Automobile pressure sensor semiconductor device based on MEMS technology
By integrating a three-dimensional heterogeneous sensing structure and a self-calibration function into a MEMS pressure sensor, the problem of insufficient perception of multi-source stress interference and complex mechanical states in the vehicle environment is solved. This achieves high-precision and stable multi-dimensional mechanical information perception and self-diagnosis, improving the reliability and adaptability of the sensor.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- DONGGUAN TONGKE ELECTRONICS CO LTD
- Filing Date
- 2026-02-09
- Publication Date
- 2026-05-19
AI Technical Summary
Existing MEMS pressure sensors face challenges such as multi-source stress interference, environmental drift, and insufficient perception of complex mechanical states in automotive environments, making it difficult to provide high-precision and stable multi-dimensional mechanical information in intelligent vehicles.
By employing a combination design of a three-dimensional heterogeneous sensing structure, a thermally activated vacuum reference cavity, an on-chip intelligent diagnostic engine, a multi-source stress isolation layer, and an anti-interference packaging shell, the sensor achieves synchronous sensing and decoupling of vertical pressure, shear force, and bending moment. Combined with self-calibration and self-diagnostic functions, the sensor's stability and reliability are enhanced.
It significantly improves the ability to comprehensively measure complex coupled stresses, reduces the impact of signal drift and mechanical interference, enhances the self-monitoring and fault warning capabilities of sensors, and provides high-precision and highly stable real-time pressure data, providing reliable underlying perception for the safety control system of intelligent vehicles.
Smart Images
Figure CN122062832A_ABST
Abstract
Description
Technical Field
[0001] This invention proposes a semiconductor device for automotive pressure sensors based on MEMS technology, belonging to the interdisciplinary field of microelectromechanical systems (MEMS) technology and intelligent automotive electronics. Background Technology
[0002] In automotive electronic control systems, pressure sensors are core components for monitoring critical parameters such as engine manifold pressure, fuel rail pressure, tire pressure, and battery thermal management system pressure. Their performance directly affects vehicle energy efficiency, safety, and emissions control. For a long time, these sensors have primarily relied on piezoresistive or capacitive sensing principles, typically encapsulating discrete sensing elements with external signal conditioning circuitry. While this traditional structure is technologically mature, it suffers from issues such as large size, sensitivity to temperature changes, and susceptibility to output signal drift. Especially in the complex in-vehicle environment, sensors not only need to sense the pressure of the target medium but also continuously endure high-temperature vibrations from the engine, mechanical impacts from the road surface, and material aging effects from long-term operation. These non-target stresses couple into the output signal, leading to measurement distortion and decreased stability.
[0003] In recent years, advancements in microelectromechanical systems (MEMS) technology have enabled the miniaturization and chip-based fabrication of pressure sensors, significantly improving response speed and reducing power consumption. However, existing MEMS-based solutions still face significant challenges in meeting the stringent requirements of real-world automotive applications. For instance, even after packaging, sensor chips are subject to assembly stress from the substrate and housing, as well as thermal stress interference caused by temperature inhomogeneity. These multi-source stresses are difficult to fully compensate for through subsequent circuitry. Furthermore, during long-term operation, the sensor's zero point and sensitivity slowly change with temperature cycling and time, exhibiting environmental drift. This necessitates external calibration to maintain accuracy, increasing system maintenance complexity. Moreover, mainstream MEMS pressure-sensitive structures are typically optimized only for vertical pressure, lacking the ability to sense and decouple complex mechanical states such as simultaneous shear forces or bending moments. This limits their application depth in safety-critical scenarios requiring higher multi-dimensional force sensing, such as chassis control and vehicle stability. Although existing research has attempted to partially alleviate the above problems by improving algorithms or structures, how to achieve effective stress isolation and decoupling at the physical level, and endow sensors with intelligent functions of self-calibration and self-diagnosis, remains a key bottleneck that current technology needs to overcome. Summary of the Invention
[0004] This invention provides a semiconductor device for automotive pressure sensors based on MEMS technology, to solve the problems mentioned in the background section above:
[0005] This invention proposes a MEMS-based automotive pressure sensor semiconductor device, comprising: a three-dimensional heterogeneous sensing structure, a thermally activated vacuum reference cavity, an on-chip intelligent diagnostic engine, a signal conditioning circuit module, a multi-source stress isolation layer, and an anti-interference packaging shell. The three-dimensional heterogeneous sensing structure is used to simultaneously sense coupled stress, and its signal output terminal is connected to the signal input terminal of the signal conditioning circuit module. The thermally activated vacuum reference cavity is thermally coupled to the three-dimensional heterogeneous sensing structure to provide a stable vacuum reference to eliminate environmental drift. The on-chip intelligent diagnostic engine is bidirectionally connected to the signal conditioning circuit module for real-time monitoring of the sensor's operating status and fault warning. The multi-source stress isolation layer is disposed between the three-dimensional heterogeneous sensing structure and the anti-interference packaging shell to isolate non-target loads. The anti-interference packaging shell protects the internal components from external environmental interference, and its power input terminal is connected to an external power source.
