Method for processing absolute position sensor signal

Abstract

The invention discloses a method for processing signals of an absolute position sensor, which relates to the technical field of signal processing, and comprises the following steps of: constructing a simplified sensor model: constructing in a manner of dividing a rotation process according to the number of teeth of a sensor gear; similarity evaluation: the similarity evaluation is used for determining the similarity between the sensor and the model to obtain accurate position information, and comprises two parts of difference calculation and maximum similarity determination; the position calculation step comprises two aspects of memory use optimization and calculation complexity optimization; the method tolerance and precision evaluation is realized by calculating the similarity of signal values of the sensor at different angles; according to the method, the unique signal combination of each position of the sensor is focused, the list containing all possible output combinations and corresponding positions thereof is constructed, the actual position is determined by evaluating the similarity between the current measurement data of the sensor and the model, the problem that a traditional method is sensitive to the characteristic change of the sensor is solved, and the reliability and precision of position calculation are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of signal processing, and particularly to a method for processing signals of an absolute position sensor. Background Art

[0002] Motors are widely used in various fields, and their control accuracy highly depends on precise position sensor signal processing. This article focuses on vernier absolute position sensors. Such sensors convert position or rotation angle into electrical signals based on the vernier principle, and their basic structure includes a sensor gear and a reading head. The difference in the number of teeth of the sensor gear (N1 and N2, and N2 = N1 - 1) is the core factor for achieving high resolution. The permanent magnet in the reading head generates a magnetic field, and the magnetoresistive element changes its resistance according to the change in magnetic field strength, thereby generating a voltage signal related to the rotation angle (each magnetoresistive element is connected in a Wheatstone bridge as Figure 1 shown). In actual application scenarios, the sensor is subject to interference from various error sources. From the dimension of error duration, permanent errors result from manufacturing tolerances and installation deviations, causing tooth profile distortion, signal average value shift, and phase shift; long-term errors are mainly caused by the aging of reading head components and temperature changes. Among them, the influence of temperature on the resistance of magnetoresistive elements is significant. Although sensor manufacturers have different views on this, existing research has confirmed that temperature will cause resistance drift; short-term errors are closely related to the misalignment problem between the rotating shaft of the mechanical device and the sensor gear, which will cause the movement of the sensor gear, resulting in a change in magnetic field strength, and ultimately causing changes in signal amplitude, frequency, and phase within a single rotation period.

[0003] Currently, the conventional method for evaluating such sensor signals follows a specific process. First, the sensor signal is amplified and A / D converted to enter the digital domain; then, the signal curve is corrected to eliminate errors such as amplitude, average value, and phase shift. The correction coefficients are determined in the initial calibration stage and updated in a timely manner according to the measurement data; subsequently, interpolation operations are performed using trigonometric functions or phase relationships to calculate the angle within the period; finally, the rotation angle of the sensor is calculated based on the vernier principle, and the phase error is corrected by means of a look-up table. However, these traditional methods have many drawbacks. For example, when using trigonometric interpolation, the position accuracy is restricted by the resolution of the input data and the resolution of the operation result, and is extremely sensitive to sensor signal errors. Once a signal value is measured incorrectly, the calculated position will be completely wrong; when using phase relationship interpolation, the calculated position and error are affected by dynamic factors such as rotation speed and angular acceleration, and the robustness to changes in sensor parameters is poor. The position error is proportional to the correction coefficient error, and is even larger than the error of the trigonometric interpolation method. Therefore, a method for processing signals of an absolute position sensor is needed to solve the above problems. Summary of the Invention

[0004] The object of the present invention is to provide a method for processing the signals of an absolute position sensor to solve the problems existing in the prior art as mentioned in the above background art.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A method for processing the signals of an absolute position sensor, comprising the following steps:

[0007] S1: Construct a simplified sensor model: It is constructed in a way that divides the rotation process based on the number of teeth of the sensor gear.

[0008] S2: Similarity evaluation: Similarity evaluation is used to determine the similarity between the sensor and the model to obtain accurate position information, including two parts: difference calculation and determination of the maximum similarity.

[0009] S3: Calculate the position step: It includes two aspects: optimization of memory usage and optimization of computational complexity.

