sunley emphasized the need to further develop a network-based perspective within the field of economic geography. However, due to limitations in data availability and methodological tools, the network approach did not initially gain widespread traction in the discipline.

With the maturation of network science—particularly methodologies derived from the product space—and the rise of big data, scholars have increasingly adopted a network perspective to study regional economic development. Concepts such as regional technological networks (Tóth et al., 2022), regional knowledge networks (Ye et al., 2024), and regional innovation networks (Cao et al., 2023) have been proposed to capture the structural dynamics of regional economies. Among them, regional technological and knowledge networks refer to networks composed of different technological and knowledge domains involved in regional innovation activities. These two are often used interchangeably in empirical research, as both rely on patent and publication data as proxies (Ye et al., 2024). In contrast, regional innovation networks focus on the network of various stakeholders engaged in regional innovation. Drawing on established literature (Tóth et al., 2022; Ye et al., 2024; Cao et al., 2023), this study adopts the concept of the regional technological network to represent the structure of the regional economy, based on the premise that the recombination of different technological domains is a direct driver of regional innovation and economic development. In a regional technological network, nodes represent technological domains, and edges reflect the connections and combinations between them. These diverse combinations enable various forms of economic innovation. While existing research has primarily examined how the structure of regional technological networks influences the emergence of new technologies and regional innovation performance, their impact on regional economic resilience remains underexplored—and their influence on industrial resilience has received even less scholarly attention (Du et al., 2022).

From a network-based perspective, regional technological networks exhibit varying levels of robustness depending on their structural configurations. These differences in robustness can, in turn, lead to divergent levels of economic resilience when regions face external shocks. First, network robustness refers to a network’s ability to continue functioning despite the loss of certain nodes or connections (Albert et al., 2000; Solé et al., 2008). In network science, assessing whether a network can maintain its core functionality typically involves examining the presence of a giant component—a connected subgraph that encompasses a substantial portion of all nodes in the network. Generally, for any given network structure, a giant component will persist and support network functionality as long as node removal remains below a certain threshold. However, once the degree of node failure exceeds this threshold, the network tends to fragment into disconnected subgraphs, ultimately leading to systemic collapse (Zitnik et al., 2019). Moreover, not only the number of nodes removed but also the attributes of the removed nodes critically influence network robustness. Specifically, targeted removal of high-degree nodes (i.e., nodes with numerous connections) tends to degrade network robustness more severely than random node removal (Tóth et al., 2022). For example, in a star-shaped regional technological network—where one central node connects multiple peripheral nodes—removing the central node disrupts the entire network structure, splitting it into isolated components. In contrast, random removal is more likely to affect peripheral nodes, allowing the network to maintain connectivity through the central hub. Second, the robustness of a regional technological network plays a crucial role in shaping the resilience of regional economies through its influence on resistance, recovery, and reconfiguration. On one hand, networks with high robustness demonstrate stronger resistance to shocks, as they are more likely to preserve their giant components when subjected to node loss. This structural stability allows economic activities to continue largely undisturbed during crises. On the other hand, such networks also facilitate recovery and reconfiguration, since their high degree of connectivity provides multiple pathways for technological recombination and innovation. This, in turn, promotes the emergence of new technologies and revitalizes regional economic activity, enabling the exploration of novel development trajectories (Tóth et al., 2022). In summary, we propose the following hypothesis:

Following the analytical framework proposed by Martin et al. (2015), this study conceptualizes industrial resilience along three temporal and process-based dimensions: resistance, recovery, and reconfiguration.

First, reconfiguration captures the processes of reorientation and renewal following external shocks. Drawing on existing literature, this study measures regional resistance using the firm survival rate, which reflects the ability of a regional industrial system to withstand external disruptions without substantial firm exits. Specifically, firms with statuses such as “active” or “in operation” are classified as surviving firms, while those marked as “deregistered,” “revoked,” or “relocated” are considered exiting firms. The survival rate is calculated as follows:

In the formula, and represent the number of surviving and exiting firms, respectively, in region c during year i. While denotes the firm survival rate of region c in year i.

Second, regional recovery is measured by the firm entry rate, which captures the ability of the industrial system to regenerate economic activity following a shock. The entry rate is calculated as follows:

In the formula, represents the number of surviving firms in region c at the end of year i, and represents the number of surviving firms in region c at the end of year i − 1.

Third, this study uses two indicators—industrial diversification index (VAR) and industrial upgrading index (UPG)—to capture the reconfiguration dimension of industrial resilience. The industrial diversification index is calculated using an entropy-based method, as expressed in the following formula:

In the formula, n denotes the number of industry categories, and represents the proportion of firms in a given industry within region c in year i.

