Marine diesel generator and synchronous generator state evaluation method

Abstract

The invention relates to a marine diesel engine and synchronous generator state evaluation method, which comprises the following steps: forming a state evaluation model through an LSTM network model, carrying out model training by using historical operation data, obtaining diesel engine health evaluation indexes through current operation data and the state evaluation model, comparing the diesel engine health evaluation indexes with reference data, and generating an equipment health report. And evaluating the health degree of the equipment. The real-time diagnosis feedback is adopted to replace the maintenance strategy of afterward maintenance, the situations of excessive maintenance and untimely maintenance are avoided, the transformation of the equipment maintenance mode from preventive maintenance to predictive maintenance is promoted, and the digital transformation of the equipment maintenance support is realized.

Description

Technical Field

[0001] The present invention relates to the field of marine power system monitoring, and in particular to a method for evaluating the status of a marine diesel engine and a synchronous generator based on an LSTM network. Background Art

[0002] Marine diesel engines and synchronous generators are key equipment for ships. Traditional condition monitoring methods mainly rely on regular inspections and manual experience, which makes it difficult to detect potential equipment failures in a timely manner. These methods cannot accurately capture the dynamic changes in equipment operation and cannot handle data with time-series dependencies well. Summary of the invention

[0003] Aiming at the existing problems of marine condition monitoring, a condition assessment method for marine diesel engines and synchronous generators is proposed. By building a full life cycle identification and diagnosis model for diesel generator sets, a health assessment of the overall operation of the equipment can be achieved.

[0004] The technical solution of the present invention is: a method for evaluating the status of a marine diesel engine and a synchronous generator, which specifically comprises the following steps:

[0005] 1) Collect data on multiple key parameters of equipment operation in real time through sensors;

[0006] 2) Preprocess the collected data, including redundant data screening and denoising, and use the convolution smoothing algorithm to filter the data and remove high-frequency noise to ensure that the data input into the model accurately reflects the operating status of the equipment; the collected operating data are: fuel pressure, lubricating oil pressure, lubricating oil temperature, cooling water temperature on the A side, cooling water temperature on the B side, exhaust temperature on the A side, exhaust temperature on the B side, speed, generator voltage, and generator current;

[0007] 3) Based on the preprocessed data, the historical operation data is sent to the LSTM network model for training, and a state assessment model for marine diesel engines and synchronous generators is constructed;

[0008] The input parameters of the marine diesel engine and synchronous generator condition assessment model are fuel pressure, lubricating oil pressure, lubricating oil temperature, A-side cooling water temperature, B-side cooling water temperature, diesel engine speed, generator voltage and generator current, and the output is the sum of the exhaust temperatures on the A and B sides;

[0009] 4) The input parameter data currently collected by the equipment is input into the marine diesel engine and synchronous generator status assessment model, and the model output value is compared with the historical health benchmark data to calculate the health of the equipment. The health value range is 0 to 100% and is used to evaluate the health of the equipment.

[0010] Furthermore, in step 2), the redundant data screening and denoising method is as follows: all the measurement point data at time T are compared with the previous time T-1. If all the measurement point data remain unchanged, the data at time T-1 is considered redundant and deleted.

[0011] Further, in step 2), the convolution smoothing algorithm filters the data to improve the accuracy of the data without changing the signal trend. The smoothing formula of the convolution smoothing algorithm is:

[0012]

[0013] Among them, x k is the kth group of data for the specific smoothing process, w is the window for the smoothing process, x k+i is the i-th data of the k-th group of data in the smoothing window, is the smoothing coefficient, which is obtained by fitting the polynomial using the least squares method. After the filtered data is filtered using the convolution smoothing algorithm, the high-frequency signal disappears while retaining the original characteristics of the data.

[0014] The beneficial effects of the present invention are as follows: the state assessment method of marine diesel engines and synchronous generators of the present invention forms a state assessment model through an LSTM network model, uses historical operation data for model training, obtains diesel engine health assessment indicators through current operation data and the state assessment model, and generates equipment health reports by comparing with benchmark data to assess the health of the equipment. Real-time diagnostic feedback is adopted to replace the maintenance strategy of post-maintenance, avoid excessive maintenance and untimely maintenance, promote the transformation of equipment repair mode from preventive maintenance to predictive maintenance, and realize the digital transformation of equipment maintenance guarantee. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 is a flow chart of the method of the present invention;

[0016] Figure 2 Comparison chart of calculation results of re-inputting part of the training data into the condition assessment model after training the diesel engine condition assessment model;

