TW201619875A - Maintenance scheduling method and system thereof - Google Patents

Abstract

A maintenance scheduling method and system thereof are provided. The maintenance scheduling method includes: obtaining cost information, record information, and sensing information, according to the cost information, record information, and the sensing information, respectively computing at least one set of candidate maintenance time, expected maintenance cost, and confidence based on the results of a plurality of maintenance procedures; when the candidate maintenance time are a plurality of candidate maintenance time, the scheduled maintenance time is determined according to the expected maintenance costs, the confidence, and the candidate maintenance time.

Description

Method and system for scheduling maintenance of machine

The present disclosure relates to a method and system for scheduling maintenance of a machine, and more particularly to a method and system for cost-oriented maintenance scheduling of a machine.

Since manufacturing is an important economic lifeline of Taiwan, due to the evolution of the times, the amount of process data available today is more and more complicated. Most of the current machine maintenance schedules refer to the maintenance time and historical maintenance records recommended by the equipment provider for part maintenance. However, the machine still has unexpected failures and causes losses, which is the most important factor affecting the company's operational risks. one. Therefore, how to provide a method and system for predicting repair time to reduce the loss of maintenance and production costs caused by unexpected failures is an issue to be solved.

In view of the above problems, the present disclosure proposes a method and system for cost-oriented machine maintenance scheduling, using a cost-oriented mechanism to provide optimal cost-effective asset equipment maintenance scheduling recommendations, and reducing unplanned downtime in the manufacturing process to achieve The best production efficiency.

According to the method for maintaining schedule of a machine according to the present disclosure, the method comprises the following steps: obtaining cost information, recording information and feeling about a part in a machine. Measurement information. Calculating at least one set of candidate maintenance time, expected maintenance cost, and confidence from the results of the plurality of maintenance programs according to the cost information, the record information, and the sensing information, wherein the candidate maintenance time, the foregoing maintenance cost, and the aforementioned confidence are in one-to-one correspondence . When the foregoing candidate maintenance time is a candidate maintenance time, the candidate maintenance time is used as the scheduled maintenance time for the machine. When the foregoing candidate maintenance time is a plurality of candidate maintenance times, the scheduled maintenance time for the machine is obtained from the candidate maintenance times based on the expected maintenance costs and the aforementioned confidence.

A machine maintenance scheduling system according to the present disclosure comprises a data acquisition module, a calculation module and a decision module. The data capture module is used to obtain a cost information, a record information and a sensing information about a part of a machine. The calculation module is configured to calculate at least one set of candidate maintenance time, expected maintenance cost, and confidence from the results of the plurality of maintenance programs according to the cost information, the record information, and the sensing information, wherein the candidate maintenance time, the foregoing expected maintenance cost, and The aforementioned confidence is one-to-one correspondence. The decision module is configured to determine a scheduled maintenance time for the machine based on the candidate maintenance time, the aforementioned maintenance cost, and the aforementioned confidence.

In summary, the present disclosure provides an optimized cost-effective machine maintenance scheduling solution that analyzes the information generated by multiple maintenance modules and then utilizes a cost-oriented mechanism to provide the most cost-effective asset equipment. Maintain scheduling recommendations to reduce unplanned downtime in the manufacturing process to increase production efficiency.

The above description of the disclosure and the following description of the embodiments are intended to demonstrate and explain the spirit and principles of the disclosure, and to provide the disclosure The scope of the patent application is further explained.

1‧‧‧ machine

11‧‧‧ Parts

2‧‧‧Database

3‧‧‧Machine Maintenance Scheduling System

31‧‧‧Data Capture Module

33A~33N‧‧‧ Maintenance procedures

35‧‧‧Computation Module

37‧‧‧Decision module

1 is a system architecture diagram of a machine maintenance schedule according to an embodiment of the present disclosure.

2 is a block diagram of a system function of a machine maintenance schedule according to an embodiment of the present disclosure.

FIG. 3 is a flow chart of a method for maintaining schedule of a machine according to an embodiment of the present disclosure.