[0006] The beneficial effects of this invention are as follows: Through its three-dimensional heterogeneous sensing structure, it can synchronously and independently sense vertical pressure, shear force, and bending moment, significantly improving the comprehensive measurement capability for complex coupled stresses and providing richer mechanical dimension information for systems such as chassis control and battery safety. The integrated, thermally activated vacuum reference chamber can dynamically adjust and maintain the internal pressure reference according to the ambient temperature, effectively suppressing signal drift caused by temperature gradients and long-term aging, thereby reducing the maintenance needs of traditional sensors that require frequent external calibration due to significant temperature drift. The on-chip intelligent diagnostic engine continuously monitors the sensor's own health status, providing early warnings of performance degradation or failure, greatly enhancing the predictability and safety of system operation and avoiding control function failures that may result from sudden, latent sensor failures. The unique stress isolation layer and anti-interference packaging physically block the transmission path of non-target interference such as mechanical vibration and assembly stress, reducing the direct impact of the external environment on the sensing core and ensuring the purity and stability of the measurement signal under harsh automotive conditions. Overall, the device achieves a leap from single sensing to multimodal decoupling and from passive measurement to active diagnosis. It can provide high-precision and highly stable real-time pressure data and provide underlying support for predictive maintenance and functional safety of intelligent vehicles, thus comprehensively improving its overall performance and reliability. Attached Figure Description
[0007] Figure 1 This is a schematic diagram of the device described in this invention. Detailed Implementation
[0008] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.
[0009] One embodiment of the present invention, such as Figure 1 As shown, a MEMS-based automotive pressure sensor semiconductor device includes: a three-dimensional heterogeneous sensing structure, a thermally activated vacuum reference cavity, an on-chip intelligent diagnostic engine, a signal conditioning circuit module, a multi-source stress isolation layer, and an anti-interference packaging shell. The three-dimensional heterogeneous sensing structure simultaneously senses coupled stresses, including vertical pressure, shear force, and bending moment; its signal output terminal is connected to the signal input terminal of the signal conditioning circuit module. The thermally activated vacuum reference cavity is thermally coupled to the three-dimensional heterogeneous sensing structure to provide a stable vacuum reference to eliminate environmental drift. The on-chip intelligent diagnostic engine is bidirectionally connected to the signal conditioning circuit module for real-time monitoring of the sensor's operating status and fault warning. The multi-source stress isolation layer is disposed between the three-dimensional heterogeneous sensing structure and the anti-interference packaging shell to isolate non-target loads, including mechanical shock, thermal expansion, and assembly stress. The anti-interference packaging shell protects internal components from external environmental interference; its power input terminal is connected to an external power source, and its power output terminal is connected to the power input terminals of the signal conditioning circuit module and the on-chip intelligent diagnostic engine.
[0010] The three-dimensional heterogeneous sensing structure is manufactured using microelectromechanical systems (MEMS) technology and contains multiple sensing units in different directions to achieve multimodal stress sensing.
[0011] The thermally activated vacuum reference cavity is constructed with a material with a low coefficient of thermal expansion, maintains a high vacuum state inside, and is equipped with thermally activated elements on the cavity wall to dynamically adjust the pressure inside the cavity.
[0012] The on-chip intelligent diagnostic engine includes a microprocessor, memory, and a self-diagnostic algorithm module, which is used to perform sensor status monitoring, fault mode identification, and early warning signal generation.
[0013] The signal conditioning circuit module is used to amplify, filter, and perform analog-to-digital conversion on the weak signal output by the sensitive structure.
[0014] The multi-source stress isolation layer uses a material with an elastic modulus gradient, and achieves the blocking of stress transmission paths through structural optimization;
[0015] The anti-interference packaging shell is made of metal-ceramic composite material, which has high airtightness and anti-electromagnetic interference capabilities.
[0016] The three-dimensional heterogeneous sensitive structure includes a vertical pressure-sensitive diaphragm, a shear force-sensitive beam structure, and a bending moment-sensitive frame.
[0017] A thermally activated vacuum reference cavity includes a vacuum chamber, a thermally activated thin film, and a temperature sensor.
[0018] The on-chip intelligent diagnostic engine includes a status monitoring module, a fault analysis module, and an early warning output module;
[0019] The signal conditioning circuit module includes a preamplifier, a bandpass filter, and an analog-to-digital converter;
[0020] The multi-source stress isolation layer includes a stress buffer layer and a stress blocking layer;
[0021] The anti-interference package includes a metal base, a ceramic cover, and a sealing ring.
[0022] The working principle of the above technical solution is as follows: By simultaneously sensing vertical pressure, shear force, and bending moment through a three-dimensional heterogeneous sensing structure, comprehensive decoupling and measurement of complex stresses are achieved, directly improving the richness of sensing dimensions and the accuracy of measurement results. The thermally activated vacuum reference cavity maintains stable internal pressure through dynamic adjustment, providing a reliable reference benchmark for the sensor. This effectively suppresses zero-point drift and sensitivity decay caused by temperature changes and material aging, significantly reducing the need for frequent external calibration of traditional sensors due to environmental drift. The on-chip intelligent diagnostic engine can analyze the sensor status in real time and warn of potential faults, enhancing the system's self-monitoring and early risk identification capabilities during operation, avoiding sudden failures that may result from the continuous accumulation of latent faults. The combination of a multi-source stress isolation layer and an anti-interference packaging shell physically blocks the transmission path of non-target loads such as external mechanical shocks and thermal stress, and shields against electromagnetic interference, reducing the direct impact of external interference signals on the measurement core. The overall device adopts chip-level integration, which not only greatly improves the long-term working stability and reliability of the system, but also enables it to adapt to more stringent vehicle operating conditions, providing more accurate and reliable underlying perception data for the safety control system of intelligent vehicles.
