Intelligent real-time monitoring and warning system for health state of electronic and electrical parts of automobile

Abstract

The invention discloses an intelligent real-time monitoring and warning system for the health state of electronic and electrical parts of an automobile. The intelligent real-time monitoring and warning system comprises a temperature sensor, a data acquisition module, a data expansion module, an LSTM prediction module, a storage module and a display and early warning module. According to the method, a high-temperature acceleration test does not need to be additionally carried out, large-scale expansion can be realized by adjusting original data in a small range through the Arrhenius model, and the cost is hardly increased. The method is suitable for various temperature-sensitive elements such as an automobile reversing radar and an electronic sensor, and can meet life monitoring requirements of different engineering scenes. And meanwhile, online real-time data acquisition and life prediction are supported, and an early warning function is matched, so that the equipment failure risk can be avoided in time.

Description

Technical Field

[0001] This invention relates to the interdisciplinary field of machine learning, life monitoring, and reliability engineering, and specifically to a real-time health status monitoring and alarm system for intelligent automotive electronic and electrical components. Background Technology

[0002] The health status of temperature-sensitive electronic and electrical components (such as automotive reversing radars and electronic sensors) is crucial to ensuring the safe operation of equipment. However, lifespan data for these components mainly relies on high-temperature accelerated testing, which has limitations such as being time-consuming, costly, and highly destructive. This results in very little raw training data that can be obtained and cannot reflect actual temperature conditions.

[0003] As an effective tool for time-series data prediction, the prediction accuracy of LSTM (Long Short-Term Memory) networks highly depends on large-scale, high-quality training data. However, existing technologies suffer from insufficient training of LSTM networks due to a lack of sufficient lifetime data, resulting in large prediction errors and weak generalization ability.

[0004] Furthermore, traditional lifetime prediction methods often rely on empirical estimations or single models, failing to balance data efficiency and prediction accuracy. This makes it difficult to meet the demands of real-time, accurate lifetime prediction for temperature-sensitive components in engineering scenarios. Therefore, there is an urgent need for a technology that combines a low-cost, efficient data augmentation scheme with a high-precision prediction model to address the pain points of existing reliability technologies. Summary of the Invention

[0005] To address the technical problems mentioned above, this invention provides a real-time health status monitoring and alarm system for intelligent automotive electronic and electrical components, comprising:

[0006] Temperature sensors are used to collect surface temperature signals of electronic and electrical components in real time and convert them into electrical signals.

[0007] The data acquisition module, connected to the temperature sensor, is used to convert electrical signals into digital temperature sequences, i.e., raw data.

[0008] A data augmentation module, connected to the data acquisition module, is used to receive the digital temperature sequence and the corresponding raw remaining life data, and to augment the raw data according to the Arrhenius model to generate an augmented dataset.

[0009] The LSTM prediction module, connected to the data augmentation module, is used to receive the augmented dataset for model training and to predict the remaining lifespan of the real-time acquired digital temperature sequence based on the trained model.

[0010] The storage module is connected to the data acquisition module, the data expansion module and the LSTM prediction module respectively, and is used to store the original data, the expanded data, the model parameters and the prediction results;

[0011] The display and warning module is connected to the LSTM prediction module and is used to display the predicted remaining lifetime in real time and trigger a warning when the remaining lifetime is lower than a preset threshold.

[0012] Preferably, the data acquisition module includes:

[0013] The signal conditioning unit, connected to the temperature sensor, is used to amplify, filter, and adjust the level of the electrical signal, and output a positive voltage signal that is compatible with the input range of the analog-to-digital converter.

[0014] An analog-to-digital converter, connected to the signal conditioning unit, is used to convert the positive voltage signal into a digital temperature sequence.

[0015] Preferably, the signal conditioning unit amplifies the electrical signal through an adjustable gain instrumentation amplifier, filters it through a second-order active low-pass filter, and then converts the two-ended differential signal into a single-ended positive voltage signal through a level adjustment circuit.

[0016] Preferably, the workflow of the data expansion module includes:

[0017] The original digital temperature sequence is filtered by moving average to obtain the filtered temperature sequence.

[0018] The filtered temperature sequence and its corresponding remaining lifetime are standardized.

[0019] Based on the standardized temperature sequence, multiple extended temperature sequences are generated by introducing random perturbation temperatures and using temperature constraint coefficients to limit the fluctuation range.

[0020] Based on the Arrhenius formula and the aforementioned expansion temperature sequence, the corresponding remaining expansion lifetime is calculated.

[0021] Preferably, after generating the expanded dataset, the data augmentation module also performs a rationality screening, and the screening rules include:

[0022] The extended temperature is within the sensor's measurement range, the remaining extended lifetime is positive and does not exceed the set upper limit of the multiple, and the similarity between the extended temperature sequence and the original temperature sequence is not lower than the set threshold.

