Keywords: new-quality productive forces; electric vehicle; intelligent curriculum; artificial intelligence; knowledge graph; SPOC; personalized learning; digital literacy.
Global economic restructuring and rapid advances in science and technology have made new-quality productive forces a central engine of high-quality social development. Under this condition, higher education must cultivate what has come to be called new-quality talent, which means young engineers who possess both creative thinking and digital literacy. In my own teaching practice, I have found that this broad educational demand has direct consequences for how we design courses related to the electric vehicle. The transition toward electrified mobility requires engineers who can synthesize mechanics, automation, electrical engineering, computing, energy storage, and new materials. Therefore, I propose a new intelligent curriculum paradigm for electric vehicle design courses. This paradigm seeks to integrate new-quality productive forces theory with intelligent technologies, digital resources, artificial intelligence platforms, and learner-centered pedagogies.
The background of this proposal is not abstract. In my experience, conventional electric vehicle design teaching often suffers from slow curriculum updating, linear knowledge delivery, insufficient multi-disciplinary integration, passive lectures, and assessment that measures memory rather than competence. Artificial intelligence opens a different path. AI can analyze educational resources, generate structured knowledge maps, recommend personalized learning materials, provide immediate feedback, and evaluate complex competencies. At the same time, small private online courses provide a flexible environment for combining classroom interactions with digital learning. My work, therefore, is based on an effort to merge digital resource construction and smart course platforms with a SPOC-based hybrid teaching model.
In the remainder of this article, I first outline the overall architecture of the new paradigm for electric vehicle design education. Then I describe the construction of digital resources and an AI-supported smart platform. Next, I explain how I implemented an SPOC teaching design and a diversified assessment system for an electric vehicle design course. Finally, I present data from one teaching cycle and explain how I use these data for continuous improvement. The entire discussion is grounded in the practical aim of nurturing new-quality talent who can guide the electric vehicle industry through innovation.
Educational Challenges in Electric Vehicle Design Courses
Electric vehicle design is a knowledge-intensive engineering discipline. It cannot be taught as a single subject because it sits at the intersection of many fields. The full development cycle of an electric vehicle involves cell chemistry, battery management, electric machine design, power electronics, embedded control, energy management, vehicle dynamics, structural layout, safety regulations, and intelligent connected systems. When I reviewed the current state of teaching, I identified several constraints that prevent these courses from meeting the demands of the new-quality productive forces.
First, content updates in textbooks cannot keep pace with the rapid evolution of electric vehicle technology. For example, new battery chemistries, high-voltage architectures, and motor topologies appear frequently. If course materials remain static, students cannot transfer the most recent industrial knowledge into their own design projects. Second, the course is highly interdisciplinary, but traditional departmental or lecture-based structures tend to divide it into disconnected modules. Third, methods of delivery are frequently based on one-way teacher talk. Students memorize formulas and structures, but they rarely manipulate engineering trade-offs. Fourth, assessment tends to rely on a final examination that emphasizes recall, not iterative design, teamwork, communication, or digital competence. These problems are common in electric vehicle design courses, but they are especially damaging when new-quality productive forces demand a shift from routine technical knowledge to flexible innovation.
To overcome these deficiencies, I argue that the course must be reconstructed around four principles: digitalization, intelligentization, precision, and personalization. Digitalization allows content to be represented through knowledge graphs, video micro-lectures, interactive simulations, and online resource libraries. Intelligentization brings AI into the teaching process to make learning visible and adaptive. Precision means that the course identifies what each student has mastered and what remains difficult. Personalization means that the student has some control over learning paths, project topics, and the pacing of self-study. Together, these principles support the production of new-quality talent with stronger innovation capacity and digital literacy.
Overall Architecture of the Intelligent Curriculum Paradigm
The holistic framework I propose for smart electric vehicle design courses can be described as a chain that connects educational goals, digital resources, platform construction, teaching methods, and assessment. From the perspective of new-quality productive forces, the expected outcome is an engineer who is not merely competent at routine design work but also capable of continuous learning, cross-disciplinary communication, responsible innovation, and ethical use of artificial intelligence. I call this outcome an innovation-constant talent, because the student must maintain a persistent disposition toward creative problem solving even as technologies change.
The conceptual model contains four layers. The first layer is digital resource construction. At this layer, I exploit large language models and knowledge-graph technology to reorganize course content. The second layer is smart platform construction. An online learning platform supports students with videos, quizzes, discussion forums, AI tutoring, and progress dashboards. The third layer is pedagogical design. I use an AI-enhanced SPOC model that bridges online and offline activities. The fourth layer is evaluation. The assessment system evaluates process, product, disciplinary depth, collaboration, digital competence, and creative output.
An important formula for thinking about the target competence of the new-quality talent is the following weighted outcome model:
$$C_{\text{target}} = \alpha_1 C_{\text{innovation}} + \alpha_2 C_{\text{digital}} + \alpha_3 C_{\text{discipline}} + \alpha_4 C_{\text{collaboration}}$$
where
$$0 \le \alpha_i \le 1, \quad \sum_{i=1}^{4} \alpha_i = 1.$$
In my institution, I have found it useful to start with weights that recognize disciplinary mastery as necessary but not sufficient. For a smart electric vehicle design course, the final competency is a vector, not a single number. A student might know battery formulas but fail to formulate a creative design question. Another student might have good software skills but cannot analyze the physical limitations of an electric vehicle powertrain. The architecture must therefore supply separate evidence streams for each competence dimension.