[0023] In one embodiment of the present invention, the multimodal stress sensing of the three-dimensional heterogeneous sensitive structure is achieved through the following method:
[0024] Multimodal raw stress signals are synchronously acquired through a three-dimensional heterogeneous sensitive structure to generate a mixed stress signal set; preliminary signal separation is performed on the mixed stress signal set to generate preliminary decoupled stress components.
[0025] Frequency domain and amplitude domain features are extracted from the initial decoupled stress components to generate stress feature vectors; independent component analysis is performed on the mixed stress signal set using the stress feature vectors to generate independent components of vertical pressure, shear force, and bending moment.
[0026] The independent components of vertical pressure, shear force, and bending moment are input into the signal conditioning circuit module for signal amplification and analog-to-digital conversion to generate digital multimodal stress data.
[0027] A thermally activated vacuum reference cavity is triggered based on digital multimodal stress data to perform reference pressure self-calibration and generate a drift compensation benchmark. The digital multimodal stress data is then compensated in real time using the drift compensation benchmark to generate environmental drift-compensated stress data.
[0028] Temporal stability analysis is performed on the stress data after environmental drift compensation to generate instantaneous sensor status indicators; the sensor health status is judged based on the instantaneous sensor status indicators to generate a health status label.
[0029] If the health status is normal, the stress data after environmental drift compensation is fused with the health status data to generate the final multimodal stress sensing data; if the health status is abnormal, the on-chip intelligent diagnostic engine is triggered to perform fault mode matching and generate a fault diagnosis report and warning signal.
[0030] The final multimodal stress sensing data or fault diagnosis report and early warning signal are output to an external system to complete a multimodal stress sensing operation. At the same time, the calibration parameters of the on-chip intelligent diagnostic engine are updated based on the data from this sensing operation to provide an optimization benchmark for the next sensing operation.
[0031] The working principle of the above technical solution is as follows: Through synchronous acquisition and independent component analysis, the effective separation and extraction of coupled stresses of vertical pressure, shear force, and bending moment are achieved, significantly improving the comprehensiveness and accuracy of multi-dimensional mechanical information perception under complex load environments. Digital stress data triggers reference cavity self-calibration and performs real-time drift compensation, effectively overcoming signal distortion caused by temperature changes and long-term aging, reducing measurement errors caused by environmental drift in traditional sensors, and enhancing the long-term stability of data output. Combining time-domain stability analysis to generate health status indicators enables the sensor to have online self-diagnostic capabilities, allowing for early identification of performance degradation or potential faults, reducing system safety risks caused by sudden failures. Finally, the sensed data and health status are fused and output, and parameters are optimized for the next operation, enabling the entire system to provide highly reliable real-time stress data while continuously learning and adjusting itself, thus maintaining excellent measurement consistency under harsh working conditions.
[0032] In one embodiment of the present invention, the step of extracting frequency and amplitude domain features from the initially decoupled stress components to generate a stress feature vector; and using the stress feature vector to perform independent component analysis on the mixed stress signal set to generate independent components of vertical pressure, shear force, and bending moment; includes:
[0033] Synchronous time-domain and frequency-domain analysis is performed on the initially decoupled stress components to extract the amplitude spectrum, phase spectrum and energy distribution characteristics of the signal, and generate a multi-dimensional stress feature matrix;
[0034] A feature space for decoupling operations is constructed using a multi-dimensional stress feature matrix, generating a decoupling feature space basis. Within this feature space, a projection transformation is performed on the mixed stress signal set to generate a projected mixed signal.
[0035] A blind source separation algorithm is applied to the projected mixed signal to iteratively calculate the statistical independence between each signal source and generate estimates of multiple independent signal source components.
[0036] The components of multiple independent signal sources are estimated and matched with the physical transfer function of a three-dimensional heterogeneous sensitive structure to generate physical meaning identifiers for each component.
[0037] Based on the physical meaning identifier, components corresponding to vertical pressure, shear force, and bending moment are selected from multiple independent signal source component estimates to generate independent components of vertical pressure, shear force, and bending moment.
[0038] The working principle of the above technical solution is as follows: By performing multi-dimensional feature analysis on the signal and constructing a specialized feature space, feature information representing different stress types can be extracted more precisely from the mixed original signal, laying the foundation for improving the accuracy of subsequent separation. Utilizing a blind source separation algorithm to iteratively calculate the statistical independence of the signal can effectively isolate the contributions of each independent source hidden in the mixed signal, significantly improving the ability to distinguish between coupled vertical pressure, shear force, and bending moment, and reducing the risk of perception distortion and misjudgment caused by signal crosstalk. Introducing a physical transfer function to match and verify the separation results gives each separated mathematical component a clear physical meaning, avoiding invalid or erroneous components that may be generated by relying solely on mathematical algorithms and are inconsistent with actual physical mechanisms. Ultimately, this method can reliably generate pure independent stress components, ensuring that the multi-dimensional mechanical information output by the sensor is not only dimensionally complete but also has clear data direction and does not interfere with each other, greatly enhancing the reliability and usability of the perception results under complex load scenarios.