[0010] S4: Method tolerance and accuracy evaluation: It is achieved by calculating the similarity of the signal values of the sensor at different angles.

[0011] Preferably, the specific steps of S1 are as follows:

[0012] For a sensor gear with the number of teeth N1 and N2, the signal sin_N1 sw and cos_N1 sw correspond to a segment length of 360 0 / N1, and the signal sin_N2 sw and cos_N2 sw correspond to a segment length of 360 0 / N2. Through this natural division, the description of the sensor signal is simplified from a complex system of equations to the form shown in formula (1), where the coefficients appear as functions of the number of periods calculated according to formula (2);

[0013]

[0014]

[0015] Where:

[0016] is the rotation angle of the virtual wheel, and its value range is (0°, 360°) or (0rad, 2πrad);

[0017] N1 and N2 are the number of periods / number of teeth in one full rotation;

[0018] P sin_N1 ,P cos_N1 ,P sin_N2and P cos_N2 is the current cycle number of the sensor model calculated according to formula (3) on each curve;

[0019] is the sensor model value at the position ;

[0020] sin() is the sine trigonometric function;

[0021] g vw,sin_N1 (P N1 ), g vw,cos_N1 (P N1 ), g vw,sin_N2 (P N2 ), g vw,cos_N2 (P N2 ) are the amplitudes of the respective signals, and the amplitude is constant within a given period;

[0022] p vw,sin_N1 (P N1 ), p vw,cos_N1 (P N1 ), p vw,sin_N2 (P N2 ), p vw,cos_N2 (P N2 ) are the phase characteristics of the respective signals within a given period;

[0023] o vw,sin_N1 (P N1 ), o vw,cos_N1 (P N1 ), o vw,sin_N2 (P N2 ), o vw,cos_N2 (P N2 ) are the averages of the respective signals, and the average is constant within a given period;

[0024] h is the index of the higher harmonics, h > 1;

[0025] g vw,sin_N1 (P N1 )[h], g vw,cos_N1 (P N1 )[h], g vw,sin_N2 (P N2 )[h], g vw,cos_N2 (P N2 )[h] are the amplitudes of the respective higher harmonics related to the period;

[0026] p vw,sin_N1 (P N1 )[h], p vw,cos_N1 (P N1 )[h], p vw,sin_N2 (PN2 )[h], p vw,cos_N2 (P N2 )[h] is the phase shift of each high - order harmonic related to the start of the period;

[0027]

[0028] Preferably, the specific steps of S2 are as follows:

[0029] When calculating the similarity, first calculate the partial differences based on the four signals of the sensor and the corresponding model signals. Assume that the signal value of the sensor at the position is and the signal value of the model at the position is sin_N1 vw , cos_N1 vw , sin_N2 vw , cos_N2 vw , then the partial difference calculation formula is:

[0030]

[0031] Thus, a set of four partial difference values is obtained. The curves generated by them change with the rotation of the sensor gear, and their average value is determined by the current value of the sensor;

[0032] The similarity is calculated by the method of the sum of absolute differences or the sum of squared differences:

[0033] Sum of absolute differences: The calculation of similarity is transformed into finding the sum of the distances between the sensor and the model signal values. Its difference function at , in the ideal case, the sensor and the model values are the same, so the partial difference value is zero and the sum is zero. And because the absolute value is non - negative, this is a local minimum;

[0034] Sum of squared differences: Based on the concept of distance in multi - dimensional space, the similarity is calculated using the Pythagorean theorem. Its difference function is similar to the sum of absolute differences and has a global minimum at .

[0035] Preferably, the specific steps of the memory usage optimization in S3 are as follows:

[0036] In terms of memory usage optimization, in each step, only for the specific rotation angle Generate the corresponding model value according to formula (1) and (2), and then use the selected similarity calculation method to combine the current sensor measurement value with the model value to calculate the difference similarity. After the calculation is completed, compare this similarity with the previously stored minimum difference value. If the newly calculated similarity is smaller, update the minimum difference value and the corresponding position information. If not, discard the current calculated value and continue to generate the model value of the next rotation angle for calculation.