[0023] Preferably, the storage module includes a static random access memory for caching real-time acquired data, and the display and warning module includes a liquid crystal display screen and an audible and visual alarm.

[0024] Preferably, the data acquisition module, data expansion module, and LSTM prediction module are integrated into an embedded processor, which communicates with the storage module and the display and warning module via a serial bus.

[0025] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0026] 1. Low data expansion cost: No additional high-temperature accelerated testing is required. Large-scale expansion can be achieved by making minor adjustments to the original data using the Arrhenius model, with almost no additional cost.

[0027] 2. High prediction accuracy: The LSTM network is trained on an expanded high-quality dataset, with prediction error controlled within 1.5 times, RMSE less than 0.1, and good fitting effect.

[0028] 3. Wide range of applications: Applicable to various temperature-sensitive components such as automotive reversing radar and electronic sensors, and can meet the life monitoring needs of different engineering scenarios.

[0029] 4. Strong real-time performance: Supports online real-time data collection and lifespan prediction. Combined with early warning functions, it can promptly avoid the risk of equipment failure. Attached Figure Description

[0030] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0031] Figure 1 This is a schematic diagram of model iteration in an embodiment of the present invention. Detailed Implementation

[0032] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0033] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0034] Example 1

[0035] This embodiment provides a real-time health status monitoring and alarm system for intelligent automotive electronic and electrical components, including: a temperature sensor, a data acquisition module, a data expansion module, an LSTM prediction module, a storage module, and a display and early warning module.

[0036] The following will describe in detail, with reference to this embodiment, how the present invention solves the technical problems in practical work.

[0037] First, the temperature sensor is mounted on the surface of temperature-sensitive components, such as reversing radars, sensors, circuit boards, and vehicle lights, to collect temperature signals in real time and convert them into electrical signals. The output voltage of the temperature sensor is linearly related to the measured temperature, as shown in the core formula below:

[0038]

[0039] in, The differential voltage output by the temperature sensor: K T V represents the voltage-temperature coefficient of the temperature sensor, determined by the sensor model; T represents the measured temperature of the sensor; T0 represents the zero-point temperature of the sensor, typically 0; V offset This is the sensor zero-point offset voltage, which is the factory calibration value.

[0040] Next, the data acquisition module conditions and converts the electrical signal from the temperature sensor from analog to digital to generate digital temperature data, i.e., the raw data. The process includes:

[0041] The instrumentation amplifier uses an INA128 model, and its amplification factor can be adjusted via an external resistor.

[0042]

[0043] Where G is the amplification factor of the instrumentation amplifier; ;R g This is an external gain adjustment resistor, adjusted according to the amplitude of the sensor output signal.

[0044] A second-order active low-pass filter (operational amplifier + RC network) is used, with the following transfer function:

[0045]

[0046] in, Complex frequency domain transfer function of the filter circuit; R is the filter resistor; C is the filter capacitor; cutoff frequency. The bandwidth is set according to the temperature signal.

[0047] Differential amplification of the dual-ended signal Convert to single-ended positive voltage Adaptable ADC input range:

[0048]

[0049] in, This is the final conditioned output voltage; This is the amplified differential voltage; The reference voltage is taken as the midpoint of the ADC input range.

[0050] The core methods of this module, such as signal buffering, differential amplification, low-pass filtering, and level adjustment, are all existing mature technologies. The improvements are as follows: for the linear output characteristics of temperature sensors, the matching relationship between the amplification factor and the filter cutoff frequency has been optimized. Through the combination of a reference voltage source and a programmable resistor, the precise calibration of the conditioning parameters has been achieved, which is suitable for the low-speed, high-precision signal acquisition requirements of temperature-sensitive components.

[0051] The data augmentation module generates an augmented dataset based on the original data. A temperature sensor collects temperature signals, which are then conditioned and converted by an ADC to obtain a digital temperature sequence.

[0052]

[0053] in, This represents a digital temperature sequence (t=1,2,3,...).

[0054] Accelerated aging tests were conducted to record the total time it took for the component to progress from its initial state (1) to its failure state (0). And by matching the temperature sequence according to the time step, the original sample pairs are obtained. .in, The remaining lifespan at any given time.

[0055] The original dataset is defined as follows:

[0056]

[0057] Where n is the number of samples; This represents the remaining lifetime at the time of the nth sample.

[0058] The data processing steps for the original dataset include:

[0059] (1) Sliding filter:

[0060]

[0061] The signal is the filtered digital temperature signal, and N is the size of the sliding window. Raw temperature signal.