[0039] In one embodiment of the present invention, the method for dynamically adjusting the intracavity pressure of the thermally activated element in the thermally activated vacuum reference cavity includes:
[0040] Real-time data on temperature and pressure inside the thermally activated vacuum reference cavity are collected to generate a set of cavity environmental parameters; simultaneously, data on the junction temperature of the sensor chip and the external ambient temperature are collected to generate a set of external thermal coupling parameters.
[0041] The integrated thermodynamic state vector is generated by fusing the intracavity environmental parameter set with the external thermal coupling parameter set; the integrated thermodynamic state vector is analyzed to generate the stability assessment result of the current vacuum reference state;
[0042] The target pressure setpoint inside the cavity is calculated based on the stability assessment results, and a pressure regulation command is generated; a thermal power loading strategy is generated based on the pressure regulation command and the historical operating parameters of the thermally activated element.
[0043] A thermal power loading strategy is implemented, applying a driving current with specific timing and amplitude to the thermally activated element to generate a dynamic thermal field distribution; this dynamic thermal field distribution causes microscopic deformation of the cavity wall material, thereby changing the internal sealed volume and generating a preliminary pressure regulation effect;
[0044] Monitor the rate and direction of intracavitary pressure change under the initial pressure regulation effect, and generate pressure regulation feedback data; use the pressure regulation feedback data to correct the thermal power loading strategy and generate a closed-loop optimized thermal power loading strategy.
[0045] The closed-loop optimized thermal power loading strategy is applied until the cavity pressure reaches the target pressure setting value, generating a steady-state vacuum reference. This steady-state vacuum reference is then output to provide a compensation reference for the pressure sensing of the three-dimensional heterogeneous sensitive structure, thus completing this dynamic adjustment operation.
[0046] The working principle of the above technical solution is as follows: By fusing internal and external thermodynamic parameters in real time, it ensures that the internal reference pressure remains highly stable even when the sensor itself heats up or the external ambient temperature changes, thus providing a reliable absolute benchmark for pressure sensing and significantly reducing the reference point drift problem caused by thermal expansion and contraction in traditional vacuum reference cavities. Pressure regulation commands are generated based on stability assessment results and closed-loop feedback control is implemented, making the regulation process predictable and adaptive, greatly enhancing the system's robustness against environmental disturbances. The thermal deformation of the thermally activated element is used to fine-tune the cavity volume, achieving stepless and precise adjustment of the reference pressure, avoiding the reliability risks and response delays that may arise from using moving parts such as mechanical valves. This closed-loop regulation mechanism can both quickly respond to sudden changes in external temperature and maintain a constant internal pressure over a long period, thereby fundamentally improving the consistency of the sensor's measurement accuracy across the entire temperature range.
[0047] In one embodiment of the present invention, the on-chip intelligent diagnostic method of the on-chip intelligent diagnostic engine includes:
[0048] Collect multi-dimensional sensor status data to generate a raw status dataset; preprocess the raw status dataset to generate a standardized status data sequence; extract features from the standardized status data sequence to generate a sensor health feature vector;
[0049] Pattern matching is performed on the sensor health feature vectors to generate preliminary health status classification results; based on the preliminary health status classification results, the corresponding fault diagnosis knowledge base is loaded to generate the current diagnosis rule set.
[0050] The standardized state data sequence is analyzed in real time using the current diagnostic rule set to generate a list of potential fault points; the original state dataset is then subjected to deep data mining based on the list of potential fault points to generate a fault evolution trend map.
[0051] Risk level assessment is performed based on fault evolution trend map to generate fault early warning level signal; information fusion is performed between fault early warning level signal and preliminary health status classification results to generate comprehensive diagnostic conclusion.
[0052] Based on the comprehensive diagnostic conclusion, a diagnostic decision is made. If the conclusion is healthy, the normal working status indicator and calibrated data are output. If the conclusion is faulty, the fault location engine is triggered to generate a high-precision fault location and type code.
[0053] The high-precision fault location and type code are sent to the early warning output module to generate a visual early warning report and maintenance suggestions; at the same time, the diagnosis process and conclusions are updated to the fault diagnosis knowledge base to complete the closed loop of this diagnosis learning.
[0054] The working principle of the above technical solution is as follows: By continuously collecting and analyzing multi-dimensional status data, it can observe the changes in the health of the sensors in real time, significantly improving the timeliness of early fault detection and warning, thereby reducing the safety risks to the entire vehicle system caused by the latent degradation or sudden failure of sensor performance. Its dual analysis mechanism based on rules and data mining enhances the rigor of the diagnostic logic and the ability to identify complex fault modes, reducing the possibility of false alarms or missed alarms. The generated fault evolution trend map and high-precision location code enable maintenance personnel to quickly and accurately understand the root cause and severity of the problem, avoiding the problems of vague fault location and time-consuming troubleshooting in traditional methods. While completing the diagnosis, this method can also update the knowledge base with the experience gained, enabling the system to provide timely and reliable status reports and maintenance guidance, while continuously accumulating experience and optimizing diagnostic capabilities over time, achieving a leap from passive monitoring to proactive evolution.
[0055] In one embodiment of the present invention, the step of performing real-time analysis on a standardized state data sequence using the current diagnostic rule set to generate a list of potential fault points; and combining the list of potential fault points with deep data mining of the original state dataset to generate a fault evolution trend map, includes:
[0056] Input the standardized state data sequence into the current diagnostic rule set, perform multi-condition parallel matching, and generate an initial rule matching result set;
[0057] The initial rule matching result set is used to calculate confidence and resolve conflicts, generating a set of high-confidence potential fault points;
[0058] Based on the severity and frequency of the fault, the set of high-confidence potential fault points is prioritized and a weighted list of potential fault points is generated.