[0017] Figure 3 This is a comparison chart of the calculation results of reselecting diesel engine operation data at another time to verify the model after training the diesel engine state assessment model. DETAILED DESCRIPTION

[0018] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0019] like Figure 1As shown in the figure, a method for evaluating the status of a marine diesel engine and a synchronous generator includes data collection, data processing, model building and training, and equipment health evaluation, as follows:

[0020] 1. Collect data on multiple key parameters of equipment operation in real time through sensors;

[0021] 2. Preprocess the collected data, including redundant data screening and denoising, and use the convolution smoothing algorithm (SG algorithm) to filter the data and remove high-frequency noise to ensure that the data input into the model accurately reflects the operating status of the equipment; the collected operating data are: fuel pressure, lubricating oil pressure, lubricating oil temperature, A-side cooling water temperature, B-side cooling water temperature, A-side exhaust temperature, B-side exhaust temperature, speed, generator voltage, generator current;

[0022] 3. Based on the preprocessed data, the LSTM network model is used to send the historical operation data to the LSTM network model for training, and a state assessment model for marine diesel engines and synchronous generators is constructed;

[0023] The input parameters of the marine diesel engine and synchronous generator condition assessment model are fuel pressure, lubricating oil pressure, lubricating oil temperature, A-side cooling water temperature, B-side cooling water temperature, diesel engine speed, generator voltage and generator current, and the output is the sum of the exhaust temperatures on the A and B sides;

[0024] 4. Input the input parameter data currently collected by the equipment into the marine diesel engine and synchronous generator status assessment model, compare the model output value with the historical health benchmark data, and calculate the health of the equipment. The health value range is 0 to 100%.

[0025] 5. Evaluate the equipment status according to the health results, generate an evaluation report, and provide corresponding maintenance suggestions. For example, when the health is above 90%, the equipment is in good condition; when the health is between 75% and 90%, minor repairs are recommended; when the health is between 60% and 75%, medium repairs are recommended; when the health is below 60%, major repairs are recommended through integrated interactive display.

[0026] The examples use marine diesel engines and synchronous generators. First, the types of measurement points collected by the model data source marine diesel engines and synchronous generators include fuel pressure, lubricating oil pressure, lubricating oil temperature, A-side cooling water temperature, B-side cooling water temperature, A-side exhaust temperature, B-side exhaust temperature, speed, generator voltage, and generator current. The acquisition frequency is 1s, and a total of 58,000 time point data are collected.

[0027] Then, data screening and data filtering are performed. The method of data screening is to compare all the measurement point data at a certain moment T with the previous moment T-1. If all the measurement point data remain unchanged, it is considered that the data at moment T-1 is redundant and needs to be deleted. Data filtering uses a convolution smoothing algorithm, the full name of which is the Savitzky-Golay algorithm, for smoothing filtering. It is a digital filter that can be applied to a set of data to smooth the data, that is, to improve the accuracy of the data without changing the signal trend. The Savitzky-Golay smoothing formula is:

[0028]

[0029] Among them, x k is the kth group of data for the specific smoothing process, w is the window for the smoothing process, x k+i is the i-th data of the k-th group of data in the smoothing window, is the smoothing coefficient, which is obtained by fitting the polynomial using the least squares method.

[0030] After the filtered data is filtered by SG algorithm, the high-frequency signal disappears while retaining the original characteristics of the data, and the filtering effect is obvious.

[0031] Next, a state assessment model for marine diesel engines and synchronous generators is constructed. According to the operating characteristics of marine diesel engines and synchronous generators, lubrication pressure, cooling water temperature on the A side, cooling water temperature on the B side, exhaust temperature on the A side, exhaust temperature on the B side, diesel engine speed, generator voltage, and generator current are selected as the types of measurement points required to build the model for model training. The model input parameters are lubrication pressure, cooling water temperature on the AB side, diesel engine speed, generator voltage, and generator current, and the output parameter is the sum of the exhaust temperatures on the A and B sides. Due to the lack of output parameter information such as torque, the generator current and voltage are used to represent the diesel engine load information; when the diesel engine is degraded, the thermal efficiency decreases, more energy is discharged out of the cylinder with the exhaust gas, and the exhaust temperature increases, so the sum of the exhaust temperature on the A side and the exhaust temperature on the B side is selected as the diesel engine health assessment indicator.

[0032] The input feature number of the marine diesel engine and synchronous generator condition assessment model is 5, the hidden layer size is 100, the number of hidden layers is 1, the first 50% of the data is used for training, and the second 50% of the data is used for verification.