Finally, for industrial upgrading index, industrial firms are classified into three categories—labor-intensive, resource-intensive, and technology-intensive. An improved structural similarity coefficient method is employed to measure the degree of industrial structural upgrading. First, the proportion of each type of industrial firm (labor-intensive, resource-intensive, and technology-intensive) in the total number of firms is treated as a component of a spatial vector, forming a three-dimensional vector =(, , ). Second, the angles、、 between and the reference vectors representing ascending levels of industrial sophistication—=(1, 0, 0), =(0, 1, 0), =(0, 0, 1)—are calculated, respectively.

The calculation formula for the Industrial Upgrading Index is as follows:

This study follows the entropy-based approach proposed by relevant scholars to construct indicators for cities’ related and unrelated technological variety (Fratesi et al., 2016; Tóth et al., 2022). According to the entropy decomposition formula by Theil, Unrelated Variety (UV) measures the entropy between groups (i.e., at the one-digit patent classification level), while Related Variety (RV) captures the entropy within groups (i.e., at the three-digit patent classification level). Specifically, the calculation formulas for unrelated and related variety are as follows:

In the formula, represents the proportion of one-digit patent category in city r relative to all patents in that city, while denotes the proportion of three-digit patent iii within the one-digit category in city r.

This study calculates the robustness of technological network in two main steps. First, we construct the city-level technology network or technological space using patent data. The core of this approach lies in assessing the relatedness between pairs of technology nodes. Specifically, we follow the co-occurrence method proposed by Hidalgo et al. (2007) and Zhang & Rigby (2022).The fundamental idea is that if two technologies frequently co-occur within the same patent document, it indicates that they likely share a similar knowledge base. Consequently, a high co-occurrence probability suggests a strong technological relatedness between the two. The formula for computing technological relatedness is as follows:

Where denotes the relatedness between technologies i and j; represents the number of patents that simultaneously contain both technologies i and j; and are the total number of patents containing technologies i and j, respectively.

Based on the calculated relatedness between technologies within a city, a city-level technology network can be constructed. In this network, the nodes represent different categories of patented technologies, and the size of each node reflects the number of patents in that category. The edges between nodes indicate the degree of relatedness between technologies—the higher the relatedness, the shorter the distance between the nodes.

The second step involves assessing the structural robustness of the city’s technology network. The core idea is to measure the proportion of nodes that must be removed to dismantle the network’s giant connected component. This is calculated using the Molloy–Reed criterion. According to this criterion, for a giant component to exist in a network, most of the nodes within it must have at least two neighbors. Specifically, if the K value of a network is greater than 2, a giant connected component exists; if the K value is less than 2, the network lacks a giant component and instead consists of many disconnected components.

Here, k represents the number of edges connected to a given node; 〈k〉 denotes the average k value across the entire network; k² refers to the square of the number of edges connected to a given node; and 〈k²〉 represents the average k² value across all nodes in the network.

Finally, using the Molloy–Reed criterion as the standard, we calculate the proportion of nodes that must be removed for the network’s K value to equal 2. Regarding the node removal methods, there are two types: random removal and targeted removal. Random removal involves randomly sampling and removing nodes from the network, while targeted removal prioritizes removing nodes with higher degrees. Through these two approaches, we ultimately determine the maximum percentage of targeted removal and the maximum percentage of random removal that the city’s technology network can withstand.

This study measures industrial resilience using data from industrial enterprises. Enterprise information was obtained from the Tianyancha website (https://www.tianyancha.com/). The selected industries include “Mining,” “Electricity, Heat, Gas, and Water Production and Supply,” and “Manufacturing.” The query filtered enterprises with registered capital of over 10 million RMB and institutional type classified as enterprises. The data cover industrial enterprises registered between January 1, 2005, and December 31, 2020, resulting in information on a total of 405,851 industrial enterprises. The data for the technology network come from patent data sourced from the China National Intellectual Property Administration website. Among the patent types in China—invention patents, utility models, and design patents—this study focuses on invention patents due to their higher technological content. The dataset includes a total of approximately 3.8146 million invention patent records from 2005 to 2020.

According to the "Yangtze River Delta Regional Integrated Development Plan Outline" approved by the State Council in 2019, the Yangtze River Delta (YRD) includes the entirety of Shanghai Municipality, Jiangsu, Zhejiang, and Anhui provinces, comprising a total of 41 prefecture-level and above cities. In 2023, the combined GDP of the three provinces and one municipality reached approximately 30.5 trillion yuan, accounting for nearly one-quarter of the national economic output. The region also recorded 819,200 granted invention patents in 2023, representing 30.21% of the total patents granted nationwide. Overall, the Yangtze River Delta is one of China’s most economically dynamic regions, boasting the strongest technological innovation capacity and the highest degree of openness to the outside world, playing a key role as the stabilizing force of China’s economy.