[0059] Using a weighted list of potential fault points as an index, the original state dataset is sliced over time and features are re-extracted to generate fault feature data blocks at multiple time scales.
[0060] Pattern evolution analysis and correlation mining are performed on fault feature data blocks at multiple time scales to generate fault mode propagation paths and evolution probabilities.
[0061] The working principle of the above technical solution is as follows: through multi-condition parallel matching and confidence calculation, truly suspicious abnormal signals can be quickly screened from a large amount of state data, significantly improving the response speed and initial screening accuracy of real-time diagnosis, and effectively reducing the frequency of false alarms caused by data noise or instantaneous interference. After generating a set of high-confidence fault points, priority ranking is performed, allowing limited diagnostic resources to focus on the most serious and frequent potential problems, enhancing the targeting and efficiency of risk management. Furthermore, by using the priority list as an index to trace back the original data for deep time series mining, the evolution process and internal correlation of fault characteristics can be revealed, thereby constructing a trend map with predictive capabilities. This reduces passive response to faults and shifts to proactively predicting their development path and probability, enabling maintenance personnel to intervene in advance and avoiding the gradual deterioration of latent defects that eventually lead to sudden sensor failure and affect driving safety. The entire process not only completes an accurate assessment of the current health status but also provides key decision-making basis for future reliability maintenance.
[0062] In one embodiment of the present invention, the step of inputting a standardized state data sequence into the current diagnostic rule set, performing multi-condition parallel matching, and generating an initial rule matching result set includes:
[0063] Iterate through each diagnostic rule in the current diagnostic rule set, extract the logical conditions and threshold constraints in the rule premises, and generate a list of matching conditions that can be executed concurrently;
[0064] For each item in the matching condition list, concurrent query and comparison operations are performed in the standardized state data sequence to generate independent matching results for each condition;
[0065] Perform a logical AND operation on the independent matching results of each condition, combine them to form a preliminary matching item that satisfies all the premises of a single rule, and generate a preliminary matching result for a single rule;
[0066] Based on the initial matching result of a single rule, the fault confidence score corresponding to the matching result is calculated, and an initial matching record with the confidence score is generated.
[0067] Integrate the initial matching records of all diagnostic rules, remove contradictory matching records, and generate an initial rule matching result set.
[0068] The working principle of the above technical solution is as follows: By decomposing multiple conditions of the diagnostic rules into a list that can be executed concurrently and performing parallel queries, the matching speed of massive state data and complex rule bases is significantly accelerated, thereby improving the response efficiency of the real-time diagnostic system. A confidence score is calculated for each preliminary matching result, providing a quantitative basis for subsequent fault determination and enhancing the credibility and refined management capabilities of the diagnostic conclusions. During result integration, contradictory matching records are proactively removed, effectively preventing self-contradictory diagnostic conclusions caused by overlapping rule boundaries or conflicting conditions, reducing the risk of false alarms, and ensuring the logical consistency of the output results. The entire matching process has a clear structure, forming an efficient and reliable analysis pipeline from condition decomposition and parallel execution to result fusion and verification. This enables the diagnostic system to handle complex multi-rule situations and ensures that the final set of potential fault points has high quality, laying a solid foundation for subsequent in-depth analysis.
[0069] In one embodiment of the present invention, the signal conditioning circuit method of the signal conditioning circuit module includes:
[0070] The system receives raw multimodal analog electrical signals from a three-dimensional heterogeneous sensitive structure to generate an initial set of analog signals; it then performs preliminary suppression of background noise and DC bias on the initial set of analog signals to generate a preliminary conditioned analog signal.
[0071] Gain matching is performed on each channel of the pre-conditioned analog signal to generate a gain-balanced analog signal; based on the amplitude range of the gain-balanced analog signal, the amplification factor of the preamplifier is dynamically adapted to generate a high dynamic range amplified signal.
[0072] Multi-stage active filtering is performed on the high dynamic range amplified signal to generate an analog signal with out-of-band noise removed; combined with the system noise spectrum characteristics fed back by the on-chip intelligent diagnostic engine, the filter cutoff frequency is adaptively fine-tuned to generate a clean target frequency band analog signal.
[0073] High-precision sampling and analog-to-digital conversion are performed on the pure target frequency band analog signal to generate the original digital signal sequence; digital filtering and sampling point calibration are performed on the original digital signal sequence to generate a high-precision digital stress data sequence.
[0074] The high-precision digital stress data sequence is sent to the on-chip intelligent diagnostic engine for data reliability verification and signal quality identification is generated. The signal quality identification is used to backtrack and evaluate the parameters of each step of the conditioning process and generate parameter optimization records.
[0075] The high-precision digital stress data sequence with signal quality identification and parameter optimization record are encapsulated to generate the final conditioning output data packet, completing one signal conditioning operation, and the final conditioning output data packet is transmitted to the subsequent processing unit.