[0062] (2) Standardization:

[0063]

[0064] Where the left side of the equation represents the standardized value, For the smallest digital temperature signal, This is the maximum digital temperature signal. This is the true maximum lifetime, i.e., the lifetime in the initial state.

[0065] Temperature at each time step in the original sensor temperature sequence (k is the time step), generate m expansion temperatures according to the following formula, where m is the expansion amount for each original temperature, ranging from 10 to 15):

[0066]

[0067] For the k-th time step, the p-th expansion temperature ( ); The original sensor temperature at the k-th time step is given by... Obtained by inverse transformation; For random perturbation temperature; k T This is a temperature constraint coefficient that limits fluctuations and prevents the measurement from exceeding the specified range.

[0068] The extended lifetime perturbation is derived from the Arrhenius formula:

[0069]

[0070] Where k is the reaction rate; A is the pre-exponential factor; E a is the activation energy; R is the gas constant; T is the absolute temperature.

[0071] The formula for the remaining lifetime L at the reference temperature is:

[0072]

[0073] Among them, t ri Real-time temperature spectrum of components; T is the time interval; R For reference temperature; .

[0074] Finally, all samples were integrated into an expanded dataset:

[0075]

[0076] Where w is the total number of samples; T exp Represents the absolute temperature spectrum; L exp This represents the corresponding remaining lifetime spectrum.

[0077] Using the above formula, without changing the core patterns of the original data, a large amount of expanded data that conforms to the component life decay pattern is generated by adjusting the range of temperature parameters (with small fluctuations), thus constructing a large-scale training dataset.

[0078] To avoid invalid augmentation samples affecting model training, a reasonable selection process is performed, and the original datasets are ultimately merged to form the training dataset.

[0079] (1) Filtering rules

[0080] Temperature validity: The extended temperature must be within the sensor's measurement range; otherwise, it will be rejected.

[0081] Lifespan rationality: The remaining lifespan must be a positive number and not exceed 1.5 times the maximum lifespan (to avoid generating lifespan samples that exceed the common sense of physics).

[0082] Similarity constraint: The temperature sequence similarity between the expanded sample and the original sample (using Pearson correlation coefficient) is ≥0.85, ensuring that the expanded data does not deviate from the core pattern of the original data.

[0083] (2) Expanding the dataset

[0084] The expanded dataset after filtering is merged with the original dataset to obtain the total training dataset, which is then divided into the following parts in a 7:2:1 ratio:

[0085] Training set: used for learning LSTM network parameters;

[0086] Validation set: used for model hyperparameter tuning and overfitting monitoring during training;

[0087] Test set: Used for final model performance verification.

[0088] Furthermore, the LSTM prediction module is used to receive real-time temperature data and augmented datasets to complete model training and remaining lifespan prediction. The specific structure of the LSTM prediction module in this embodiment is shown in Table 1.

[0089] Table 1

[0090] .

[0091] The LSTM unit, through the synergistic action of the forget gate, input gate, cell state, and output gate, performs feature extraction and long-term dependency modeling on time-series temperature data acquired by the sensor. The core computational formula used is the classic basic formula of the LSTM network, without additional modifications, ensuring the stability and adaptability of the model structure. The iteration curve of the model in this embodiment is shown below. Figure 1 As shown.

[0092] The storage module in this embodiment is used to store raw data, expanded data, training parameters, and prediction results. The display and warning module is used to determine the life safety factor based on the vehicle reliability decomposition index, display the remaining life in real time, and trigger a warning when the life is lower than the threshold.

[0093] Example 2

[0094] This embodiment provides the application scenario and operation process for the present invention.

[0095] (a) Hardware deployment

[0096] 1. Install a high-precision temperature sensor on the surface of temperature-sensitive components (such as car reversing radar), and connect the sensor output to the data acquisition module.

[0097] 2. The data acquisition module uses a 16-bit analog-to-digital converter with a sampling frequency of 200kHz to ensure the accuracy and real-time performance of temperature data acquisition.

[0098] 3. The data augmentation module and the LSTM prediction module are deployed in an embedded processor (such as a DSP or FPGA). The processor communicates with the data acquisition module, storage module, and display module via a serial bus.

[0099] 4. The storage module uses SRAM to cache raw data, expand the dataset, and network training parameters; the display and early warning module uses an LCD screen and an audible and visual alarm to display the prediction results in real time and trigger early warnings.

[0100] (II) Software Implementation Steps

[0101] 1. Raw Data Acquisition: Continuously collect component operating temperature data using temperature sensors, and simultaneously record the actual remaining lifespan of the components, forming a raw dataset containing 10-20 sets of data. Alternatively, raw data can be obtained through accelerated lifespan experiments or other methods.

[0102] 2. Data augmentation: Based on the Arrhenius model, the temperature fluctuation range was set to ±5K. The original data was augmented using a formula to generate 2000-3000 sets of augmented data, which were then merged with the original data to form the training dataset.