This study calculates the industrial resilience scores for 41 prefecture-level cities in the YRD from 2005 to 2020. To ensure comparability across cities, the results are standardized using the Z-score method. Spatial visualizations of industrial resilience for four benchmark years—2005, 2010, 2015, and 2020—are presented in Fig. 1. The following key findings emerge: First, industrial resilience in the YRD shows a general trend of rising initially and then declining. The mean resilience scores for the four selected years are 0.386 (2005), 0.606 (2010), 0.374 (2015), and − 0.986 (2020). The significant increase in 2010 suggests that the YRD demonstrated strong industrial resilience in the 2008 global financial crisis. In contrast, the sharp decline in 2020 can be largely attributed to the accelerated exit of industrial firms under the impact of the COVID-19 pandemic. Second, regional disparities in industrial resilience across the YRD initially narrowed but later widened. The standard deviation of resilience scores for the four years are 1.713 (2005), 1.344 (2010), 1.342 (2015), and 1.926 (2020), indicating that disparity decreased after the global financial crisis but increased again during the pandemic. Third, a clear pattern of inter-provincial variation in industrial resilience emerges. Shanghai consistently exhibits higher levels of resilience, followed by Jiangsu, Anhui, and Zhejiang in 2015 and 2020. Notably, in 2010, Anhui outperformed Jiangsu in terms of industrial resilience, possibly reflecting the effects of targeted regional development policies. Fourth, in terms of spatial clustering, the high-resilience cities in the YRD do not exhibit a strong and consistent spatial concentration pattern over the years. However, certain years—particularly 2010 and 2015—do show visible clusters of high resilience in northern Jiangsu and northern Anhui. These findings highlight the temporal dynamics and spatial heterogeneity of industrial resilience in the YRD and underscore the need to consider both firm-level and regional structural factors in resilience-building strategies.

To assess the robustness of technological networks across the YRD, we compute network robustness under two scenarios—random removal and targeted removal—for four benchmark years. The results are illustrated in Figs. 2 and 3. A spatial analysis of the technological network robustness in the YRD reveals several key insights (Fig. 2): First, technological networks are significantly more robust under random removal than under targeted removal. In 2020, the average robustness threshold for targeted removal across the 41 cities in the YRD was 2.9%, compared to 25% under random removal. This indicates that removing just the top 2.9% of highest-degree nodes causes the network to lose its giant connected component and collapse, whereas a network can endure random removal of up to 25% of its nodes before disintegration. Second, under targeted removal, cities such as Shanghai, Hangzhou, Suzhou, and Nanjing exhibit relatively high levels of network robustness. Under random removal, the high-robustness cities include Shanghai, Hangzhou, Nanjing, Suzhou, and Hefei. Overall, whether under random or targeted node removal, municipalities directly under central government jurisdiction (Shanghai), provincial capitals (Nanjing, Hangzhou, Hefei), and economically advanced cities (Suzhou) demonstrate stronger technological network robustness. However, Hefei’s lower robustness under targeted removal highlights its vulnerability to disruptions at critical technological nodes, suggesting an insufficient capacity to address potential “chokepoints” in key technology areas.

Figure 3 presents the distribution characteristics of technological network robustness in the YRD for the years 2005, 2010, and 2015. Several key patterns emerge: In 2005, under both random and targeted removal scenarios, only Shanghai exhibited high robustness, while most other cities were categorized as low-value regions. This suggests that, aside from Shanghai, the technological networks in the YRD were generally at a low and evenly distributed development stage, characterized by low node degrees and weak connectivity, resulting in limited network robustness. By 2010, Shanghai remained the only city in the high-value category across both removal scenarios. However, mid-high value cities such as Nanjing, Hangzhou, and Suzhou began to emerge, especially under random removal, where their network robustness saw significant improvement. This indicates the beginning of a more differentiated and hierarchical regional technological network structure. In 2015, under targeted removal, Shanghai, Suzhou, and Nanjing entered the high-robustness category, while Hefei, Wuxi, Hangzhou, Zhenjiang, Fuyang, and Wenzhou developed into mid-high value regions. Under random removal, Shanghai and Suzhou remained highly robust, while Wuxi, Nanjing, Hefei, Hangzhou, and Wuhu were classified as mid-high value regions. These changes reflect a significant enhancement in technological network robustness across YRD cities during this period. Overall, both under random and targeted removal, the robustness of technological networks in the YRD increased steadily over time. Robustness was consistently higher under random removal than under targeted removal. Shanghai maintained its position as the most robust city throughout the period, while Nanjing, Suzhou, Hangzhou, and Hefei—as provincial capitals or regional economic hubs—experienced rapid growth in network robustness.