[0076] The working principle of the above technical solution is as follows: Through multi-stage cascaded conditioning steps, the conversion quality and reliability from weak analog signals to high-precision digital data are significantly improved. Initial suppression of background noise and DC bias effectively reduces signal baseline drift, providing a cleaner starting point for subsequent amplification and processing. Gain matching of each channel signal eliminates inherent errors introduced by minor differences in the sensitive unit's manufacturing process, improving the consistency between multi-modal data. A strategy of dynamically adapting amplifier gain avoids signal clipping or insufficient resolution, enhancing the system's ability to capture a wide range of stress variations. Adaptive filtering combined with diagnostic engine feedback optimizes the filtering effect in real time based on the noise characteristics of the current operating condition, reducing the risk of poor performance of fixed-parameter filters in complex electromagnetic environments. The final digitized data undergoes rigorous reliability verification, and the verification results are used to inversely optimize the conditioning parameters, enabling the entire conditioning loop to have self-learning and adjustment capabilities, preventing performance from gradually deteriorating over time or with environmental changes, thus maintaining high-precision and high-stability signal output over the long term.
[0077] In one embodiment of the present invention, the step of performing gain matching on each channel of the preliminary conditioning analog signal to generate a gain-balanced analog signal; and dynamically adapting the preamplifier's amplification factor based on the amplitude range of the gain-balanced analog signal to generate a high dynamic range amplified signal, includes:
[0078] Multi-channel parallel analysis is performed on the preliminary conditioning analog signal to extract the baseline amplitude and dynamic range of each channel signal and generate a channel gain characteristic table;
[0079] Using the channel with the widest dynamic range in the channel gain characteristic table as a reference, calculate the compensation gain coefficients required for the remaining channels and generate a set of channel gain compensation coefficients.
[0080] By applying a set of channel gain compensation coefficients, amplitude compensation is performed on the preliminary conditioned analog signals of each channel to generate gain-balanced analog signals with amplitude alignment between channels;
[0081] Analyze the overall amplitude distribution of the gain-balanced analog signal, identify the peak and effective value ranges of the signal, and generate statistical characteristics of the signal amplitude.
[0082] Based on the statistical characteristics of the signal amplitude, the preamplifier gain configuration table is queried to select the optimal amplification factor and generate amplifier control parameters;
[0083] The amplifier control parameters are loaded into the preamplifier to amplify the gain-balanced analog signal, generating a high dynamic range amplified signal.
[0084] The working principle of the above technical solution is as follows: By analyzing and extracting the gain characteristics of each channel in parallel, the output differences of each sensitive unit can be accurately quantified, thus providing a reliable basis for subsequent compensation and effectively improving the consistency and comparability of the amplitudes of the multiple signals output by the three-dimensional heterogeneous sensitive structure. Amplitude compensation is performed based on the channel with the widest dynamic range, ensuring that the signals of all channels are adjusted to the same "starting line," significantly reducing the inherent deviations between channels caused by manufacturing tolerances or temperature drift, and enhancing the accuracy of multimodal stress data fusion. The optimal amplification factor is dynamically selected based on the statistical characteristics of the compensated signal, ensuring adaptability to signals of different intensities. This avoids clipping distortion caused by strong signals exceeding the measurement range after amplification, and also prevents weak signals from being submerged in quantization noise due to insufficient amplification. This intelligent adaptation mechanism enables the system to maintain a high signal-to-noise ratio and high resolution under wide amplitude input, improving the overall dynamic performance and measurement accuracy of the signal conditioning process.
[0085] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.
Claims
1. A semiconductor device for automotive pressure sensors based on MEMS technology, characterized in that, The device comprises: a three-dimensional heterogeneous sensing structure, a thermally activated vacuum reference cavity, an on-chip intelligent diagnostic engine, a signal conditioning circuit module, a multi-source stress isolation layer, and an anti-interference packaging shell. The three-dimensional heterogeneous sensing structure is used to simultaneously sense coupled stress, and its signal output terminal is connected to the signal input terminal of the signal conditioning circuit module. The thermally activated vacuum reference cavity is thermally coupled to the three-dimensional heterogeneous sensing structure to provide a stable vacuum reference to eliminate environmental drift. The on-chip intelligent diagnostic engine is bidirectionally connected to the signal conditioning circuit module for real-time monitoring of the sensor's operating status and fault warning. The multi-source stress isolation layer is disposed between the three-dimensional heterogeneous sensing structure and the anti-interference packaging shell to isolate non-target loads. The anti-interference packaging shell protects internal components from external environmental interference; its power input terminal is connected to an external power supply, and its power output terminal is connected to the power input terminals of the signal conditioning circuit module and the on-chip intelligent diagnostic engine.
2. The automotive pressure sensor semiconductor device based on MEMS technology according to claim 1, characterized in that, The three-dimensional heterogeneous sensing structure is manufactured using microelectromechanical systems (MEMS) technology and contains multiple sensing units in different directions to achieve multimodal stress sensing. The thermally activated vacuum reference cavity is constructed with a material with a low coefficient of thermal expansion, maintains a high vacuum state inside, and is equipped with thermally activated elements on the cavity wall to dynamically adjust the pressure inside the cavity. The on-chip intelligent diagnostic engine includes a microprocessor, memory, and a self-diagnostic algorithm module, which is used to perform sensor status monitoring, fault mode identification, and early warning signal generation. The signal conditioning circuit module is used to amplify, filter, and perform analog-to-digital conversion on the weak signal output by the sensitive structure. The multi-source stress isolation layer uses a material with an elastic modulus gradient, and achieves the blocking of stress transmission paths through structural optimization; The anti-interference packaging shell is made of metal-ceramic composite material, which has high airtightness and anti-electromagnetic interference capabilities.