[0037] Preferably, the specific steps of optimizing the computational complexity in S3 are:

[0038] In terms of computational complexity optimization, it is achieved by combining the full search mode EM and the fast search mode FM. The full search mode is used when the system is started or when the position of the mechanism cannot be determined for some reason. In this mode, the entire sensor model needs to be traversed. Although the calculation time is long, it can ensure that the accurate position is found in the initial stage or under abnormal conditions, providing a basis for subsequent fast searches; during normal operation, the fast search mode is used. Since the rotation speed of the mechanism is limited when it runs in a closed loop, the maximum angle that the mechanism may rotate within a calculation cycle can be pre-determined according to its maximum rotation speed, thereby setting a suitable search area; in the fast search mode, the sensor model is calculated only within the set limited search area, which significantly shortens the time to calculate the new position. The calculation cycle T FM , maximum rotation speed ω max The formula ω must be satisfied between the size of the search area max T FM <The size of the search area is used to ensure that the position information can be updated promptly and accurately during the rotation of the mechanism.

[0039] Preferably, the specific steps of S4 are:

[0040] To determine the tolerance, the similarity of the sensor signal values ​​at different angles is calculated using the following formula:

[0041]

[0042] in, and The two angles are separated by the middle length of the cycle. After calculating the similarity, the difference value of the sensor curve in adjacent cycles can be determined through normalization.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] The present invention focuses on the unique signal combination at each position of the sensor, constructs a "list" containing all possible output combinations and their corresponding positions, determines the actual position by evaluating the similarity between the current measurement data of the sensor and the model, overcomes the problem that the traditional method is sensitive to changes in sensor characteristics, improves the reliability and accuracy of position calculation, and is implemented by adapting to the FPGA circuit. Description of the Drawings

[0045] Figure 1 It is the equivalent circuit schematic diagram of the reading head of the cursor type absolute position sensor.

[0046] Figure 2 It is the schematic diagram of the mode switching of the numerical method of the present invention. Detailed Embodiments

[0047] In order to make the technical means, creative features, achieved purposes and effects of the present invention easy to understand, the present invention will be further described below in conjunction with specific embodiments.

[0048] Please refer to Figure 1-2 , the present invention provides the following technical solutions:

[0049] A method for processing the signals of an absolute position sensor, comprising the following steps:

[0050] S1: Construct a simplified model of the sensor: It is constructed by dividing the rotation process based on the number of teeth of the sensor gear; Although the ideal model can accurately reproduce the sensor curve, it has high memory requirements and is difficult to update; The complete model has an advantage in accuracy but is too complex and consumes a huge amount of memory; The simplified model balances calculation and memory, describes the signal by dividing the rotation section according to the number of teeth of the sensor gear, and although its coefficient update method has limitations, it is feasible and provides data support for subsequent calculations. Its accuracy is affected by changes in sensor parameters, which in turn is related to the accuracy of the overall method.

[0051] The specific steps of the said S1 are:

[0052] For a sensor gear with the number of teeth N1 and N2, the signal sin_N1 sw and cos_N1 sw corresponding to a segment length of 360 0 / N1, the signal sin_N2 sw and cos_N2 sw corresponding to a segment length of 360 0 / N2. Through this natural division, the description of the sensor signal is simplified from a complex system of equations to the form shown in formula (1), where the coefficients appear as functions of the number of cycles calculated according to formula (2);

[0053]

[0054] Wherein:

[0055] is the rotation angle of the virtual wheel, and its value range is (0°, 360°) or (0rad, 2πrad);

[0056] N1 and N2 are the number of cycles / teeth in one full rotation;

[0057] P sin_N1 , P cos_N1 , P sin_N2 and P cos_N2 are the current number of cycles of the sensor model calculated according to formula (3) on each curve;

[0058] is the sensor model value at the position ;

[0059] sin() is the sine trigonometric function;

[0060] g vw,sin_N1 (P N1 ), g vw,cos_N1 (P N1 ), g vw,sin_N2 (P N2 ), g vw,cos_N2 (P N2 ) are the amplitudes of each signal, and the amplitude is constant within a given period;

[0061] p vw,sin_N1 (P N1 ), p vw,cos_N1 (P N1 ), p vw,sin_N2 (P N2 ), p vw,cos_N2 (P N2 ) are the phase characteristics of each signal within a given period;