| Architectural layer | Main action | Technology or method | Expected contribution to electric vehicle education |
|---|---|---|---|
| Digital resource construction | Use AI and large language models to build course knowledge graphs | Concept extraction, topic modeling, knowledge representation | Creates a living map of electric vehicle content instead of a fixed textbook |
| Smart platform construction | Deploy online videos, quizzes, discussion, AI tutors, learning analytics dashboards | Online learning platform, AI question bank, learning record system | Supports self-paced study before and after class |
| Hybrid pedagogy | Blend offline classroom and online SPOC tasks | BOPPPS lesson planning, project-based learning, AI assistance | Enables teacher-student-machine interaction for electric vehicle projects |
| Diversified assessment | Measure learning process, project quality, and digital behavior | Rubrics, online logs, peer review, AI-assisted scoring | Provides reliable evidence about student creativity and digital ability |
The structure above can be summarized as a function:
$$P_{\text{electric vehicle}} = f\Bigl(D_{\text{resources}}, S_{\text{platform}}, T_{\text{teaching}}, A_{\text{assessment}}\Bigr)$$
In this expression, each capital letter represents a set of course design decisions. In the following sections, I explain how I instantiate those decisions in an actual course named “Fundamentals of Electric Vehicle Design.”
Digital Resource Construction for an Electric Vehicle Design Course
The initial step in my proposed paradigm is not to start with lectures. I first convert the professional content of an electric vehicle design course into a structured digital knowledge space. Large language models are useful at this point because they can process the syllabus, digital textbooks, prior examination questions, published standards, and recent industry reports. I use these models to extract key terms, identify core concepts, and reveal latent relationships among modules. After the extraction process, the content is organized into the fundamental domains of electric vehicle design: the battery system, the electric drive system, the electronic control system, and the complete vehicle architecture.
In the battery domain, the course covers cell types, performance indicators, pack structure, thermal behavior, state estimation, charging protocols, and safety. In the electric drive domain, students study permanent-magnet synchronous motors, induction motors, switched reluctance motors, motor losses, efficiency maps, and inverter constraints. In the electronic control domain, the course emphasizes energy management strategies, vehicle controllers, torque allocation, regenerative braking, and communication protocols. In the complete vehicle domain, learners examine longitudinal dynamics, system-level parameter matching, lightweight structures, thermal integration, and the layout of electric vehicle components. These four modules do not remain isolated. Their relationships define the overall course graph.
Formally, the knowledge graph for the electric vehicle design course is represented as:
$$G_{\text{EV}} = (V, E, W)$$
where \(V\) would be the set of knowledge nodes, \(E\) the set of directed or undirected relationships, and \(W\) the edge strength. Since all formulas are rendered with double dollar delimiters, the mathematical expression is:
$$V = \{v_1, v_2, \dots, v_n\}, \quad E \subseteq V \times V, \quad W: E \rightarrow [0,1].$$
Each node \(v_i\) may represent a concept, a formula, a design rule, a failure mode, or an industrial standard. In my course, I make sure that the graph has both hierarchical edges, such as “is-a” or “consists-of,” and procedural edges, such as “must be calculated before” or “is optimized together with.” This graph helps students see why a battery management system decision affects motor torque control and why a vehicle-level target influences component design.
To build the graph, I first define a set of modules:
$$M = \{M_{\text{battery}}, M_{\text{motor}}, M_{\text{control}}, M_{\text{vehicle}}\}.$$
Each module contains local knowledge nodes. Then I calculate semantic similarity between any two nodes using a Jaccard-style overlap of associated digital documents. For two nodes \(v_i\) and \(v_j\), the relation weight is:
$$w_{ij} = \frac{\left|D_{v_i} \cap D_{v_j}\right|}{\left|D_{v_i} \cup D_{v_j}\right|}$$
where \(D_{v_i}\) is the digital content associated with node \(v_i\). The resulting structure helps me plan videos, examples, and tests. For example, if the weight between “battery thermal management” and “power limitation under high temperature” is high, I place both concepts close together in the instructional sequence. The knowledge graph is not just a static map. It is the backbone for generating personalized learning paths in the smart platform.
| Electric vehicle domain | Representative core knowledge nodes | Digital resources I produced | Connection to new-quality productive forces |
|---|---|---|---|
| Battery system | Cell voltage, energy density, power density, internal resistance, thermal runaway, state of charge estimation | Micro-lectures, battery test data, animation of internal reactions, interactive thermal simulation | Builds ability to evaluate new battery chemistry such as solid-state batteries |
| Electric drive system | Motor torque-speed curve, efficiency map, magnetic flux weakening, inverter switching, loss distribution | Motor design videos, Ansys data sets, comparison tables for permanent-magnet and induction motors | Encourages students to question conventional motor design assumptions |
| Electronic control | VCU functions, torque arbitration, state machine logic, CAN communication, energy management strategies | Simulink model examples, fault-case quizzes, coding examples for rule-based and optimization-based control | Strengthens programming and intelligent decision-making spirit |
| Complete vehicle | Driving cycles, resistance calculation, gear ratio selection, battery-motor matching, lightweight structure | Vehicle simulation cases, MATLAB scripts, project briefs with real-world electric vehicle parameters | Develops systems engineering view of electric vehicle development |
From the knowledge graph, I identify core knowledge points that every student must master. For instance, the performance index of a power battery is a prerequisite for many later tasks. Such points become required videos and required quiz items. Other advanced topics, such as digital twin battery management or AI-based thermal prediction, become elective materials. The distinction between required and elective knowledge is crucial for SPOC design because it preserves the baseline of disciplinary quality while still allowing personalized exploration.
Construction of an Online and Offline Smart Platform
A smart curriculum for electric vehicle design cannot exist without an online platform. I built the platform around a web-based learning environment that supports short video modules, embedded quizzes, discussion forums, assignment submission, group management, and AI-based learning analytics. The platform stores a record of every student’s behavior. These records include video watching time, pause events, test attempts, correct response rates, forum messages, project downloads, and optional-content browsing. Such data are indispensable for precision education in the electric vehicle field.