3. The automotive pressure sensor semiconductor device based on MEMS technology according to claim 2, characterized in that, The three-dimensional heterogeneous sensitive structure includes a vertical pressure-sensitive diaphragm, a shear force-sensitive beam structure, and a bending moment-sensitive frame. A thermally activated vacuum reference cavity includes a vacuum chamber, a thermally activated thin film, and a temperature sensor. The on-chip intelligent diagnostic engine includes a status monitoring module, a fault analysis module, and an early warning output module; The signal conditioning circuit module includes a preamplifier, a bandpass filter, and an analog-to-digital converter; The multi-source stress isolation layer includes a stress buffer layer and a stress blocking layer; The anti-interference package includes a metal base, a ceramic cover, and a sealing ring.
4. The automotive pressure sensor semiconductor device based on MEMS technology according to claim 2, characterized in that, The multimodal stress sensing of the three-dimensional heterogeneous sensitive structure is achieved through the following method: Multimodal raw stress signals are synchronously acquired through a three-dimensional heterogeneous sensitive structure to generate a mixed stress signal set; preliminary signal separation is performed on the mixed stress signal set to generate preliminary decoupled stress components. Frequency domain and amplitude domain features are extracted from the initially decoupled stress components to generate stress feature vectors; Independent component analysis of the mixed stress signal set is performed using stress eigenvectors to generate independent components of vertical pressure, shear force, and bending moment. The independent components of vertical pressure, shear force, and bending moment are input into the signal conditioning circuit module for signal amplification and analog-to-digital conversion to generate digital multimodal stress data. A thermally activated vacuum reference cavity is triggered based on digital multimodal stress data to perform reference pressure self-calibration and generate a drift compensation benchmark. The digital multimodal stress data is then compensated in real time using the drift compensation benchmark to generate environmental drift-compensated stress data. Time-domain stability analysis is performed on the stress data after environmental drift compensation to generate real-time state indicators of the sensor. The sensor's health status is determined based on the sensor's real-time status indicators, and a health status label is generated. If the health status is marked as normal, the stress data after environmental drift compensation will be fused with the health status to generate the final multimodal stress sensing data. If the health status is marked as abnormal, the on-chip intelligent diagnostic engine is triggered to perform fault mode matching and generate a fault diagnosis report and warning signal. The final multimodal stress sensing data or fault diagnosis report and early warning signal are output to an external system to complete a multimodal stress sensing operation. At the same time, the calibration parameters of the on-chip intelligent diagnostic engine are updated based on the data from this sensing operation to provide an optimization benchmark for the next sensing operation.
5. The automotive pressure sensor semiconductor device based on MEMS technology according to claim 4, characterized in that, The frequency domain and amplitude domain features of the initially decoupled stress components are extracted to generate a stress feature vector. Independent component analysis is performed on the mixed stress signal set using stress eigenvectors to generate independent components of vertical pressure, shear force, and bending moment; including: Synchronous time-domain and frequency-domain analysis is performed on the initially decoupled stress components to extract the amplitude spectrum, phase spectrum and energy distribution characteristics of the signal, and generate a multi-dimensional stress feature matrix; A feature space for decoupling operations is constructed using a multi-dimensional stress feature matrix, generating a decoupling feature space basis. Within this feature space, a projection transformation is performed on the mixed stress signal set to generate a projected mixed signal. A blind source separation algorithm is applied to the projected mixed signal to iteratively calculate the statistical independence between each signal source and generate estimates of multiple independent signal source components. The components of multiple independent signal sources are estimated and matched with the physical transfer function of a three-dimensional heterogeneous sensitive structure to generate physical meaning identifiers for each component. Based on the physical meaning identifier, components corresponding to vertical pressure, shear force, and bending moment are selected from multiple independent signal source component estimates to generate independent components of vertical pressure, shear force, and bending moment.
6. The automotive pressure sensor semiconductor device based on MEMS technology according to claim 2, characterized in that, The method for dynamically adjusting the intracavity pressure of the thermally activated element in the thermally activated vacuum reference cavity includes: Real-time data on temperature and pressure inside the thermally activated vacuum reference cavity are collected to generate a set of cavity environmental parameters; simultaneously, data on the junction temperature of the sensor chip and the external ambient temperature are collected to generate a set of external thermal coupling parameters. The integrated thermodynamic state vector is generated by fusing the intracavity environmental parameter set with the external thermal coupling parameter set; the integrated thermodynamic state vector is analyzed to generate the stability assessment result of the current vacuum reference state; The target pressure setpoint inside the cavity is calculated based on the stability assessment results, and a pressure regulation command is generated; a thermal power loading strategy is generated based on the pressure regulation command and the historical operating parameters of the thermally activated element. A thermal power loading strategy is implemented, applying a driving current with specific timing and amplitude to the thermally activated element to generate a dynamic thermal field distribution; this dynamic thermal field distribution causes microscopic deformation of the cavity wall material, thereby changing the internal sealed volume and generating a preliminary pressure regulation effect; Monitor the rate and direction of intracavitary pressure change under the initial pressure regulation effect, and generate pressure regulation feedback data; use the pressure regulation feedback data to correct the thermal power loading strategy and generate a closed-loop optimized thermal power loading strategy. The closed-loop optimized thermal power loading strategy is applied until the cavity pressure reaches the target pressure setting value, generating a steady-state vacuum reference. This steady-state vacuum reference is then output to provide a compensation reference for the pressure sensing of the three-dimensional heterogeneous sensitive structure, thus completing this dynamic adjustment operation.