Figure 4 is a line graph corresponding to the candidate maintenance time for the expected maintenance cost calculated by the cost-oriented maintenance scheduling method in accordance with an embodiment of the present disclosure.

The detailed features of the disclosure are described in detail below in the embodiments, which are sufficient to enable anyone skilled in the art to understand the technical contents of the disclosure and to implement the invention, and the scope of the disclosure, the scope of the application, and the drawings, Those skilled in the art will readily appreciate the relevant objects of the present disclosure. The following examples are intended to further illustrate the present disclosure, but are not intended to limit the scope of the disclosure.

Please refer to FIG. 1 . FIG. 1 is a system architecture diagram of a machine maintenance schedule according to an embodiment of the present disclosure. As shown in the first figure, the machine 1 has a component 11 and is coupled to the database 2 to store information about the operation of the component 11, such as operational records, sensing information transmitted by the sensor, and the like. The part 11 has a fixed life. For example, if it is a laser light source, the light source emitter may need to be replaced for half a year. However, the light source emitter may also occur unexpectedly within half a year. damage. Therefore, the machine maintenance scheduling system 3 connected to the database 2 retrieves the relevant information, and determines the optimal maintenance time of the part 11 to improve the operation efficiency of the machine 1.

For a detailed structure of the machine maintenance scheduling system 3, please refer to FIG. 2 together. FIG. 2 is a system function block diagram of the machine maintenance schedule according to an embodiment of the present disclosure. As shown in FIG. 2, the system 3 of the machine maintenance schedule of the present disclosure includes a data acquisition module 31, a calculation module 35, a decision module 37, and a plurality of maintenance programs 33A to 33N. The data capture module 31 is configured to obtain cost information, record information, and sensing information in the foregoing database. The plurality of maintenance programs 33A to 33N may respectively calculate a candidate maintenance time based on the cost information, the record information, and the sensing information obtained by the data capture module 31, and may also include a confidence level. The calculation module 35 is configured to calculate a set of candidate maintenance time and expected maintenance cost based on the results of the plurality of maintenance programs 33A to 33N according to the cost information, the record information, and the sensing information obtained by the data capture module 31. And confidence, in which candidate maintenance time, expected maintenance cost and confidence are in one-to-one correspondence. The decision module 37 is configured to determine a scheduled maintenance time based on the aforementioned candidate maintenance time, the aforementioned expected maintenance cost, and the aforementioned confidence.

The cost information is selected from one of a group consisting of a user interface input, a reading mode, and a database system. The record information is selected from one of a group consisting of a reading mode and a database system. The sensing information is selected from one of a group consisting of a reading method, a data capture interface, and a database system. The above data acquisition module obtains cost information, record information and sensing information from the above interface, and the input interface of the machine can be implemented in different ways. Not limited to this.

In addition, the main function of the aforementioned maintenance program is to provide candidate maintenance time or predicted confidence value according to the information of the machine. The maintenance program may be a preventive maintenance program, a conditional maintenance program or a predictive maintenance program, and each program may be different. The model provides asset equipment candidate maintenance time recommendations. The aforementioned model can be implemented by different algorithms, such as a neural network, a support vector machine, and a Cost-Based Predictive Maintenance Model disclosed later in the disclosure. Etc., this disclosure is not limited to this.

With regard to the cost-oriented maintenance model proposed in the present disclosure, the model is established based on minimizing the maintenance cost of the training data, that is, the model minimizes the maintenance cost of the training data and the actual part life, wherein the maintenance cost is calculated. When unintended damage occurs before the maintenance time of the part, each maintenance cost is calculated by adding an original cost of the part plus a loss cost, and dividing by an unexpected service life, where the unintended service life is Calculated based on an opening time of the part and the time the part was unexpectedly damaged. If there is no unanticipated damage before the maintenance time of this part, each maintenance cost is divided by an original cost of the part by an expected service life, which is based on an activation time of the part and this time point. Calculated, and then based on the calculation of the maintenance cost, establish a training rule to establish the maintenance model. More specifically, in one embodiment, maintenance costs The system is based on the following equation: among them For the predicted remaining service life of the part, r is the actual remaining service life of the part, C _p is the original cost of the part, C _u is the loss cost of the unintended damage of the part, t _f is the actual usable life of the part, ( t _f - △ r ) is the expected service life of the part, Δ r is the actual remaining life of the part minus the predicted remaining life of the part ( r - ). If the parameter of the maintenance model is θ , the predicted remaining service life It can be calculated by the function f _θ (.), so each training data can calculate the remaining service life through f _θ (.) And according to the equation (1), the maintenance cost of the data is obtained, and finally the maintenance cost of all the training materials is added to obtain the total maintenance cost. In order to train the model parameters that minimize the total maintenance cost, the total maintenance cost can be The parameter θ is differentially differentiated and the parameter update is performed using the steepest gradient method (steepest decent) until convergence.