I used the following design concept: every face-to-face lecture has a corresponding online learning path. Before class, students watch a required video about the main concept of an electric vehicle system and complete a short diagnostic quiz. During class, I use the diagnostic result to decide what to emphasize. After class, students review additional videos, solve practice problems, and begin project-related tasks. In the new-quality productive forces context, the online platform also introduces students to digital tools for design, such as simulation software, data analysis scripts, and AI-driven design assistants.
Artificial intelligence supports the smart platform in several ways. First, AI helps generate many variants of practice problems from a small number of expert-authored templates. For example, a question about electric vehicle range estimation can be instantiated with different battery capacities, ambient temperatures, driving cycles, and initial states of charge. Second, AI creates intelligent tutoring feedback. When a student gives an incorrect response, the platform does not simply reveal the correct answer. Instead, it suggests a micro-lesson related to the missing prerequisite. Third, AI makes personalized project recommendations. At the beginning of the project phase, the system asks each student about their interest, self-rated skill with MATLAB, experience with CAD, and preference for battery, motor, control, or complete vehicle tasks. Based on this vector, the system returns a ranked list of electric vehicle design directions.
A simplified mathematical model for personalized topic recommendation is:
$$\hat{y} = \arg\max_{c \in C} \left( \beta_1 \operatorname{sim}(x, c) + \beta_2 h_c \right)$$
where \(x\) is the student feature vector, \(c\) is a candidate project cluster, \(h_c\) is the current industrial heat of that cluster, and \(\beta_1\) and \(\beta_2\) are balancing parameters. In practice, \(h_c\) might be larger for “solid-state battery pack integration” than for “traditional lead-acid battery replacement.” The result is that the electric vehicle design course can offer different project paths to different students while still maintaining the same learning objectives.
This approach improves the student experience. Instead of requiring all students to complete exactly the same design task, the platform allows them to choose an electric vehicle subsystem that matches their curiosity. One student may investigate battery pack placement and crash safety. Another student may focus on motor control and regenerative braking. A third student may propose a cloud-based battery health monitoring strategy. The classroom becomes a heterogeneous studio, and each student develops deep knowledge in one area while sharing findings with peers in other areas.
| Platform function | Examples for electric vehicle design | AI contribution |
|---|---|---|
| Video learning | Micro video on battery current distribution, motor flux weakening, electric vehicle braking strategy | Generates captions, summaries, and follow-up questions |
| Intelligent quiz bank | Range estimation, energy consumption, torque distribution, safety standards for electric vehicle components | Creates multiple variants with random parameters and individual feedback |
| Project recommendation | Battery pack lightweight design, thermal management strategy, electric vehicle powertrain matching | Ranks options based on student profile and labor market demand |
| Learning analytics | Heat maps showing which electric vehicle concept is most difficult for a class | Detects patterns and suggests interventions |
| AI tutoring | Dialogic help for interpreting battery equivalent-circuit models or motor equations | Uses large language model trained with electric vehicle course material |
The smart platform is therefore not merely a repository of files. It is an active agent in the pedagogy of electric vehicle design. It makes students’ thinking visible and gives me, as the teacher, reliable, high-frequency feedback. When I see that the class as a whole struggles with a particular electric vehicle control concept, I can modify the following classroom session.
AI-Enhanced SPOC Teaching Practice
The pedagogical core of this project is a small private online course combined with face-to-face classroom activity. SPOC, which stands for Small Private Online Course, sits between a fully open massive open online course and a traditional class. It is small enough to maintain community, private enough to preserve confidentiality, and online enough to allow flexible digital learning. In my electric vehicle design course, the SPOC mode includes three stages: preparation before class, participatory learning in class, and consolidation after class. These three stages are informed by the BOPPPS model, which includes six parts: bridge-in, objective, pre-assessment, participatory learning, post-assessment, and summary.
Before class, I publish the learning path for the upcoming electric vehicle design topic. The platform presents a bridge-in question, states the objective, and gives a small pre-assessment quiz. For example, before teaching electric vehicle energy management, I might ask students to watch a short video about why battery temperature affects range. The pre-assessment checks whether they understand the relationship between internal resistance and energy loss. This stage also includes optional frontier tasks, such as reading recent news about solid-state batteries or watching an expert interview about next-generation electric vehicle motors.
During class, I do not repeat the online video content. Instead, I orchestrate active learning. I begin by showing the knowledge graph and asking students to connect the online pre-assessment to the new design problem. Then I present a realistic electric vehicle design scenario. For example, the scenario may involve designing a powertrain for an urban micro electric vehicle with a given target range, acceleration, and maximum speed. Students work in small groups. They calculate road load, choose a motor, estimate battery energy, and discuss charging constraints. AI tools are available as assistive agents. A student who is unsure about the formula for rolling resistance can ask the AI tutor for a hint or a worked example. The teacher moves from group to group, diagnosing obstacles and challenging assumptions.
The AI-enhanced classroom also includes quick interactive quizzes. I call this the teacher-student-machine triangle. The machine collects quiz responses, displays a histogram on the screen, and suggests which concepts require more explanation. The teacher then uses this information to decide whether to revisit a topic or move forward. The student becomes an active producer of data. This design is especially useful for electric vehicle design because many topics are quantitative and depend on the correct application of nonlinear models.
The post-class phase is designed for consolidation and extension. Students complete a homework set, participate in an online forum, and make progress on the personalized project. The project may be related to the same electric vehicle scenario but with different component selection. For example, one group may choose a permanent-magnet motor for a city car, while another group explores an induction motor for a more aggressive highway cycle. Each group submits a short report explaining its assumptions, calculations, and recommendation. The AI platform scores the quantitative parts with a rubric and provides immediate feedback. The teacher later uses the reports to guide a whole-class comparison.