[0062] o vw,sin_N1 (P N1 ), o vw,cos_N1 (P N1 ), o vw,sin_N2 (P N2 ), o vw,cos_N2 (P N2 ) are the averages of each signal, and the average is constant within a given period;

[0063] h is the index of the high-order harmonic, h > 1;

[0064] g vw,sin_N1 (P N1 )[h], g vw,cos_N1 (P N1)[h], g vw,sin_N2 (P N2 )[h], g vw,cos_N2 (P N2 )[h] is the amplitude of each high - order harmonic related to the period;

[0065] p vw,sin_N1 (P N1 )[h], p vw,cos_N1 (P N1 )[h], p vw,sin_N2 (P N2 )[h], p vw,cos_N2 (P N2 )[h] is the phase shift of each high - order harmonic related to the start of the period;

[0066]

[0067] In terms of coefficient update, there are two main methods: One is to process all the data within each period to calculate the average value. However, this method is affected by the rotational speed. If the speed is not constant, the calculated average value will have a deviation, and when calculating the amplitude, since the DC component needs to be removed first, errors will also occur when the rotational speed is uneven. The other is to only process a pair of extreme values (minimum and maximum) in each segment, and calculate the coefficients through the formulas g=(max - min) / 2 and o=(max + min) / 2. However, this method is sensitive to measurement errors. In practical applications, a low - pass filter is often used to eliminate noise, and combined with the detection of the period start point to improve the accuracy of coefficient calculation. The period start point can be determined based on the quadrant and trend of the sensor signal. The new period start point usually occurs when transitioning from the fourth quadrant to the first quadrant. And due to the characteristics of segment division, even using approximate signal average value information can roughly determine the period start point. At the same time, the periodic change of the position tracking curve calculated can also be used to assist in detecting new segments.

[0068] The accuracy of the simplified model is restricted by the degree of change in sensor parameters. When the change in sensor parameters is small, the coefficients of adjacent segments are similar, and the model can approximate the actual curve well. On the contrary, if the parameters change greatly, the limited number of coefficients will reduce the model accuracy. Compared with the actual signal curve of the sensor, there is a certain error in the simplified model because its parameters change in a jump rather than continuously. However, due to the numerical characteristics of the new method, there is a certain tolerance for this inaccuracy. In terms of memory requirements, it has a significant advantage compared to the complete model. The memory required to store its coefficients depends on the number of teeth. For example, for four signals, there is a set of three coefficients (amplitude, average value, phase) for each period. According to the formula the number of coefficients can be calculated, and combined with the data bit resolution, the memory requirements can be estimated (as shown in Table 1). It is much more efficient than the complete model under the same configuration and is more conducive to application in actual systems.

[0069] Memory Requirements of the Simplified Model of the Position Sensor

[0070]

[0071] S2: Similarity Evaluation: Similarity evaluation is used to determine the similarity between the sensor and the model to obtain accurate position information, including two parts: difference calculation and determination of the maximum similarity; there are two methods, the sum of absolute differences and the sum of squared differences. The similarity is determined by calculating the signal difference between the sensor and the model. The former has a global minimum under specific conditions but does not have rotational invariance, while the latter has this property. Its result affects the accuracy of position calculation, provides a basis for screening out the combination closest to the actual position from the model, and collaborates with the sensor model to guide the direction of position calculation.

[0072] The specific steps of the above-mentioned S2 are as follows:

[0073] When calculating the similarity, first calculate some differences based on the four signals of the sensor and the corresponding model signals. Assume that the signal value of the sensor at position is and the signal values of the model at position are sin_N1 vw , cos_N1 vw , sin_N2 vw , cos_N2 vw , then the formula for some differences is:

[0074]

[0075]

[0076] Thus, a set of four partial difference values is obtained. The curves generated by them change with the rotation of the sensor gear, and their average value is determined by the current value of the sensor.

[0077] The similarity is calculated using the method of the sum of absolute differences or the sum of squared differences:

[0078] Sum of Absolute Differences: The similarity calculation is transformed into finding the sum of the distances between the signal values of the sensor and the model. Its difference function At , in the ideal case, the sensor and the model values are the same, so the partial difference value is zero and the sum is zero. And because the absolute value is non - negative, this is a local minimum; also, due to the uniqueness of the combination of the absolute position sensor values, this point is also the global minimum.