7. The automotive pressure sensor semiconductor device based on MEMS technology according to claim 1, characterized in that, The on-chip intelligent diagnostic method of the on-chip intelligent diagnostic engine includes: Collect multi-dimensional sensor status data to generate a raw status dataset; preprocess the raw status dataset to generate a standardized status data sequence; extract features from the standardized status data sequence to generate a sensor health feature vector; Pattern matching is performed on the sensor health feature vectors to generate preliminary health status classification results; based on the preliminary health status classification results, the corresponding fault diagnosis knowledge base is loaded to generate the current diagnosis rule set. The standardized state data sequence is analyzed in real time using the current diagnostic rule set to generate a list of potential fault points; the original state dataset is then subjected to deep data mining based on the list of potential fault points to generate a fault evolution trend map. Risk level assessment is performed based on fault evolution trend map to generate fault early warning level signal; information fusion is performed between fault early warning level signal and preliminary health status classification results to generate comprehensive diagnostic conclusion. Based on the comprehensive diagnostic conclusion, a diagnostic decision is made. If the conclusion is healthy, the normal working status indicator and calibrated data are output. If the conclusion is faulty, the fault location engine is triggered to generate a high-precision fault location and type code. The high-precision fault location and type code are sent to the early warning output module to generate a visual early warning report and maintenance suggestions; at the same time, the diagnosis process and conclusions are updated to the fault diagnosis knowledge base to complete the closed loop of this diagnosis learning.
8. The automotive pressure sensor semiconductor device based on MEMS technology according to claim 7, characterized in that, The standardized state data sequence is analyzed in real time using the current diagnostic rule set to generate a list of potential fault points; Deep data mining is performed on the original state dataset using a list of potential failure points to generate a failure evolution trend map; including: Input the standardized state data sequence into the current diagnostic rule set, perform multi-condition parallel matching, and generate an initial rule matching result set; The initial rule matching result set is used to calculate confidence and resolve conflicts, generating a set of high-confidence potential fault points; Based on the severity and frequency of the fault, the set of high-confidence potential fault points is prioritized and a weighted list of potential fault points is generated. Using a weighted list of potential fault points as an index, the original state dataset is sliced over time and features are re-extracted to generate fault feature data blocks at multiple time scales. Pattern evolution analysis and correlation mining are performed on fault feature data blocks at multiple time scales to generate fault mode propagation paths and evolution probabilities.
9. The automotive pressure sensor semiconductor device based on MEMS technology according to claim 8, characterized in that, The step of inputting the standardized state data sequence into the current diagnostic rule set, performing multi-condition parallel matching, and generating an initial rule matching result set includes: Iterate through each diagnostic rule in the current diagnostic rule set, extract the logical conditions and threshold constraints in the rule premises, and generate a list of matching conditions that can be executed concurrently; For each item in the matching condition list, concurrent query and comparison operations are performed in the standardized state data sequence to generate independent matching results for each condition; Perform a logical AND operation on the independent matching results of each condition, combine them to form a preliminary matching item that satisfies all the premises of a single rule, and generate a preliminary matching result for a single rule; Based on the initial matching result of a single rule, the fault confidence score corresponding to the matching result is calculated, and an initial matching record with the confidence score is generated. Integrate the initial matching records of all diagnostic rules, remove contradictory matching records, and generate an initial rule matching result set.
10. The automotive pressure sensor semiconductor device based on MEMS technology according to claim 1, characterized in that, The signal conditioning circuit method of the signal conditioning circuit module includes: The system receives raw multimodal analog electrical signals from a three-dimensional heterogeneous sensitive structure to generate an initial set of analog signals; it then performs preliminary suppression of background noise and DC bias on the initial set of analog signals to generate a preliminary conditioned analog signal. Gain matching is performed on each channel of the pre-conditioned analog signal to generate a gain-balanced analog signal; based on the amplitude range of the gain-balanced analog signal, the amplification factor of the preamplifier is dynamically adapted to generate a high dynamic range amplified signal. Multi-stage active filtering is performed on the high dynamic range amplified signal to generate an analog signal with out-of-band noise removed; combined with the system noise spectrum characteristics fed back by the on-chip intelligent diagnostic engine, the filter cutoff frequency is adaptively fine-tuned to generate a clean target frequency band analog signal. High-precision sampling and analog-to-digital conversion are performed on the pure target frequency band analog signal to generate the original digital signal sequence; digital filtering and sampling point calibration are performed on the original digital signal sequence to generate a high-precision digital stress data sequence. The high-precision digital stress data sequence is sent to the on-chip intelligent diagnostic engine for data reliability verification and signal quality identification is generated. The signal quality identification is used to backtrack and evaluate the parameters of each step of the conditioning process and generate parameter optimization records. The high-precision digital stress data sequence with signal quality identification and parameter optimization record are encapsulated to generate the final conditioning output data packet, completing one signal conditioning operation, and the final conditioning output data packet is transmitted to the subsequent processing unit.