From an instructional viewpoint, the SPOC model can be represented as a learning loop:
$$L_{\text{total}} = L_{\text{pre}} + L_{\text{class}} + L_{\text{post}}$$
where \(L_{\text{pre}}\), \(L_{\text{class}}\), and \(L_{\text{post}}\) are measured learning gains at each stage. In my course, I compute the knowledge mastery for each electric vehicle module as:
$$M_k = \frac{1}{J_k} \sum_{j=1}^{J_k} \frac{R_{j,k}}{N_{j,k}}$$
where \(R_{j,k}\) is the number of correct responses to assessment item \(j\) in module \(k\), \(N_{j,k}\) is the number of attempts, and \(J_k\) is the number of items associated with that module. Monitoring this quantity over time tells me whether the SPOC cycle is working.
| SPOC stage | Teacher action | Student action | AI/machine action | Electric vehicle design task example |
|---|---|---|---|---|
| Pre-class | Release path, set objective, diagnose prior knowledge | Watch micro-lecture, complete pre-quiz, post questions | Records time, suggests materials, identifies weak nodes | Review driving resistance model before range calculation |
| In-class | Facilitate scenario, lead discussion, clarify errors | Calculate, argue, prototype, debug, present preliminary result | Runs live polls, generates hints, stores all interactions | Select motor and battery for an urban electric vehicle architecture |
| Post-class | Design assignment, grade project, adapt next lesson | Complete homework, revise project report, discuss online | Auto-grades numerical tasks, forms concept clusters, sends reminders | Optimize gear ratio or battery state-of-charge strategy |
One of the most valuable parts of AI-enhanced SPOC is that it supports individual project selection. Traditionally, all students in an electric vehicle design course answer the same problem set. I now allow the platform to generate personalized project briefs from a matrix of battery, motor, control, and vehicle parameters. The teacher reviews the AI-generated briefs to make sure they are technically feasible and aligned with course objectives. Then the student chooses a brief. This process preserves autonomy while preventing students from drifting into topics that require prerequisites they do not have.
The personalized topic generation process can be expressed as a set of constraints:
$$\text{Project}_{s} = \left( M_{\text{battery}} \times B_s,\; M_{\text{motor}} \times E_s,\; M_{\text{control}} \times C_s,\; M_{\text{vehicle}} \times V_s \right)$$
where \(B_s\), \(E_s\), \(C_s\), and \(V_s\) are binary or continuous preference weights given by student \(s\). A student interested in battery management might set \(B_s=1\), \(E_s=0\), \(C_s=0.5\), \(V_s=0.5\). The AI then designs a project around battery thermal limitations and vehicle-level performance constraints. The result is a richer educational experience because every student is using the same underlying theory but applying it to a different subsystem of the electric vehicle.
Diversified Process-Oriented Assessment System
New-quality productive forces require more than examinations. A smart electric vehicle design course should evaluate students on their ability to navigate uncertainty, use information, model systems, communicate technical results, and reflect on their own learning. For this reason, I redesigned the course assessment as a diversified process. The final grade is no longer a one-shot indicator; it is a composite of online activity, classroom interactions, project work, and final exam performance. In my course, the total score is divided into online and offline components:
$$S_{\text{total}} = S_{\text{process}} + S_{\text{final}}$$
where \(S_{\text{final}}\) is the score of a closed-book final examination and \(S_{\text{process}}\) contains all coursework collected during the semester. The process grade itself has the following structure:
$$S_{\text{process}} = 5 P_{\text{pre-task}} + 5 H_{\text{homework}} + 5 C_{\text{class}} + 10 Q_{\text{chapter}} + 25 R_{\text{project}}$$
Each weight is expressed as a percentage of the total course score. Thus, pre-class tasks count for 5%, homework for 5%, class participation for 5%, chapter tests for 10%, and the personalized project report for 25%. The final examination accounts for the remaining 50%.
In my electric vehicle design course, the pre-class tasks are either required or optional. Required tasks include watching the core video and answering a question about the specified electric vehicle concept. Optional tasks might include reading an article about a new power battery or listening to a short podcast about electric vehicle charging infrastructure. Optional tasks provide bonus evidence of curiosity. Homework questions are generated from the knowledge graph, with each item connected to one or more knowledge nodes. The chapter tests are timed online quizzes that occur after completing a module. The personalized project is the largest process component; it requires students to apply the full design cycle to an electric vehicle subsystem of their choice.
One central question is how to assess innovation and digital literacy, which are not easily measured by a multiple-choice quiz. I use rubrics that are aligned with new-quality talent. For the project, the rubric includes problem definition, literature research, modeling depth, use of simulation or programming, feasibility, creativity, collaboration, and reflection. The AI platform provides quantitative assistance by detecting whether students use certain digital tools consistently. The teacher evaluates creativity and physical reasoning through the written report and face-to-face defense.
From an assessment-theory viewpoint, the reliability of a composite score depends on the interrelationship among its components. I define the overall digital behavior index as:
$$D_{\text{behavior}} = \frac{1}{T_{\text{total}}} \sum_{i=1}^{N} t_i \cdot q_i$$
where \(t_i\) is the time spent on online activity \(i\), \(q_i\) is the quality score of that activity, and \(T_{\text{total}}\) is the total active learning time. This index is not used as a hard grade but as diagnostic information. If a student has a high final exam score but low online engagement, I may need to ask whether the exam rewarded shallow learning. If another student has high project quality but low chapter-test scores, I may recommend additional review of fundamental electric vehicle calculations.