[0079] Sum of Squared Differences: Based on the concept of distance in multi - dimensional space, the similarity is calculated using the Pythagorean theorem. Its difference function is similar to the sum of absolute differences. At There is a global minimum here, and it is proved by numerical simulation that this method is invariant to the rotation of the sensor gear and the rotation of the model, has more advantages in position calculation, can provide a more stable and reliable basis for finding the actual position subsequently, plays a key screening and judgment role in the entire new method system, and ensures that the calculated position is closer to the true value.

[0080] S3: Steps for calculating the position: including two aspects of memory usage optimization and computational complexity optimization; in terms of memory optimization, the traditional overall calculation and storage are abandoned, and instead, data is generated and processed step by step to reduce memory occupancy; in terms of computational complexity optimization, a combination of full search and fast search modes is used. A search area is set according to the rotation speed of the mechanism. The full search is used for initialization or anomalies, and the fast search is updated regularly. Each step is closely linked. Based on the sensor model and the similarity evaluation result, the position is calculated efficiently according to the established algorithm flow to ensure the practicality of the new method.

[0081] The specific steps of the memory usage optimization in the above S3 are as follows:

[0082] In terms of memory usage optimization, the traditional calculation method needs to generate the entire sensor model, calculate and store all difference function values, which leads to huge memory requirements and low computational efficiency. The new method abandons this way and instead calculates step by step. In each step, only for a specific rotation angle of the current sensor model The corresponding model value is generated according to formulas (1) and (2), and then using the selected similarity calculation method, the difference similarity is calculated by combining the current measurement value of the sensor and this model value. After the calculation is completed, this similarity is compared with the previously stored minimum difference value. If the newly calculated similarity is smaller, the minimum difference value and the corresponding position information are updated. If not, the current calculated value is discarded, and the model value of the next rotation angle is generated for calculation. During the whole process, only the key information of the current minimum difference value needs to be stored, which greatly reduces the memory occupancy and improves the computational efficiency.

[0083] The specific steps of the computational complexity optimization in the above S3 are as follows:

[0084] In terms of optimizing computational complexity, it is achieved by combining the full-search mode EM and the fast-search mode FM. The full-search mode is used when the system starts or when the position of the mechanism cannot be determined due to certain reasons. In this mode, the entire sensor model needs to be traversed. Although the computational time is relatively long, it can ensure finding the accurate position in the initial stage or abnormal situations, providing a basis for subsequent fast search. During normal operation, the fast-search mode is adopted. Since the rotation speed of the mechanism is limited when running in a closed loop, the maximum angle that the mechanism may rotate within a computational cycle can be determined in advance according to its maximum rotation speed, thereby setting an appropriate search area. In the fast-search mode, the sensor model is only calculated within the set limited search area, significantly shortening the time for calculating the new position, and the computational cycle is T FM , the maximum rotation speed ω max and the size of the search area need to satisfy the formula ω max T FM <the size of the search area to ensure that the position information can be updated in a timely and accurate manner during the rotation of the mechanism. As Figure 2 shown, the two modes are flexibly switched according to the actual situation. For example, when there are situations such as pausing the calculation of the position, updating the sensor model coefficients, or detecting an incorrect calculated position, it will switch from the fast-search mode to the full-search mode to recalculate the position, ensuring the accuracy and reliability of the position calculation.

[0085] S4: Method tolerance and accuracy evaluation: It is achieved by calculating the similarity of the signal values of the sensor at different angles. In the full-search mode, the change of sensor parameters affects the difference function, and the tolerance and the number of teeth are related by analyzing its relationship with the local minimum. The fast-search mode ensures robustness in a suitable search area. The accuracy is better than the traditional method through simulation and is still effective in extreme situations such as signal loss. The resolution also affects the accuracy, comprehensively reflecting the accuracy of the new method in calculating the position under different conditions. It is restricted by the sensor model and the calculation process, and in turn acts on the optimization of the model and steps, prompting the continuous improvement of the new method.