To further examine the long-term dynamics of technological network robustness, Fig. 4 presents the tolerance levels of four representative cities—Shanghai, Nanjing, Hangzhou, and Hefei—to targeted and random removal from 2005 to 2020. In the figure, the green line represents targeted removal, the yellow line indicates random removal, and the red dashed line marks the critical threshold (K = 2) at which the giant component collapses. The intersection of the green line and the red line represents the maximum proportion of targeted removal a city's technological network can tolerate before structural collapse, while the intersection of the yellow line and the red line denotes the maximum tolerance to random removal. The results show that Shanghai demonstrates the highest robustness, withstanding up to 27% targeted removal and 89% random removal, significantly outperforming the other cities. Nanjing, Hangzhou, and Hefei show targeted removal tolerances of 20%, 20%, and 15%, respectively, and random removal tolerances of 84%, 82%, and 81%. These findings reinforce that, over a long temporal scale, Shanghai consistently exhibits the most robust technological network in the YRD. Notably, its robustness under targeted node attacks far surpasses that of the other cities. While Nanjing and Hangzhou show relatively strong performance under random attacks, their networks remain vulnerable to targeted attacks. Hefei, in particular, displays limited robustness under targeted removal.

This study employs regional industrial resilience as the dependent variable, while the explanatory variables include: technological network robustness, related variety (RV), unrelated variety (UV). Technological network robustness is captured using six indicators, namely: the maximum tolerable percentage of targeted node removal (Targeted Removal), the maximum tolerable percentage of random node removal (Random Removal), a series of scenario-based indicators reflecting robustness under varying levels of node removal—specifically: 20% targeted (or 80% random) removal (target_02); 40% targeted removal (target_04); 60% targeted removal (target_06); 80% targeted removal (target_08). To investigate how technological related and unrelated variety, as well as technological network robustness, influence regional industrial resilience, this study constructs a panel ordinary least squares (OLS) regression model. The general form of the model is specified as follows:

In the above equation, the meanings of all variables are as previously defined, and represents the regression residual.

The regression results of the econometric models are presented in Table 1. Regarding technological network robustness, Model (1), which uses the maximum tolerance for random removal as the key explanatory variable, shows a significantly positive impact on industrial resilience. This indicates that the higher the robustness of a city's technological network, the stronger its industrial resilience—thus confirming Hypothesis 3. Model (6), which employs the maximum tolerance for targeted removal, similarly demonstrates a statistically significant and positive relationship between network robustness and industrial resilience, further validating Hypothesis 3. Across Models (1) through (6), related variety exhibits a significantly negative effect on industrial resilience. In other words, cities with higher levels of related technological variety tend to exhibit lower levels of industrial resilience, thereby failing to support Hypothesis 1. This result suggests that although related variety can facilitate knowledge spillovers among cognitively proximate industries and enhance regional economic dynamism, it may also reduce a system’s ability to withstand external shocks. Specifically in industrial systems—which tend to be more closed and structurally interdependent—high relatedness reinforces technological coupling and path dependency. As a result, risk transmission becomes more efficient, system redundancy is reduced, and overall industrial resilience is weakened. This finding contrasts with previous research that generally associates related variety with greater economic resilience, highlighting the distinct structural characteristics of industrial systems. Conversely, the regression results across all models reveal a significantly positive effect of unrelated variety on industrial resilience, thereby supporting Hypothesis 2. This suggests that a more diversified and unrelated technological structure contributes to enhanced regional industrial resilience by reducing systemic vulnerability. Furthermore, by comparing the regression coefficients in Models (1) and (6), it is evident that the effects of technological variety (both related and unrelated) are weaker than those of technological network robustness (based on both random and targeted node removal). This highlights the analytical advantage of using a network-based robustness metric over traditional variety indicators in assessing regional industrial resilience. (All independent variables were standardized in the regression models, ensuring comparability across coefficients.)

A comparison between Model (1) and Model (6) reveals that the regression coefficient representing technological network robustness under targeted removal (6.677) is significantly greater than that under random removal (1.240). To further explore this difference, Models (2) through (5) incorporate varying proportions of targeted and random removal. The regression results under each removal scenario are statistically significant, and a clear pattern emerges: as the proportion of targeted node removal increases, the corresponding regression coefficient—indicating the impact of technological network robustness on industrial resilience—also increases. These findings suggest that in the process of enhancing industrial resilience through improving technological network robustness, greater attention should be paid to high-degree nodes within the network. Strengthening the robustness of these key nodes can substantially improve the overall stability of the urban technological network, thereby reinforcing regional industrial resilience.