Finally, the assessment system is not static. I use regular feedback surveys to adjust the weighting scheme when necessary. The goal is to ensure that students do not game the assessment system by focusing exclusively on one part of the electric vehicle course. Because the industry itself is changing, the course must remain responsive to emerging priorities such as cybersecurity, lifetime value of electric vehicle batteries, and the environmental impact of rare-earth motors.
| Assessment component | Weight | Primary electric vehicle skill measured | Tools and evidence |
|---|---|---|---|
| Pre-class required and optional tasks | 5% | Self-management; preparation for electric vehicle theory | Video logs, online quiz, optional article access |
| Homework | 5% | Application of electric vehicle equations to concrete cases | Numerical assignments, short design problems |
| Class participation | 5% | Communication, argumentation, collaborative electric vehicle design | Teacher observation, live polls, discussion questions |
| Chapter tests | 10% | Mastery of battery, motor, control, and vehicle knowledge nodes | Online objective tests, automatically generated variants |
| Personalized project report | 25% | Innovation, digital modeling, system-level thinking | AI-recommended topic, report, defense, peer evaluation |
| Final exam | 50% | Integrated understanding of electric vehicle design fundamentals | Closed-book written examination |
This assessment architecture allows me to claim that the course targets new-quality talent because it rewards both the process and the product of learning. It also makes the electric vehicle design process transparent to students. They know from the beginning that active digital behavior, project initiative, and reflective collaboration will be counted. This transparency changes their attitude from grade chasing to competence building.
Implementation Effect and Data-Driven Improvement
After one complete semester of this intelligent electric vehicle design course, I collected student interaction data from the smart platform. The data reveal important information about learning interests and difficulties. The platform allows me to see which optional frontier topics are clicked most often and which knowledge graph nodes produce the lowest test scores. In this section, I report representative data from the course and explain how I use it to improve the next session.
The online platform gives me a measure of thematic interest. For optional frontier videos and articles related to electric vehicle technology, I define the click rate as follows:
$$C_{\text{theme}} = \frac{n_{\text{clicks}}}{n_{\text{active students}}} \times 100\%$$
where \(n_{\text{clicks}}\) is the number of unique or repeated clicks on an electric vehicle theme and \(n_{\text{active students}}\) is the number of students who entered the optional resource library. A higher click rate indicates stronger student curiosity about that emerging electric vehicle topic.
From my teaching data, I found that solid-state battery materials were the most attractive frontier topic. The click rate reached 95.3%. This finding is interesting because solid-state batteries are often regarded as the next major breakthrough for electric vehicle range and safety. Students also showed strong interest in blade battery structure, with 94.5% click rate, perhaps because of its widespread use in popular electric vehicle models. Hairpin motor technology was also very popular. The learning analytics system logged 87 questions or comments about hairpin winding design. Fuel-cell electric vehicle technologies and energy recovery topics also received high attention.
| Optional frontier topic in electric vehicle design | Click rate | Evidence of interaction | Implication for course adaptation |
|---|---|---|---|
| Solid-state battery | 95.3% | High click rate and many forum questions | Add a new mini-lecture on solid-state electrolyte materials; provide more project options |
| Blade battery structure | 94.5% | Students compared blade battery with conventional pouch cells | Insert practical pack architecture cases into the battery module |
| Hairpin winding motor | 91.2% | 87 questions or comments | Add an interactive electric motor winding simulation |
| Fuel-cell electric vehicle | 88.6% | Many students asked about hydrogen storage safety | Invite a guest lecture or show a virtual tour of a hydrogen refueling station |
| Regenerative braking and energy recovery | 87.9% | Students requested more braking torque distribution examples | Provide a MATLAB exercise on cooperative braking between electric motor and friction brake |
These data support the value of an intelligent course architecture. It reveals that today’s students are not passive. They want to understand the technologies that will shape the future electric vehicle market. At the same time, the data help me decide how much frontier material to weave into the core curriculum. The new-quality productive forces view holds that such materials should become part of the normal electric vehicle design process, not mere side attractions.
In addition to interest data, I analyzed the chapter-test error rate. The difficulty of each knowledge node is easy to measure because each quiz item is linked to a node in the electric vehicle knowledge graph. For each node, I record the correct response rate. I use the following definition:
$$\rho_k = \frac{N_{\text{correct},k}}{N_{\text{attempt},k}} \times 100\%$$
where \(k\) is the electric vehicle knowledge node. In the first teaching cycle, five difficult nodes were especially important. The lowest correct rate was associated with the calculation of electric vehicle power battery parameters. Only 80.0% of students could correctly solve problems that required combining battery energy, capacity, state of charge, and effective range. Another difficult node involved motor parameter calculation, with 82.0% correct. The configuration of hybrid electric vehicle architectures also created obstacles, with 83.0% correct. The calculation of powertrain parameter matching was likewise difficult, with 83.0% correct. Energy management strategy problems were slightly easier, at 85.0%, but still required attention.
| Knowledge weak point in electric vehicle design | Correct response rate | Likely cause identified from learner data |
|---|---|---|
| Power battery related calculations | 80.0% | Students confuse capacity, energy, C-rate, and range; unit conversions are weak |
| Electric motor parameter calculation | 82.0% | Students memorize the motor equation but cannot interpret torque-speed limits at high speed |
| Hybrid electric vehicle architecture diagram | 83.0% | Students cannot distinguish series, parallel, and power-split architectures under load conditions |
| Powertrain parameter matching | 83.0% | Difficulty coordinating motor peak torque, battery discharge power, and vehicle mass |
| Energy management strategy | 85.0% | Students struggle to understand rule-based versus optimization-based control for plug-in electric vehicles |
These weak points are not surprising in a complex electric vehicle design course. What matters is that I can now see them clearly. In previous teaching, I might have discovered only after the final examination that many students misunderstood powertrain matching. Now, the smart platform detects the problem as soon as the chapter test is complete. This detection is part of precision education.