[0103] 3. LSTM Network Construction: Build an LSTM network with 3 hidden layers, with an input dimension of 1 (temperature data) and an output dimension of 1 (remaining lifetime). The number of hidden layer units is 64, the activation function is ReLU, and the optimizer is Adam.

[0104] 4. Model Training: Divide the merged dataset into training and test sets in a 7:3 ratio, set the number of iterations to 1550, and the batch size to 32. Monitor the loss function in real time during training until the loss converges (RMSE less than 0.1).

[0105] 5. Real-time prediction: After training, the system receives data collected by the temperature sensor in real time, inputs it into the LSTM network to obtain the remaining lifetime prediction value, and displays it through the display module. When the predicted lifetime is lower than the preset threshold (such as 10% of the total lifetime), the audible and visual alarm is triggered to warn.

[0106] (III) Performance Verification

[0107] A car reversing radar was selected as the test object. The original data consisted of 150 sets, which were expanded to 2500 sets using the Arrhenius model. The LSTM network training took 50 minutes and 25 seconds and converged after 1550 iterations. Four sets of samples were randomly selected for validation. The model output fit the original data well, the detail error was within an acceptable range, and the predicted lifespan error was controlled within 1.5 times, meeting the requirements of engineering applications.

[0108] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A real-time monitoring and alarm system for the health status of intelligent automotive electronic and electrical components, characterized in that, include: Temperature sensors are used to collect surface temperature signals of electronic and electrical components in real time and convert them into electrical signals. The data acquisition module, connected to the temperature sensor, is used to convert electrical signals into digital temperature sequences, i.e., raw data. A data augmentation module, connected to the data acquisition module, is used to receive the digital temperature sequence and the corresponding raw remaining life data, and to augment the raw data according to the Arrhenius model to generate an augmented dataset. The LSTM prediction module, connected to the data augmentation module, is used to receive the augmented dataset for model training and to predict the remaining lifespan of the real-time acquired digital temperature sequence based on the trained model. The storage module is connected to the data acquisition module, the data expansion module and the LSTM prediction module respectively, and is used to store the original data, the expanded data, the model parameters and the prediction results; The display and warning module is connected to the LSTM prediction module and is used to display the predicted remaining lifetime in real time and trigger a warning when the remaining lifetime is lower than a preset threshold.

2. The real-time monitoring and alarm system for the health status of intelligent automotive electronic and electrical components according to claim 1, characterized in that, The data acquisition module includes: The signal conditioning unit, connected to the temperature sensor, is used to amplify, filter, and adjust the level of the electrical signal, and output a positive voltage signal that is compatible with the input range of the analog-to-digital converter. An analog-to-digital converter, connected to the signal conditioning unit, is used to convert the positive voltage signal into a digital temperature sequence.

3. The real-time monitoring and alarm system for the health status of intelligent automotive electronic and electrical components according to claim 2, characterized in that, The signal conditioning unit amplifies the electrical signal through an adjustable gain instrumentation amplifier, filters it through a second-order active low-pass filter, and then converts the two-ended differential signal into a single-ended positive voltage signal through a level adjustment circuit.

4. The real-time health status monitoring and alarm system for intelligent automotive electronic and electrical components according to claim 1, characterized in that, The workflow of the data augmentation module includes: The original digital temperature sequence is filtered by moving average to obtain the filtered temperature sequence. The filtered temperature sequence and its corresponding remaining lifetime are standardized. Based on the standardized temperature sequence, multiple extended temperature sequences are generated by introducing random perturbation temperatures and using temperature constraint coefficients to limit the fluctuation range. Based on the Arrhenius formula and the aforementioned expansion temperature sequence, the corresponding remaining expansion lifetime is calculated.

5. The real-time monitoring and alarm system for the health status of intelligent automotive electronic and electrical components according to claim 4, characterized in that, After generating the expanded dataset, the data augmentation module also performs a rationality screening, and the screening rules include: The extended temperature is within the sensor's measurement range, the remaining extended lifetime is positive and does not exceed the set upper limit of the multiple, and the similarity between the extended temperature sequence and the original temperature sequence is not lower than the set threshold.

6. The real-time monitoring and alarm system for the health status of intelligent automotive electronic and electrical components according to claim 1, characterized in that, The storage module includes a static random access memory for caching real-time acquired data, and the display and warning module includes an LCD screen and an audible and visual alarm.

7. The real-time monitoring and alarm system for the health status of intelligent automotive electronic and electrical components according to claim 1, characterized in that, The data acquisition module, data expansion module, and LSTM prediction module are integrated into an embedded processor, which communicates with the storage module and the display and warning module via a serial bus.

Images