[0086] The specific steps of S4 described above are as follows:

[0087] To determine the tolerance, it is achieved by calculating the similarity of the signal values of the sensor at different angles, using the following formula:

[0088]

[0089] where, and are two angles at the middle length of the adjacent cycles. After calculating the similarity and normalizing it, the difference value of the sensor curve in adjacent cycles can be determined. The tolerance of this method is only related to the number of teeth of the sensor gear, and as the number of teeth decreases, the tolerance shows a decreasing trend.

[0090] Robustness of the full search mode: Changes in sensor parameters alter the difference function, which may lead to changes in the position or value of the global minimum. By analyzing the relationship between adjacent local minima, it is found that an incorrect cycle jump may occur when the sensor parameters change by approximately 4.5% (for sensors with a specific number of teeth). The tolerance can be determined by calculating the similarity of signal values at different angles of the sensor, which is only related to the number of teeth of the sensor, and the tolerance decreases as the number of teeth decreases.

[0091] In the fast search mode, it is also necessary to analyze from the two dimensions of robustness and accuracy. When the search area is set to be smaller than a specific critical value (such as |P1 - P2|), this method is robust because there is usually only one local minimum in this area, and it can naturally "slide" to the correct position during the search process. For the analysis of accuracy, it is mainly completed by means of a large number of numerical simulations. The simulations are based on position sensors with different numbers of teeth, using an ideal sensor model and related sensor mathematical models. The change in sensor parameters is simulated as a two-way movement of the average value on four signals, and a comparison calculation is performed with the traditional cursor method with angle interpolation. In the case of changes in sensor parameters, the accuracy of this method far exceeds that of the traditional method. Even in extreme situations, such as when one signal of the sensor is lost, the new method can still roughly determine the correct position. Moreover, if a coefficient update mechanism is added during the position calculation process, subsequent position calculations can return to normal, further demonstrating the good performance of this method in the fast search mode.

[0092] In addition, the resolution also has a significant impact on the calculation accuracy. In terms of the bit resolution of the input signal, the simulation results show that at least 8 bits are required, and 12 bits can basically meet the accuracy requirements; there is a lower limit for the angle resolution of the sensor model. Below this lower limit, the new method will no longer be applicable. After comprehensively considering various factors and conducting simulations, it is found that the choice of similarity calculation method has a relatively small impact on the position accuracy, while the position calculation accuracy of sensors with fewer teeth is relatively low. The main reason is that the model line segments of these sensors are longer, making it difficult to accurately approximate the amplitude and phase change characteristics. However, under appropriate resolution conditions, the impact of model error on accuracy can usually be ignored.

[0093] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for processing an absolute position sensor signal, characterized in that: The following steps are involved: S1: Construct a simplified model of the sensor: The rotation process is divided according to the number of teeth on the sensor gear; S2: Similarity evaluation: Similarity evaluation is used to determine the similarity between the sensor and the model to obtain accurate location information, including difference calculation and maximum similarity determination; S3: Calculation position step: including memory usage optimization and computational complexity optimization; S4: Method tolerance and accuracy evaluation: This is achieved by calculating the similarity of the sensor signal values ​​at different angles.