Please refer to FIG. 3, which is a flow chart of a method for maintaining schedule of a machine according to an embodiment of the present disclosure. As shown in FIG. 3, in step S301, cost information, record information, and sensing information about the machine are obtained. In step S303, a set of more than one candidate maintenance time, expected maintenance cost, and confidence are calculated based on the cost information, the record information, and the sensing information, and the results of one or more maintenance programs. In step S305, the scheduled maintenance time for the machine is obtained according to the foregoing candidate maintenance time, the aforementioned expected maintenance cost, and the aforementioned confidence.

In step S303, the candidate maintenance time, the expected maintenance cost, and the confidence level are in one-to-one correspondence. The expected maintenance cost is calculated based on the corresponding candidate maintenance time, multiple maintenance costs, and a service life probability distribution of one part of the machine. I hope to get it. The service life probability distribution may be a multinomial distribution, or any other model that can simulate the probability of service life, and the disclosure is not limited thereto.

In calculating the expected maintenance cost, the formula (1) can be used to calculate the service life probability distribution calculation of the part. For example, please refer to FIG. 4 , which is a line graph corresponding to the candidate maintenance time for the expected maintenance cost calculated by the cost-oriented maintenance scheduling method according to an embodiment of the disclosure. As shown in Figure 4, the horizontal axis is the time point and the vertical axis is the estimated maintenance cost that is calculated to be amortized to each day. Assume that the past record information is the service life of 10 parts, the days are 55, 55, 55, 56, 58, 59, 62, 62, 68, and the corresponding polynomial distribution probability is p (55) = 0.3, p (56 ) = 0.1, p (58) = 0.2, p (59) = 0.1, p (62) = 0.2, p (68) = 0.1. Under the assumption of this polynomial distribution, it is assumed that the actual service life of this part can only be 55, 56, 58, 62 and 68, and the probability is 0.3, 0.1, 0.2, 0.1, 0.2 and 0.1, respectively. The original cost of the part is 60,000 yuan, the cost of damage caused by unintended damage is 100,000 yuan. When the candidate maintenance time point is 50 days, the expected maintenance cost is as follows: When the candidate maintenance time is 60 days, if the actual service life is 55, 56 and 58, unintended damage will occur, and if the actual service life is 62 and 68, no unexpected damage will occur, so the maintenance cost is expected. for: When the candidate maintenance time point is 70 days, only unexpected damage will occur, so the expected maintenance cost is: By analogy, the candidate maintenance time point is calculated from 50 days to 70 days, and the maintenance cost expectation value is amortized to each day (as shown in Figure 4).