I use the data to improve the course in multiple ways. First, I subdivide the problematic videos into smaller micro-lessons. For example, the original video about battery capacity and range is about twelve minutes. I break it into shorter segments: one segment on capacity and energy, one segment on effective range under real driving cycles, and one segment on battery discharge limits. This segmentation helps students find and repair the exact point of confusion. Second, I train the AI tutoring assistant to address the specific electric vehicle misconception. If a student asks about a battery calculation, the assistant now asks whether the student is confusing kilowatt-hours with ampere-hours and then offers a worked example. Third, I modify the classroom discussion to include more comparison of electric vehicle architectures. When the data show that hybrid architecture diagrams are difficult, I schedule a group activity in which students draw and compare the power flows of different electrified powertrains.
The improvement mechanism can be quantified as a gain score:
$$\Delta \rho_k = \rho_k^{\text{after intervention}} – \rho_k^{\text{before intervention}}$$
My goal is to raise the correct response rate for each weak electric vehicle concept by at least five percentage points. For the most difficult battery-related calculation, I expect that a combination of micro-videos, visual examples, and AI tutoring will produce such an improvement. If it does not, I will examine whether the prerequisite math skills are low and then adjust the course design accordingly.
| Detected difficulty in electric vehicle design | Refined video or resource action | AI tutoring action | Classroom intervention |
|---|---|---|---|
| Power battery calculation | Split video into capacity, current, energy, and range examples | Generate step-by-step battery sizing questions with random data | Add a live calculation challenge using real electric vehicle specifications |
| Motor parameter calculation | Publish torque-speed map as an interactive graph | Quiz students about operating points above rated power | Use a motor dynamometer virtual lab to observe heating and efficiency |
| Hybrid architecture drawing | Provide architecture flowcharts with animated power paths | Ask students to match architecture type with driving condition | Conduct a think-pair-share activity in which each group explains one architecture |
| Powertrain matching | Create a spreadsheet-based matching tool | Give tailored hints about battery and motor limits | Organize an optimization contest for a target electric vehicle range |
| Energy management | Provide a rule-based control flowchart in Python or MATLAB | Generate scenario questions about charge-depleting and charge-sustaining modes | Debate whether to prioritize battery life or energy efficiency in an electric vehicle |
This continuous improvement cycle is essential for new-quality productive forces. The electric vehicle industry is evolving so quickly that no fixed curriculum can remain valid forever. By collecting data from every learning episode, I can update the course more frequently and with stronger evidence. The intelligent platform also encourages teachers to become lifelong learners. I must understand AI recommendations, interpret learning analytics, and evaluate new electric vehicle technologies before I can guide students.
Relationship Between the Paradigm and New-Quality Talent
The entire purpose of this course redesign is to produce new-quality talent for the electric vehicle ecosystem. New-quality talent is not defined by a static list of skills. It is defined by the ability to contribute to the development of new industries, new business models, and new technologies. In the context of electric vehicle design, such talent must be able to move between the physical world of motors and batteries and the abstract world of data and algorithms. My proposed paradigm advances this goal by linking five types of capability.
The first capability is innovation. The personalized project requires each student to propose a design improvement, benchmark it, and defend it. The second capability is digital competence. Students use online platforms, knowledge graphs, AI assistants, simulation software, and data analytics. The third capability is disciplinary depth. The SPOC design preserves a strong foundation in electric vehicle engineering principles. The fourth capability is systems thinking. The knowledge graph and the vehicle-level architecture force students to see interactions between battery, motor, control, and vehicle. The fifth capability is lifelong learning. The frontier optional resources and AI-supported feedback teach students how to learn on their own after the course has ended.
From a cognitive point of view, one can model the transformation process as:
$$\text{Learning gain} = \int_{0}^{T} \frac{\partial C_{\text{knowledge}}(t)}{\partial t} dt + \gamma \cdot \text{innovation score}$$
where \(T\) is the duration of the electric vehicle course and \(\gamma\) is a coefficient describing how strongly creativity is promoted. Although such a continuous model is an abstraction, it emphasizes that the course is a dynamic system. The teacher provides resources, the student enacts learning activities, and the platform connects the two. This dynamic approach is more suitable for the electric vehicle field than a static lecture-based model.
In my own teaching, I observed that the AI-enhanced course motivated students to go beyond the minimum requirements. Several students who started with an interest in battery design chose optional resources about solid-state batteries and then used those resources in their final project. The project reports demonstrated a higher level of original thinking than previous cohorts. Students were using real electric vehicle data, constructing their own simulation models, and explaining trade-offs in ways that had not been common before the implementation of the new paradigm.
Challenges and Reflections on Intelligent Electric Vehicle Education
Although the new paradigm has many benefits, I must be honest about its challenges. First, building and maintaining a high-quality smart platform requires time and institutional support. The teacher cannot expect to produce dozens of micro-videos, build a knowledge graph, train an AI assistant, and analyze large amounts of student data without released time. Second, AI-generated content must be reviewed for correctness. Electric vehicle engineering is safety-critical, and hallucinations in AI-generated examples can be dangerous. I always verify equations, design assumptions, and numerical examples before presenting them to students. Third, some students initially feel uncomfortable with personalized learning. They expect the teacher to provide a single answer and a single pace. I address this by explaining the educational philosophy at the beginning of the course and by providing accessible scaffolding.