2. A method for processing an absolute position sensor signal according to claim 1, characterized in that: The specific steps of S1 are: For a sensor gear with teeth N1 and N2, the signal sin_N1 sw and cos_N1 sw The corresponding segment length is 360 0 / N1, signal sin_N2 sw and cos_N2 sw The corresponding segment length is 360 0 / N2, through this natural division, the sensor signal description is simplified from a complex system of equations to a form as shown in formula (1), where the coefficients appear as a function of the number of periods calculated according to formula (2); in: is the rotation angle of the virtual wheel, ranging from (0°, 360°) or (0rad, 2πrad); N1 and N2 are the number of periods / teeth in one complete rotation; P sin_N1 , P cos_N1 , P sin_N2 and P cos_N2 is the current cycle number of the sensor model on each curve calculated according to formula (3); is in position The sensor model value at ; sin() is the sine trigonometric function; g vw,sin_N1 (P N1 ), g vw,cos_N1 (P N1 ), g vw,sin_N2 (P N2 ), g vw,cos_N2 (P N2 ) is the amplitude of each signal, and the amplitude is constant within a given period; p vw,sin_N1 (P N1 ), p vw,cos_N1 (P N1 ), p vw,sin_N2 (P N2 ), p vw,cos_N2 (P N2 ) is the phase characteristic of each signal in a given period; o vw,sin_N1 (P N1 ), o vw,cos_N1 (P N1 ), o vw,sin_N2 (P N2 ), o vw,cos_N2 (P N2 ) is the average value of each signal, and the average value is constant within a given period; h is the index of higher harmonics, h>1; g vw,sin_N1 (P N1 )[h],g vw,cos_N1 (P N1 )[h],g vw,sin_N2 (P N2 )[h],g vw,cos_N2 (P N2 )[h] is the amplitude of each higher harmonic associated with the period; p vw,sin_N1 (P N1 )[h],p vw,cos_N1 (P N1 )[h],p vw,sin_N2 (P N2 )[h],p vw,cos_N2 (P N2 )[h] is the phase shift of each higher harmonic relative to the start of the cycle; 3. The method for processing an absolute position sensor signal according to claim 1, characterized in that: The specific steps of S2 are: When calculating the similarity, first calculate the partial differences based on the four sensor signals and the corresponding model signals. Suppose the sensor is at position The signal value at Model in position The signal value at is sin_N1 vw 、cos_N1 vw 、sin_N2 vw 、cos_N2 vw , then the partial difference calculation formula is: This results in a set of four partial difference values, each of which generates a curve that changes as the sensor gear rotates, and whose average value is determined by the current value of the sensor; The similarity is calculated using the sum of absolute differences or the sum of squared differences: Sum of absolute differences: The similarity calculation is converted into finding the sum of the distances between the sensor and the model signal values. The difference function exist At , since the sensor and model values ​​are the same in an ideal situation, some of the difference values ​​are zero, the sum is zero, and since the absolute value is non-negative, this is a local minimum; Sum of squared differences: Based on the concept of multidimensional space distance, the Pythagorean theorem is used to calculate similarity, and its difference function Similar to the sum of absolute differences, There is a global minimum at .

4. The method for processing an absolute position sensor signal according to claim 1, characterized in that: The specific steps for optimizing memory usage in S3 are: In order to optimize memory usage, at each step, only the specific rotation angle of the current sensor model is used. Generate the corresponding model value according to formula (1) and (2), and then use the selected similarity calculation method to combine the current sensor measurement value with the model value to calculate the difference similarity. After the calculation is completed, compare this similarity with the previously stored minimum difference value. If the newly calculated similarity is smaller, update the minimum difference value and the corresponding position information. If not, discard the current calculated value and continue to generate the model value of the next rotation angle for calculation.

5. The method for processing an absolute position sensor signal according to claim 1, characterized in that: The specific steps of optimizing the computational complexity in S3 are: In terms of computational complexity optimization, it is achieved by combining the full search mode EM and the fast search mode FM. The full search mode is used when the system is started or when the position of the mechanism cannot be determined for some reason. In this mode, the entire sensor model needs to be traversed. Although the calculation time is long, it can ensure that the accurate position is found in the initial stage or under abnormal conditions, providing a basis for subsequent fast searches; during normal operation, the fast search mode is used. Since the rotation speed of the mechanism is limited when it runs in a closed loop, the maximum angle that the mechanism may rotate within a calculation cycle can be pre-determined according to its maximum rotation speed, thereby setting a suitable search area; in the fast search mode, the sensor model is calculated only within the set limited search area, which significantly shortens the time to calculate the new position. The calculation cycle T FM , maximum rotation speed ω max The formula ω must be satisfied between the size of the search area max T FM <The size of the search area is used to ensure that the position information can be updated promptly and accurately during the rotation of the mechanism.

6. The method for processing an absolute position sensor signal according to claim 1, characterized in that: The specific steps of S4 are: To determine the tolerance, the similarity of the sensor signal values ​​at different angles is calculated using the following formula: in, and The two angles are separated by the middle length of the cycle. After calculating the similarity, the difference value of the sensor curve in adjacent cycles can be determined through normalization.

Images