In addition, the confidence level corresponding to each of the foregoing candidate maintenance times is obtained according to at least one maintenance program that provides candidate maintenance time in the foregoing maintenance procedure. When a maintenance program of the maintenance program belongs to a preventive maintenance module or a conditional maintenance module, the corresponding predicted confidence value is not provided. When one of the maintenance programs of the maintenance program belongs to the predictive maintenance module, the confidence corresponding to the candidate maintenance time is selected from a maintenance confidence program provided by each maintenance program of the foregoing maintenance program and the aforementioned maintenance program. The ratio of candidate maintenance time selected in these maintenance procedures is selected, and the latter two are taken as large. More specifically, in an embodiment, the confidence corresponding to the candidate maintenance time is based on the following equation: Where C ( t ) is a confidence equation, C ( t ) is the confidence corresponding to the candidate maintenance time t , S ( t ) is the set of predicted confidence values corresponding to the candidate maintenance time t , and NP is a certain preventive maintenance model. The group or conditional maintenance module provides candidate maintenance time t , but does not provide predictive confidence values. NA indicates that a predictive maintenance module provides a candidate maintenance time t , but this does not provide a predictive confidence value. Max( S ( t )- NA ) is to find the prediction confidence value with the largest set of prediction confidence values, ∥ S ( t )- NP ∥ is the number of elements of the prediction confidence value set S ( t ) at the candidate maintenance time t . The ratio of candidate maintenance time t to the candidate maintenance time of all predictive maintenance scheduling modules, that is, the degree of approval of each candidate maintenance module to support this candidate maintenance time t , the more predictive maintenance module selection candidate maintenance At time t , the higher the value. In other words, the confidence level of the candidate maintenance time t is that both the prediction confidence value of the maximum confidence value set and the approval degree of the candidate maintenance time t are both large. In summary, even if the maintenance module does not provide the predicted confidence value, as long as at least one maintenance module that selects the candidate maintenance time t is a predictive maintenance module, the confidence can still be predicted by the candidate maintenance time t. The ratio of the candidate maintenance time of the maintenance module is calculated based on the approval of each maintenance module to support this candidate maintenance time t .

For example, if the set of confidence values provided by the maintenance module is S (September 1st) = {NP, NP}, S (September 5th) = {NA, 0.8}, S (September 12) ={NA}, through equation (2), in the candidate maintenance time September 1st, because the maintenance scheduling module does not provide predictive confidence value for the preventive maintenance module or conditional maintenance module, so C (September 1) = NP. In the candidate maintenance time September 5, there are two maintenance modules predicted and one maintenance module provides a confidence value of 0.8, so C (September 5) = max (0.8, 2 / (2 + 1)) =0.8. In the candidate maintenance time September 12, there is only one maintenance module prediction, so C (September 12) = max (-, 1 / (2 + 1)) = 0.33.

Next, in step S305, the decision module of the machine maintenance scheduling system of the present disclosure determines the scheduled maintenance time for the machine from the candidate maintenance time. In an embodiment, the decision module selects the corresponding one from the candidate maintenance time. The confidence level exceeds at least one candidate maintenance time of a confidence threshold, and the candidate maintenance time with the lowest expected maintenance cost is selected as the scheduled maintenance time from the candidate maintenance time in which the confidence exceeds the confidence threshold. .

For example, if the machine maintenance scheduling system is based on the information provided by the aforementioned maintenance module, the candidate maintenance time is September 1st, September 5th, September 10th, September 12th, 9th. On the 15th of the month, the corresponding confidence levels are 0.1, 0.6, 0.7, 0.2, 0.3, and the expected maintenance costs are 1, 4, 3, 5, 7 (ten thousand yuan) respectively. If the confidence threshold is 0.5, the decision is made. The module judges each confidence one by one. The candidate maintenance time of more than 0.5 is September 5 (0.6) and September 10 (0.7), and the lowest expected maintenance cost is September 10 (30,000 yuan). Therefore, the final scheduled maintenance time is September 10.

In another embodiment, the decision module uses the confidence level as a plurality of weights from the maintenance, and the corresponding candidate maintenance time weighted average is used as the scheduled maintenance time.

For example, if the machine maintenance scheduling system is based on the information provided by the maintenance module, the candidate maintenance time is September 1, September 5, September 10, September 12, September. On the 15th, and their corresponding confidence levels are 0.1, 0.6, 0.7, 0.2, 0.3, the expected maintenance costs are 1, 4, 3, 5, 7 (ten thousand yuan), assuming the current date is August 16th. The candidate maintenance time is 15 days for the parts remaining on September 1, 19 days for September 5, 21 days for September 10, 23 days for September 12, and 26 days for September 15. The weights of these five candidate maintenance times are

September 1st: (0.1/(0.1+0.6+0.7+0.2+0.3))=5.3%

September 5th: (0.6/(0.1+0.6+0.7+0.2+0.3))=31.6%

September 10th: (0.7/(0.1+0.6+0.7+0.2+0.3))=36.8%

September 12th: (0.2/(0.1+0.6+0.7+0.2+0.3))=10.5%

September 15th: (0.3/(0.1+0.6+0.7+0.2+0.3))=15.8%

After the weighted average calculation, since (15*5.3%+19*31.6%+21*36.8%+23*10.5%+26*15.8%)=21.05, the remaining life that can be obtained is 21 days, and then push Estimated scheduled maintenance time is September 10.