Another challenge is that digital learning analytics can threaten student privacy. The platform records video watching time, test answers, forum activity, and even hesitation patterns. I use these records only for pedagogical purposes and follow ethical guidelines. Students are informed of what data are collected and why they matter. The goal is not surveillance; it is compassionate adaptation to learner needs. In the electric vehicle domain, engineers must learn that data-driven decision-making, which is essential in the field, also requires ethical awareness. My course treats responsible data use as part of new-quality talent.
The assessment system also requires careful vigilance. No formula can fully capture creativity, ethical judgment, or physical intuition. The composite score is a proxy. Therefore, I use several qualitative indicators, such as project defense, peer review, and self-reflection essays. When the automated grade and the teacher’s professional judgment differ, I conduct a deeper review. This cautious attitude is important for the electric vehicle industry because a design decision that looks acceptable on paper may be poor when tested under extreme conditions.
I also recognize that not every institution has the same access to technological infrastructure. A smart electric vehicle design course requires reliable internet, learning management systems, computing devices, simulation software, and perhaps laboratory hardware. In regions with less robust infrastructure, a simplified version of the course can still begin with knowledge graphs and offline classroom interaction. The most important principle is not the specific technology but the idea that teaching should be designed as a feedback loop. Digital resources can be modest in scale as long as the teacher is willing to use student data to improve the next lesson.
During my own implementation, I maintained an iterative design cycle. I reviewed every module after completion and asked several questions. Did the required electric vehicle concept appear clearly? Did the optional resources extend the knowledge graph in a way that was noticeable? Did the AI tutor provide accurate feedback? Were the chapter test items aligned with both disciplinary standards and industrial practice? How did the students experience the workload? These questions led to many small modifications. For instance, I realized that the AI-generated project topics were sometimes too broad. A brief saying “design a battery for an electric vehicle” is too vague. I therefore refined the prompt to require a specific vehicle class, driving cycle, thermal constraint, and performance target.
Future Directions for Intelligent Electric Vehicle Courses
Looking ahead, I plan to expand the use of generative AI in ways that better reflect industrial design workflows. In the real electric vehicle industry, engineers use AI for concept generation, optimization, simulation, and validation. My course already uses AI for personalized learning and feedback, but I want students to use AI as a design tool. For example, students could use an AI assistant to generate initial electric vehicle architecture alternatives. Then they would calculate whether those architectures satisfy range and cost constraints. They would compare human-designed and AI-generated solutions, developing critical trust in artificial intelligence. This activity directly supports the digital literacy expected by new-quality productive forces.
I also intend to strengthen the connection between the course and industry data. Real electric vehicle data are often scattered across test reports, public databases, vehicle telematic records, and charging station logs. I can build a course resource repository that contains real anonymized data from an electric vehicle fleet. Students can learn to clean data, analyze energy consumption, detect anomalies, and propose improved control strategies. This experience makes the electric vehicle design course more authentic and allows students to encounter the uncertainty that exists in actual engineering. Authentic data also help students understand that electric vehicle performance depends on climate, driver behavior, road topography, and battery aging.
Another promising direction is cross-course collaboration. An intelligent electric vehicle design course should not exist as an isolated island. It can collaborate with courses in artificial intelligence, embedded systems, materials science, human-computer interaction, and business strategy. The knowledge graph can be shared with other instructors. Suppose a control course and an electric vehicle design course both refer to regenerative braking. The control course could focus on the algorithm that coordinates braking torque, while the design course focuses on system-level brake and motor sizing. By exchanging student data and course rubrics, the two courses can create a better vertical integration of the curriculum. This collaborative model is itself a reflection of the interdisciplinary nature of electric vehicle engineering.
In the domain of assessment, I plan to introduce more scenario-based and project-based evaluation. Traditional final examinations are limited by time and cannot fully assess whether a student can repair an unexpected electric vehicle control failure. I am working on an open-ended design challenge in which students are given a partially completed electric vehicle project and asked to diagnose problems, propose modifications, and predict performance using simulation. Such a challenge could be conducted in a digital laboratory where students interact with AI-generated fault reports. The challenge would test not only knowledge but also resilience and creativity.
There is also an opportunity to make the course part of the larger “industry-university-research” ecosystem. In the field of electric vehicles, industry partners have access to the latest high-voltage architecture and battery testing equipment. If the course invites practitioners to provide real problems, students will benefit from authentic scenarios. At the same time, the smart platform can support collaboration among geographically separated teams. Students from one university can co-design an electric vehicle component with students from another university or from industry R&D centers. The SPOC model is a natural fit for such distributed collaboration. I hope to pilot this type of teamwork in a future offering of the course.
Quantitative Model for Course Improvement
One important contribution of my work is the creation of a feedback loop in which course design itself is a system to be optimized. I define a course health index for the electric vehicle design curriculum by combining learning outcome data, student satisfaction, project quality, and instructional efficiency. The index can be written as:
$$H_{\text{course}} = \sum_{j=1}^{4} \mu_j H_j$$
where
$$H_1 = \text{average normalized student test score over electric vehicle modules},$$
$$H_2 = \text{project creativity rating normalized by a 5-point rubric},$$
$$H_3 = \text{student satisfaction and self-efficacy score},$$
$$H_4 = \text{platform engagement index}.$$
The weights \(\mu_j\) are decided by the teaching team. I use this index each semester to diagnose overall strengths and weaknesses. For example, if \(H_2\) is high but \(H_1\) is low, the course may emphasize project work without enough fundamental calculation practice. Conversely, if \(H_1\) is high and \(H_2\) is low, the course may be too drill-based and not sufficiently challenging for innovation.