This study conducts a robustness check of the baseline regression results by replacing key variables. Considering that regional industrial resilience comprises three dimensions—resistance, recovery, and reconfiguration—we substitute the original dependent variable, regional industrial resilience, with resistance. By replacing the dependent variable with a single-dimension measure, we test the stability and consistency of the regression results (in Table 2). The results show that the regression outcomes using resistance are consistent with those obtained using overall industrial resilience, indicating that the findings are robust.

Based on invention patent authorization data and industrial enterprise data from 2005 to 2020, this study quantitatively evaluates the industrial resilience and technological network structure of 41 cities in the YRD. It analyzes the spatiotemporal evolution of network robustness under both random and targeted removals, with a particular focus on the tolerance levels of four representative cities—Shanghai, Nanjing, Hangzhou, and Hefei—under different node removal scenarios. Finally, a panel regression model is employed to explore how related/unrelated technological variety and network robustness influence industrial resilience. Drawing from network science, the study offers new insights into regional industrial resilience and reaches the following key conclusions:

(1) With respect to the temporal and spatial evolution of industrial resilience, industrial resilience in the YRD region shows an overall trend of rising and then declining. A significant drop is observed in 2020. Regional disparities in resilience are notable, with Shanghai consistently exhibiting high resilience levels. However, high-value clusters of industrial resilience are not spatially prominent.

(2) Concerning the technology network robustness, technological networks are significantly more robust under random node removals than under targeted ones. Regardless of the type of node removal, network robustness in the YRD has improved over time. Shanghai remains the most robust city in terms of technological networks, while provincial capitals and regional centers such as Nanjing, Suzhou, Hangzhou, and Hefei have seen rapid improvements.

(3) Concerning the long-term robustness patterns in key cities, over the long term, Shanghai maintains the highest level of technological network robustness in the YRD, especially in response to targeted removals. While cities like Nanjing and Hangzhou perform well under random removal scenarios, their robustness against targeted removals remains relatively weak. Hefei, in particular, shows lower robustness in handling targeted attract.

(4) With regard to the impact of technological network structure on industrial resilience, regression analysis reveals a significant positive relationship between technological network robustness and industrial resilience. The greater the proportion of targeted node removal, the stronger the effect of network robustness on resilience. Among all measures, robustness under targeted removal shows the strongest positive impact. Moreover, related variety negatively affects industrial resilience, while unrelated variety has a significant positive effect. However, the influence of technological diversity on resilience is considerably weaker than that of network robustness.

This study follows the classical analytical framework of regional economic resilience research by constructing an evaluation system for industrial resilience. Starting from the perspective of regional economic structure, it focuses on how technological factor endowments—as a key component of economic structure—influence regional industrial resilience. The study introduces two key innovations: First, building upon the traditional resilience framework, this research employs micro-level firm data to quantitatively assess regional industrial resilience across three process-oriented dimensions: resistance, recovery, and reconfiguration. This approach not only enriches the methods for measuring industrial resilience but also enhances the granularity and specificity of the analysis. Second, the study incorporates a novel network science perspective by introducing the concept and metrics of technological network robustness. This allows for a more refined depiction of regional technological network structures and provides a new lens through which to examine their effects on industrial resilience. By doing so, the study aims to offer fresh insights and understanding for resilience research in the context of regional industrial systems.

Based on the main findings, this study offers the following policy recommendations for optimizing technological network structures and enhancing industrial resilience in the YRD. First, strengthen protection of core technologies. The findings highlight that targeted removal of high-degree nodes significantly undermines network robustness. Therefore, safeguarding key technologies and ensuring that critical technical capabilities remain under local control is crucial for maintaining the structural integrity of technology networks and enhancing industrial resilience. Second, improve network robustness in key cities. Major cities in the YRD—especially regional hubs—must reinforce the robustness of their technological networks, particularly against targeted attacks on key nodes. Aside from Shanghai, most cities demonstrate inadequate robustness under such conditions. Even provincial capitals such as Nanjing, Hangzhou, and Hefei exhibit limited robustness, underscoring the urgent need to fortify the network structures of central cities. Third, leverage network structure for resilience enhancement. Technological network structure offers a promising pathway to improving regional industrial resilience. Enhancing network robustness (whether under targeted or random removal scenarios) and promoting unrelated variety in technology development are effective strategies to strengthen a region's capacity to withstand and adapt to external shocks.