In summary, the present disclosure provides an optimized cost-effective machine maintenance scheduling solution that analyzes the information generated by multiple maintenance modules to develop a part health assessment and fault pre-diagnosis prediction model. Use cost-oriented mechanisms to provide optimal cost-effective asset equipment maintenance scheduling recommendations and reduce unplanned downtime in the manufacturing process to increase production efficiency.

Although the disclosure is disclosed above in the foregoing embodiments, it is not intended to limit the disclosure. All changes and refinements are beyond the scope of this disclosure. Please refer to the attached patent application for the scope of protection defined by this disclosure.

1‧‧‧ machine

11‧‧‧ Parts

2‧‧‧Database

3‧‧‧Machine Maintenance Scheduling System

Claims (25)

A method for maintaining a schedule of a machine includes: obtaining a cost information, a record information, and a sensing information about a part of a machine; and based on the cost information, the record information, and the sensing information, As a result of the maintenance process, at least one candidate maintenance time, expected maintenance cost, and confidence are calculated, wherein the candidate maintenance time, the expected maintenance cost, and the confidence level are in one-to-one correspondence; when the at least one candidate maintenance time is a candidate maintenance time Taking the candidate maintenance time as a scheduled maintenance time for the machine; and when the at least one candidate maintenance time is a plurality of candidate maintenance times, according to the expected maintenance cost and the at least one confidence degree, The scheduled maintenance time is obtained for the machine in the candidate maintenance time. The method of claim 1, wherein the cost information is selected from one of a group consisting of a user interface input, a reading mode, and a database system. The method of claim 1, wherein the record information is selected from one of a group consisting of a read mode and a database system. The method of claim 1, wherein the sensing information is selected from one of a group consisting of a reading mode, a data capture interface, and a database system. The method of claim 1, wherein the maintenance program includes a cost guide To the maintenance program, based on the sensing information of the machine and/or a health indicator of the part, aiming at minimizing maintenance cost, a cost-oriented maintenance model is established to calculate the corresponding candidate maintenance time. The method of claim 1, wherein each of the expected maintenance costs of the at least one expected maintenance cost is based on the corresponding candidate maintenance time, the plurality of expected maintenance costs, and a service life probability distribution of the part of the machine. Calculated by calculating the expected value. The method of claim 6, wherein each of the maintenance costs of the maintenance costs is an original cost of the part plus a loss cost when it is determined that the part will be unintendedly damaged before the candidate maintenance time. And dividing by an unexpected service life, wherein the unintended service life is calculated based on an activation time of the part and a time when the part is unintendedly damaged. The method of claim 6, wherein, when it is determined that the part does not undergo unintended damage before the candidate maintenance time, each of the maintenance costs of the maintenance costs is divided by an original cost of the part by an intended use. Lifetime, wherein the expected service life is calculated based on an activation time of the part and the candidate maintenance time. The method of claim 1, wherein the confidence corresponding to each of the candidate maintenance times is obtained according to the at least one maintenance program that provides the candidate maintenance time in the maintenance procedures. The method of claim 9, wherein at least one of the at least one maintenance program providing the candidate maintenance time is a predictive maintenance module corresponding to the The confidence level of the candidate maintenance time is selected from the ratio of the largest of the predicted confidence values provided by each of the at least one maintenance program to the maintenance programs selected for the candidate maintenance time among the maintenance programs, Take it big. The method of claim 1, wherein the step of obtaining the scheduled maintenance time from the candidate maintenance time according to the expected maintenance cost and the at least one confidence level is performed from the candidate maintenance The at least one candidate maintenance time in which the confidence level exceeds a confidence threshold is selected in the time, and the candidate maintenance time with the lowest expected maintenance cost is selected from the at least one candidate maintenance time. This schedules maintenance time. The method of claim 1, wherein the step of obtaining the scheduled maintenance time for the machine from the candidate maintenance times is performed according to the expected maintenance cost and the at least one confidence level, and the confidence level is used The