[0033]

[0034] In the formula Calculate the value for the model, y iis the actual value of the measuring point, and y is the average value of all actual measuring points. When the actual operating state is closer to the healthy state, that is, the smaller the error between the exhaust temperature of the actual operation and the exhaust temperature calculated by the healthy model, the closer R2 is to 1. When the health degree is between 90 and 100, it indicates that the equipment is healthy and in good operating condition. When the health degree is between 75 and 90, the equipment needs to be repaired. It is recommended to check whether the appearance of the whole machine has corrosion, deformation, cracks, etc. At the same time, the key components are maintained. The maintenance parts mainly include: cylinder head, main moving parts, engine body, speed regulation, intake and exhaust system, fuel system, lubrication system, cooling system, compressed air system and monitoring system. When the health degree is between 60 and 75, the equipment needs to be repaired. It is recommended to check whether the appearance of the whole machine has corrosion, deformation, cracks, etc. At the same time, the key components are maintained. The maintenance parts mainly include: cylinder head, engine body, speed regulation, intake and exhaust system, fuel system, lubrication system, cooling system, compressed air system and monitoring system. When the health level is lower than 60, the equipment needs to be overhauled. The content of the overhaul includes the repair scope of the medium repair. At the same time, the performance of the whole equipment needs to be checked and the overhaul shall be carried out by professionals from the HDN after-sales service department.

[0035] Finally, the state assessment model of marine diesel engine and synchronous generator is verified. After training the state assessment model of marine diesel engine and synchronous generator, part of the training data is re-input into the state assessment model. The calculation results are as follows: Figure 2 As shown, the accuracy R2 = 0.9964. Reselect the diesel engine operation data of another period of time to verify the model, and the calculation results are as follows Figure 3 As shown, the accuracy R2 = 0.97. It can be seen that after data screening and filtering, the marine diesel engine and synchronous generator state assessment model based on LSTM has high accuracy and can realize the accurate calculation of diesel engine health.

[0036] The above-mentioned embodiments only express the specific implementation of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the invention patent. It should be pointed out that for ordinary technicians in this field, several modifications and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be based on the attached claims.

Claims

1. A method for evaluating the status of a marine diesel engine and a synchronous generator, characterized in that: The specific steps include: 1) Collect data on multiple key parameters of equipment operation in real time through sensors; 2) Preprocess the collected data, including redundant data screening and denoising, and use the convolution smoothing algorithm to filter the data and remove high-frequency noise to ensure that the data input into the model accurately reflects the operating status of the equipment; the collected operating data are: fuel pressure, lubricating oil pressure, lubricating oil temperature, cooling water temperature on the A side, cooling water temperature on the B side, exhaust temperature on the A side, exhaust temperature on the B side, speed, generator voltage, and generator current; 3) Based on the preprocessed data, the historical operation data is sent to the LSTM network model for training, and a state assessment model for marine diesel engines and synchronous generators is constructed; The input parameters of the marine diesel engine and synchronous generator condition assessment model are fuel pressure, lubricating oil pressure, lubricating oil temperature, A-side cooling water temperature, B-side cooling water temperature, diesel engine speed, generator voltage and generator current, and the output is the sum of the exhaust temperatures on the A and B sides; 4) The input parameter data currently collected by the equipment is input into the marine diesel engine and synchronous generator status assessment model, and the model output value is compared with the historical health benchmark data to calculate the health of the equipment. The health value range is 0 to 100% and is used to evaluate the health of the equipment.

2. The method for evaluating the state of a marine diesel engine and a synchronous generator according to claim 1, characterized in that: The redundant data screening and denoising method in step 2) is as follows: compare all the measurement point data at time T with the previous time T-1. If all the measurement point data remain unchanged, the data at time T-1 is considered redundant and deleted.

3. The method for evaluating the state of a marine diesel engine and a synchronous generator according to claim 2, characterized in that: In step 2), the convolution smoothing algorithm filters the data to improve the accuracy of the data without changing the signal trend. The smoothing formula of the convolution smoothing algorithm is: Among them, x k is the kth group of data for the specific smoothing process, w is the window for the smoothing process, x k+i is the i-th data of the k-th group of data in the smoothing window, is the smoothing coefficient, which is obtained by fitting the polynomial using the least squares method. After the filtered data is filtered using the convolution smoothing algorithm, the high-frequency signal disappears while retaining the original characteristics of the data.

Images