From data in a recent semester, I found that the course health index improved after I introduced optional frontier resources. In particular, the project creativity score improved because students could pursue topics they found fascinating. However, the platform engagement index also showed that some students spent an excessive amount of time watching optional videos, which may have reduced the time available for structured homework. This observation indicates a need for better study-time management in future SPOC design. I plan to add a recommendation system that balances learner interest with recommended video duration.
Another useful model is a difficulty response curve for each electric vehicle course topic. For each topic, I can define the proportion of students who have mastered the prerequisite knowledge and the proportion who have completed the learning path. The likelihood that a particular student masters a target concept can be expressed as a logistic function:
$$P(\text{mastery}) = \frac{1}{1 + e^{-z}}$$
where
$$z = a_0 + a_1 x_{\text{pre}} + a_2 x_{\text{time}} + a_3 x_{\text{interaction}} + a_4 x_{\text{quality}}$$
and \(x_{\text{pre}}\) is the pre-assessment score, \(x_{\text{time}}\) is online learning time, \(x_{\text{interaction}}\) is the number of quiz attempts or forum posts, and \(x_{\text{quality}}\) is the quality of interactive behavior. Logistic regression can identify which factors matter most in the electric vehicle course. Such a model can help me predict which students are at risk before they fail a chapter test. Early warning is one of the most promising applications of artificial intelligence in higher education.
By using this model, I can move from descriptive analytics, which tells me what happened, to predictive analytics, which tells me what is likely to happen. For example, if a student in an electric vehicle design course has a very low pre-assessment score on battery equations and spends less than twenty minutes on the related video, the model may predict a probability of mastery of only 40%. I can then send the student a personalized checklist. This kind of targeted intervention would have been difficult to implement in a traditional course.
The continuous improvement of the electric vehicle course is based on both qualitative and quantitative signals. I do not rely on technology alone. I hold focus group discussions with students, paying special attention to those who are anxious about the course. I ask them to describe their mental model of how an electric vehicle works. Their drawings and explanations often reveal more than any multiple-choice test. The intensity of this reflective process is necessary if we want the curriculum itself to embody new-quality productive forces. A curriculum that cannot learn is as limited as a static engineer who refuses to update their knowledge.
Implications for Other Engineering Courses
The proposed paradigm is not limited to the electric vehicle domain. The same principles can be transferred to other engineering fields, such as robotics, renewable energy systems, aerospace systems, and intelligent manufacturing. The starting point is always a knowledge graph. The graph defines the core competence space of the discipline and reveals the relationships among concepts. Then an intelligent platform collects behavioral data as students navigate that space. The teacher uses AI and data analytics to design personalized learning paths and project tasks. Finally, output evaluation includes not only tests but also authentic performances, such as design proposals, prototypes, simulation reports, and collaborative problem-solving.
Nevertheless, the electric vehicle context is special in several ways. The first is industrial urgency. The transition to electric mobility is moving quickly, and educational institutions must produce talented engineers who are ready for this transformation. The second is multi-disciplinary complexity. The electric vehicle cannot be reduced to one engineering science. It is a true cyber-physical system that includes electrochemical energy storage, electromagnetic conversion, real-time control, thermal management, electromagnetic compatibility, and large-scale data analysis. The third is safety and social significance. An electric vehicle is not only a product; it is part of the transportation infrastructure. Its battery supply chain, charging network, and end-of-life recycling raise environmental and ethical questions. A smart course should make room for these broader societal reflections.
For students, the intelligent electric vehicle course creates both challenges and opportunities. The challenge is that they must accept a higher degree of autonomy. They cannot wait passively for a lecture to tell them everything. The opportunity is that they develop the kind of self-regulated learning behavior that is essential in a profession where technologies change every year. The knowledge graph gives them a map, the SPOC platform supports them along the path, and the teacher becomes a mentor rather than a broadcaster. In this way, new-quality productive forces, artificial intelligence, and curriculum design form a coherent whole.
Conclusion
In this first-person account, I have proposed and implemented a new intelligent curriculum paradigm for electric vehicle design courses under the guidance of new-quality productive forces. The paradigm includes digital resource construction, smart platform development, AI-enabled SPOC pedagogy, and diversified assessment. At its center is the knowledge graph of electric vehicle content, which connects battery systems, electric drive systems, electronic control systems, and complete vehicle architecture. The course platform records learning behavior, the AI tutor provides individualized feedback, and the teacher uses real-time data to adapt the course. This dynamic system enables more precise instruction, more personalized projects, and better support for innovation.
I have shown that data from the smart platform can reveal learning hotspots and weak points in the electric vehicle design curriculum. Frontier topics such as solid-state batteries and hairpin motors attract high attention, while calculations involving battery sizing, motor parameters, and powertrain matching present persistent difficulties. I use these insights to split videos into smaller modules, train AI assistants to respond to specific misconceptions, and design classroom activities that address the source of the misunderstanding. By monitoring improvement through correct-response rates, I close the loop between outcome data and course revision.
From a broader educational perspective, the new-quality productive forces view challenges us to move beyond fixed content and standardized examinations. The future of the electric vehicle industry depends on engineers who can collaborate with intelligent machines, extract insights from data, generate creative concepts, and make responsible decisions under uncertainty. My course redesign is an attempt to create such engineers. It is not a final solution but a continuing experiment. I will continue to improve the electric vehicle design course and to share the results with other educators who are interested in the intelligent transformation of higher education.
The most important message is that curriculum design itself must be intelligent. If we teach future electric vehicle designers through static methods, we fail to model the very innovation that they will be expected to produce. By integrating AI, knowledge graphs, SPOC, and diversified assessment, the intelligent course becomes not just an instrument for teaching but also a dynamic system for cultivating new-quality talent. I hope the practices described here will inspire further innovation in electric vehicle engineering education and in all fields that need to meet the demands of the changing world.