scheduled maintenance time is calculated by using a weighted average method for the plurality of weights and the corresponding candidate maintenance times. A machine maintenance scheduling system includes: a data acquisition module for obtaining a cost information, a record information and a sensing information about a part of a machine; a calculation module for The cost information, the record information, and the sensing information are calculated by the results of the plurality of maintenance programs to calculate at least one set of candidate maintenance time, expected maintenance cost, and confidence, wherein the candidate maintenance time, the expected maintenance cost, and the confidence level are a correspondence; and a decision module for maintaining time and expectations based on the at least one set of candidates Maintenance costs and confidence determine a scheduled maintenance time for the machine. The system of claim 13, wherein the data capture module is selected from one of a group consisting of a user interface input, a read mode, and a database system. The system of claim 13, wherein the data capture module is selected from one of a group consisting of a read mode and a database system to obtain the record information. The system of claim 13, wherein the data capture module is selected from one of a group consisting of a read mode, a data capture interface, and a database system to obtain the sensing information. The system of claim 13, wherein the maintenance program includes a cost-oriented maintenance program, which is based on the sensing information of the machine and/or a health indicator of the part, and aims to minimize maintenance cost. A cost-oriented maintenance model is established to calculate the corresponding candidate maintenance time. The system of claim 13, wherein each of the expected maintenance costs of the at least one expected maintenance cost is based on the corresponding candidate maintenance time, the plurality of maintenance costs, and a use of the part of the machine. The life probability distribution is obtained by calculating the expected value. The system of claim 18, wherein when the computing module determines that the part will experience unanticipated damage before the candidate maintenance time, each of the maintenance costs of the maintenance costs is increased by an original cost of the part. The last lost cost is divided by an unexpected service life, wherein the unintended service life is calculated based on an activation time of the part and a time when the part is unexpectedly damaged. The system of claim 18, wherein each maintenance cost of the maintenance costs is divided by an original cost of the part when the computing module determines that the part does not experience unanticipated damage before the candidate maintenance time At an expected service life, wherein the expected service life is calculated based on an activation time of the part and the candidate maintenance time. The system of claim 13, wherein the confidence level corresponding to each of the candidate maintenance times is obtained by the at least one maintenance program that provides the candidate maintenance time according to the maintenance programs. The system of claim 21, wherein at least one of the at least one maintenance program providing the candidate maintenance time is a predictive maintenance module, the confidence corresponding to the candidate maintenance time is selected from the at least one The ratio of the largest of the predicted confidence values provided by each of the maintenance programs in the maintenance program to the maintenance programs selected for the candidate maintenance time in the maintenance programs, which are both large. The system of claim 13, wherein the at least one candidate maintenance time is a candidate maintenance time when the decision module determines the scheduled maintenance time according to the at least one candidate maintenance time, the expected maintenance cost, and the confidence level. The decision module uses the candidate maintenance time as the scheduled maintenance time for the machine. If the at least one candidate maintenance time is a plurality of candidate maintenance times, the decision module is based on the expected maintenance costs and the At least one confidence level, the scheduled maintenance time for the machine is obtained from the candidate maintenance times. The system of claim 23, wherein the decision module selects from the candidate maintenance times when the at least one candidate maintenance time is a candidate maintenance time And selecting the at least one candidate maintenance time that the confidence level exceeds a confidence threshold, and selecting, from the at least one candidate maintenance time, the candidate maintenance time with the lowest expected maintenance cost as the scheduled maintenance time. The system of claim 23, wherein the step of obtaining the scheduled maintenance time for the machine from the candidate maintenance times is performed according to the expected maintenance cost and the at least one confidence level, and the confidence level is used The scheduled maintenance time is calculated by using a weighted average method for the plurality of weights and the corresponding candidate maintenance times.